9. Oscilloscopes
When I started the LHC project it soon became apparent I would have to upgrade the equipment in my lab, slowly collected over more than 30 years. I recognised I would probably need to measure short optical and electrical pulses with sub-nanosecond rise times and very high voltages, particularly for the laser Q-switch driver circuitry. A key measurement tool in any electronics lab is an oscilloscope.
My fastest oscilloscope was a rare Tek 2455B 4-channel 250MHz non-storage analogue scope with a dodgy channel and a 1.4ns rise time but it could not see fast single pulses. I also had an excellent Rigol DS1102CD 2-channel 100MHz 0.4GS/s digital Storage Oscilloscope (DSO) [I13] which could see 2ns single pulses, but its 3.5ns rise time renders the view meaningless. To paraphrase Roy Scheider [G24]:
'You're gonna need a bigger scope', and a faster DSO was my ideal choice,
Not to scale: Tek 2455B Rigol DS1102CD
CHOOSING A DSO
https://www.tek.com/en/blog/which-oscilloscope-to-buy
BANDWIDTH
[E57] ETS Current Probes Page 42:
'Bandwidth is the maximum frequency of an input signal which can pass through the analogue front end of the scope with minimal amplitude loss (from the tip of the probe to the input of the oscilloscope ADC). It is specified as the frequency at which a sinusoidal input signal is attenuated to 70.7% of its original amplitude (the -3 dB point).'
SAMPLE RATE
https://www.tek.com/en/documents/application-note/real-time-versus-equivalent-time-sampling
[Q4, Q4e] [I23] Tek Primer - Fundamentals of Digital Signal Integrity P.13 sampling 5x rule:
'To determine the oscilloscope bandwidth needed to accurately characterize signal amplitude for a specific application, the “5 Times Rule” is useful: An oscilloscope selected using the 5 Times Rule will have less than a ±2% measurement error. In general, a higher bandwidth will provide a more accurate reproduction of the signal of interest:
Oscilloscope bandwidth ≥ highest frequency component of signal x5'
However a close approximation can be obtained with a minimum of x3:
From [E58] - Bandwidth vs Sample Rate – Electronic Product Design webpage:
'Sample rate is the maximum number of samples the scope can take per second and will usually be across all of the channels. Generally you are interested in the "real time" sample rate and not a repetitive signal sample rate.
To capture the true shape of the signal, you need a scope with a bandwidth large enough to capture several of the signals harmonics, so ideally use a scope with 3x to 5x the bandwidth you calculated for your signal.
For general analogue signals use a scope with a bandwidth specification at least 2x the highest frequency component in the signal.
For digital signals where you care about seeing what noise is occurring – determine the fastest signal rise/fall time you are interested in and work out the bandwidth using the formula
Bandwidth = 0.35 / Rise time.' [Q2a]
So to see a pulse with a 1ns rise time, bandwidth = 0.35 / 1ns = 350MHz. [Q2a]
But for more accuracy, aim for 0.4 / rise time = 400MHz, see HP's worked example: [Q4b]
A 1ns rise time is desirable but necessitates at least a 350MHz bandwidth and from [Q4e], achieving that on all 4 channels would require a sample rate of 350MHz x5 x4 = 7GS/s, beyond my budget.
Another issue with DSOs is the very reason they were invented: apart from the convenience of storage and instant measurements, they are cheaper to manufacture. However IMO they are less reliable than their analogue predecessors, particularly the modern approach which is to embed a standard PC running Windows,
often with a HDD. I have seen DSOs of this type on eBay with the familiar DOS prompt failing to recognise the broken drive, rendering the hardware of the scope useless.
There are a number of DSO manufacturers but generally I've stuck with the two most predominant - Tektronix and HP/Agilent as they have manufactured the most and there are bargains to be had as a result. LeCroy DSOs often come up relatively cheaply on eBay but I've been lead to believe they have reliability issues, and I don't like their kludge of adding external hardware to maximise channel sampling. I thought Japanese company Iwatsu was good until I read 'Iwatsu was OEM for some upper low end (WaveJet) and midrange (WaveRunner LT) LeCroy oscilloscopes (the WaveJets are rebadged Iwatsu scopes, the Waverunners have been developed together by LeCroy and Iwatsu for LeCroy).'
https://www.eevblog.com/forum/testgear/iwatsu-ts-series-oscilloscopes/
Regardless, the higher the bandwidth and sample rate, the more complex the electronics and the need for bespoke gate arrays with fine pitch surface mount packages and tiny gaps between pins, often running at elevated temperatures causing issues not only with the chips, but the accompanying surface mount supply bulk decoupling capacitors. Tek's TDS-series DSOs are infamous for surface mount electrolytic capacitors drying out on the acquisition board (a relatively easy fix, see this great video:
http://www.youtube.com/watch?v=X8N9O3a9jiM ) but if anything else goes in such a complex beast, it's usually a case of chucking the whole thing in the bin. Generally the older the equipment, the simpler the components and the easier for an enthusiast to replace, countered by a less friendly interface.
If it were not for the high probability of catastrophic failure of high end second hand DSOs due their age (age being the thing making them affordable), I would have sought out a Tek TDS694C 3GHz 10GS/s 120kpts [I29] (infamous for a heat-related trigger failure on channels 3 & 4), or the better
HP54854A 4GHz 20GS/s 262kpts [I24] which similarly runs Windows. Even the 1GHz probes that came with the TDS640 had a habit of overheating and failing. Lastly, their sheer size and weight put me off. My lab is small, with no room for such beasts. That said, many modern smaller DSOs like the Tek TDS3064B DPO 600MHz 5GS/s have an Achilles Heel of a tiny 10kpts acquisition memory, and whilst single channel performance is outstanding, none of the above DSOs can support more than 1-channel at full bandwidth.
Not to scale: Tek TDS694C HP54854A Tek TDS3064B
FAST ANALOGUE OSCILLOSCOPES ARE CHEAPER THAN EQUIVALENT DSOs
Due to the above I abandoned the immediate plan to upgrade my Rigol DS1102CD DSO, and instead
looked to the past for an affordable high speed alternative.
CALIBRATION
Accuracy of measurement brings another problem - calibration. I cannot afford to have my equipment professionally repaired or calibrated, so I have made a point of buying equipment I can service and calibrate myself, see [References: Calibration], which means it must have documentation including schematics. Fortunately such equipment exists from the past era when instruments were built to last and are now very affordable, in particular Tektronix 7000 series analogue oscilloscopes and indeed Tek's range of test equipment from that age. http://www.tek.com/Measurement/Support/scopes/faq/history.html
Here's a huge list of Tektronix part numbers together with a brief description of each entry:
http://www.davmar.org/temodel.html#tek
Here's a list with rare Tek manuals (and others), and instrument modification documents:
http://www.hakanh.com/dl/kits.htm
FAST ANALOGUE OSCILLOCOPES
Tek's fastest analogue mainframe was the 350ps 7104 and its micro-channel plate (MCP) accelerator meant it could trigger on such events but it had to be used sparingly so even if I got one, I could not use it all the time. It had no storage but I could probably capture the CRT with a digital camera. What put me off was finding one with a good MCP, but they were very expensive even on eBay.
I found a German website describing the combination of a 400MHz Tek 7834 storage mainframe and 7S14 (100ps/div) sampler as capable of seeing a 1ns pulse out of a Tek 109 mercury switch pulse generator at its nominal 640Hz; the site also revealed the 7S14 has hidden batteries powering its sampler circuits, that can give the impression the unit is faulty:
http://www.amplifier.cd/Test_Equipment/Tektronix/Tektronix_other/109.html
http://www.amplifier.cd/Test_Equipment/Tektronix/Tektronix_7000_series_special/sampler_7S14.htm
http://www.amplifier.cd/Test_Equipment/Tektronix/Tektronix_7000_series_mainframe/7834.htm
'Viewing the curve with a 7S14 sampler was very easy, the plug-in has a good trigger capability'.
Tekwiki has a brief description of the 7S14, together with further details on the hidden batteries:
http://w140.com/tekwiki/wiki/7S14
http://w140.com/tekwiki/wiki/7S14_repair
Looking on eBay I found several 7S14s going for around $200 and 7834s with plugins for a similar price, mainly I think because it's quite ancient and buyers tend to go for its 7934 500MHz successor. The 7S14 samples at around 20MHz but based on the website above I decided if I built my high speed circuits to run repetitively at least as fast as the 109 (around 750Hz) for debug, I could use it, so I bought one for around $200 including plugins and a 'working' 7S14 for around $90.
Left, the 7834
Right, the 7S14 with four AG10 batteries in holders I added to the input sampler chambers.
The 7834 turned out to be a beautiful beast, but the 7S14 was a dud even after replacing its batteries. I managed to get a refund and bought another 'tested' but that too was faulty and it was too expensive to return it, see [Repairs: Tek 7S14 Repairs]. Later, another working 7S14 appeared on eBay Germany starting at $50, but it went for over $500 and at this point I decided to look for another solution. I still needed something that could capture single shot events, and my research moved to the Tek 7912AD digital transient recorder.
Footnote
I had assumed from the German site the 7834 needed something like the 7S14 to see the 109. In 2022 when I had my own Tek 109 by then modified to produce 1Hz pulses [Projects: Tek 109 1-Shot & LF Mod] I found the 7834 alone is capable of triggering and displaying those 1Hz pulses using its variable persistence storage mode, and a 7A29 produced much the same waveform as on a 7104; even a 7A26 wasn't that bad.
TRANSIENT DIGITISERS
The Tek 7912AD programmable digitiser is capable of capturing a single pulse with a 500ps rise time and has a claimed equivalent sample rate of 100GS/s: http://w140.com/tekwiki/wiki/7912
By chance one appeared on UK eBay for £80 and I bought it, the description: 'Broken switches on front panel, easy fix'. We know what that means, don't we? I plugged it in and slowly brought up its voltage on the lab mains variac. The fan started hesitatingly, and when I pressed the ON/OFF button lights appeared but there was no video output, and then there was smoke...see [Repairs: Tek 7912AD Repairs]!
Below is a table I drew up, comparing the rise time of various oscilloscopes corresponding to their bandwidth in MHz, and from the thesis curve (below), the signal degradation I could expect from the 7912AD (see [C4], [C5], [T12] for dB conversions). Below, the Tek 7912AD
I had bought a DVD service manual for the 7912AD and during the process of teardown as I looked on the web for reports from other owners with similar problems, I found more documents describing its capabilities. One of these was a thesis at Lockheed Martin which described the operation of its scan converter tube in detail, and modifications to increase the bandwidth to 3.5GHz. On page 79 was a very revealing curve of actual measured bandwidth, which surprised me how far it fell short of Tek's description of the 7912AD. Below is a table I drew up, comparing the rise time of various oscilloscopes corresponding to their bandwidth in MHz, and from the thesis curve (below), the signal degradation I could expect from the 7912AD at these frequencies (see [C5], [T12] for dB conversions):
Other oscilloscope bandwidths vs 7912AD rise time degradation:
Trise Bndwth Model to compare Bandwidth Loss Percent with 7912AD + 7A29 or 7A21N
3500ps = 100MHz Rigol DS1102CD ( 100MHz) - 0.0dB:100.0% -
2330ps = 150MHz ( 150MHz) - 0.0dB:100.0% -
1750ps = 200MHz ( 200MHz) - 0.2dB: 95.5% 10V in->9,6V, 5V->4.78V (7A29)
1400ps = 250MHz Tek 2455B ( 250MHz) - 0.5dB: 89.0% 10V in->8.9V, 5V->4.45V (7A29)
1000ps = 350MHz ( 350MHz) - 0.9dB: 81.0% 10V in->8.1V, 5V->4.05V (7A29)
875ps = 0.400GHz Tek 2465B/67B 7834 7854 ( 400MHz) - 1.2dB: 75.9% 10V in->7.6V, 5V->3.80V (7A29)
800ps = 0.435GHz min to measure 4ns Trise ( 435MHz) - 1.3dB: 74.1% 10V in->7.4V, 5V->3.71V (7A29)
700ps = 0.500GHz (7912AD & 7A29) ( 500MHz) - 2.0dB: 63.0% 10V in->6.3V, 5V->3.15V (7A29)
583ps = 0.600GHz Iwatsu TS-80600 ( 600MHz) - 3.0dB: 50.0% 10V in->5.0V, 5V->2.50V (7A29)
500ps = 0.700GHz ( 700MHz) - 5.0dB: 32.0% 10V in->3.2V, 5V->1.60V (7A29)
467ps = 0.750GHz ( 750MHz) - 6.0dB: 25.0% 10V in->2.5V, 5V->1.25V (7A29)
389ps = 0.900GHz ( 900MHz) - 9.0dB: 12.0% 10V in->1.2V, 5V->0.63V (7A29)
350ps = 1.000GHz Tek 7104/SCD1000/TS-81000 (1000MHz) -12.0dB: 6.3% 10V in->0.6V, 5V->0.32V (7A21N)
Rise times for faster DSOs covered in this section:
175ps = 2.000GHz Rigol MSO8204 (2000MHz)
100ps = 3.500GHz Tek SCD5000TD (3500MHz) Transient Digitiser
70ps = 5.000GHz (5000MHz)
62ps = 6.000GHz Agilent 54855A 6GHz (6000MHz) Option 008
58ps = 6.000GHz Tek 7250 (6000MHz) Transient Digitiser
[W3] The Iwatsu TS-81000 (TS80600=600MHz) is a 1GHz scope with a CCD scan converter & LCD screen:
https://www.iti.iwatsu.co.jp/en/products/ss/ts81000/ts_spec_e.html
Iwatsu was OEM for some upper low end (WaveJet) and midrange (WaveRunner LT) LeCroy oscilloscopes (the WaveJets are rebadged Iwatsu scopes, the Waverunners have been developed together by LeCroy and Iwatsu for LeCroy) https://www.eevblog.com/forum/testgear/iwatsu-ts-series-oscilloscopes/
The 7912AD seems a great choice for capturing sub-ns single shot transients but the following curve, translated into the equivalent dB losses shown in my table above right, suggests its specification is not as good as implied by Tek (scribblings from my LHC project notebook):
At first I was taken in by Tek's assertion the 7912AD provides an equivalent sampling rate of 100GS/s but it soon became evident from this article and others, that this is not the case.
Furthermore to attain the claimed 1GHz response it is necessary to rewire the 7912AD circuitry to accommodate the 7A21N input plugin, rendering any other plugin unusable and even then, it barely comes up to scratch.
The reality is at -3dB with its companion 7B90P timebase & 7A21N '1GHz' plugin it's a 600MHz digitiser, and only 500MHz with a 7A29.
This curve was taken from page 79/380 of the following (Lockheed) document (34MB):
http://www.scottpages.net/MESCthesis.pdf
The 7104 oscilloscope far surpasses the 7912AD (see curve to left):
7912AD:
~750MHz -6dB: 10V in->2.5V, 5V->1.25V (7A29)
~900MHz -9dB: 10V in->1.2V, 5V->0.60V (7A29)
~1000MHz -12dB: 10V in->0.63V passive 7N21
7104:
1000MHz -2dB: 10V in->6.3V, 5V->3.15V (7A29)
Tek's later 1GHz SCD1000 and 3.5GHz SCD5000 digitisers appear markedly superior to the 7912AD. They still use a scan converter tube but also provide a PC VGA display output as well an optional front panel with LCD screen
and manual input. In both cases tekwiki provides more detail, including internal photos for the SCD1000:
http://w140.com/tekwiki/wiki/SCD1000
http://w140.com/tekwiki/wiki/SCD5000
Tek's later 6GHz 7250 is better still, but is huge and heavy and like the SCDs, I could find no service manual with schematics:
PROBES
At this point it is worth mentioning probes; probes affect rise time measurement. A comparison of high speed and high voltage oscilloscope probes can be found in [Research: High Speed kV Probe].
A good rule of thumb is to aim for probe bandwidth having 1.5x the maximum intended frequency of use. Probe rise time is also an issue when investigating non-repetitive high speed signals, and watch out for probes that only work with specific models of oscilloscopes, such as Tek's PP series.
I have collected a few good voltage probes over the years, but the following chart I found in a Tek catalogue (unfortunately I can no longer find which it came from, but page 219 of the 1987 Tek General Catalogue from [I31] is very close), reveals some bare truths about their probes' claimed performance vs reality when paired with plugins and mainframes.
For instance, the P6201 900MHz high impedance probe is only rated up to 660MHz with the fastest listed combination of a 1GHz 7A29 and 7104 mainframe. When exactly does it work up to 900MHz?
HIGH SPEED TIMING REFERENCE
As I was looking for an affordable way of recording high speed waveforms, I was also considering how to verify the rise time of my pockels cell driver without it being swamped by oscilloscope rise time, or clobbered by a 50Ω load. I began to look for an affordable, similarly high speed pulse generator to use as a timing reference and researched Tek's range of yesteryear instruments.
First, I found the Type 109 mercury (Hg) switch pulse generator which uses a combination of a mercury wetted reed relay, biasing magnet and length of coaxial cable to produce a short pulse with a sub 250ps rise time. I thought it would be a good idea if I had such a generator as a reference but even if I had a source of its obsolete reed relay, running at ~750Hz gives it a very limited mechanical life, although it is capable of supporting pulses up to ±300V. (I later acquired one of these and added low frequency and single pulse capability to it, see [Projects: Tek 109 1-Shot & LF Mod] ).
The newer 284, often called up in Tek service manuals, uses a tunnel diode. However it rarely comes up in auction and when it does it's usually expensive, and what if it went the same way as the 7S14?
I also considered building my own tunnel diode pulse generator, but I realised the chances of getting a sub-ns rise time would be pretty small without RF design knowledge and I still had nothing that could see it, particularly as TD generators produce pulse amplitudes in fractions of a volt.
TIME DOMAIN REFLECTOMETER
Investigating fast rise time pulse generators took me to the 7S12 Time Domain Reflectometer (TDR) plugin that I had ignored up until now. I discovered it could measure cable integrity by injecting a 25ps rise time pulse and capturing the return signal with an equally fast sampling head. I also found it takes a trigger plugin that turns it into a general purpose 1GHz sampler, and the timebase goes down to 20ps/div! I noticed 7S12s came up on eBay USA quite often and despite the impressive specs they were not that expensive. I realised I should have got this instead of the 7S14, although the German site that used the 7S14 said the TDR could not see the 109 pulse due to its lower speed sampling interval. I reasoned I could get around this by triggering a HV pulse generator from the TDR pulse generator head.
The 7S12 plugin has itself two plugins, usually a sampler on the left and a trigger or pulse generator on the right. Below is a list of several heads with specifications gleaned from Tekwiki and manuals:
S-1 SAMPLING HEAD
http://w140.com/tekwiki/wiki/S-1
Rise time: 350ps
Bandwidth: 1GHz
Input impedance: 50Ω (GR-874)
Noise: <2mV
Maximum signal voltage: 1Vpp
Maximum signal voltage: ±1Vpp
Features
A trigger pickoff is provided for internal triggering.
S-2 SAMPLING HEAD
http://w140.com/tekwiki/wiki/S-2
Rise time: 75ps
Bandwidth: 4.66GHz
Input impedance: 50Ω (GR-874)
Noise: <10mV
Maximum signal voltage: ±1V
Maximum input voltage: ±5V
S-3/S-3A SAMPLING HEAD
http://w140.com/tekwiki/wiki/S-3A
Input rise time (x1): <350ps (> with attenuator tips)
Bandwidth: 1GHz
Input impedance (BNC): 100kΩ || 2.3pF at LF
10x: 1MΩ || 2.0pF / 100x 1.7pF
Noise: <3mV
Maximum signal voltage:
1x (no attenuator): x1 switch ± 1V, x2 switch ± 2V
10x attenuator: x1 switch ± 10V, x2 switch ± 20V
100x attenuator: x1 switch ±100V, x2 switch ±200V
Maximum input voltage:
1x (no attenuator): ±100Vpk, ± 20V sine
10x / 100x attenuator: ±350Vpk, ±200V sine
http://groups.io/g/TekScopes/topic/7s_7t_sampling_system/948674?p=Created,,,20,0,0,0
Alternative fix for S3a 1GHz sampling probe bridge failure: 'I found that the basic spec, most critically reverse leakage and capacitance were matched rather well with Agilent microwave Schottkys HSMS8202. These are diode pairs in a single SOT23 surface mount package. By flipping one of them over, and soldering the pair of pins one edge together you end up with a ~$3 bridge, full 1GHz bandwidth and aberrations well within spec. I suspect that this approach would work well with all the earlier samplers up to around 4GHz, including the S1 and S2.'
S-4 SAMPLING HEAD
http://w140.com/tekwiki/wiki/S-4
Rise time: 25ps (observed with S-50 or S-52, 35ps)
Bandwidth: 14.5GHz
Input impedance: 50Ω (terminated SMA)
Noise: <5mV
Maximum signal voltage: ±500mV
Maximum input voltage: ±5V
http://groups.io/g/TekScopes/topic/7s_7t_sampling_system/948674?p=Created,,,20,0,0,0'I have been told that construction of S4 is quite terrible, so sampling pulses blow by on the bridge is significant and low level signals can be drown in that noise. I gentleman who told me about that, said that he almost scrapped an amplifier that he was building, but power meter showed no noise, so he went out and found S6, which confirmed the amplifier was all right. The
S-5 SAMPLING HEAD
'The Tektronix S-5 sampling head plug-in is the slowest of the S-series samplers.'
http://w140.com/tekwiki/wiki/S-5
Rise time: 1ns
Bandwidth: 350MHz
Input impedance: 1MΩ (BNC)
Noise: 500µV
Maximum signal voltage: ±1V
Maximum input voltage: 100Vdc
S-6 SAMPLING HEAD
http://w140.com/tekwiki/wiki/S-6
Rise time: 30ps
Bandwidth: 11.5GHz
Input impedance: 50Ω (unterminated SMA)
Noise: <5mV of noise
Maximum signal voltage: ±1V
Maximum input voltage: ±5V
S-50 PULSE GENERATOR HEAD
http://w140.com/tekwiki/wiki/S-50#Comparison_of_S-50_and_S-52
Rise time <25ps
Amplitude 400mV into 50Ω (SMA)
Output impedance varies, not specified
Sequencing of internal events analogue timing circuit
Pre-trigger lead time 75ns
Pre-trigger to pulse jitter 15ps +1V into 50Ω (BSM)
Pulse repetition rate 25kHz
Trigger (with pulse) output 200mV into 50Ω
S-52 PULSE GENERATOR HEAD
http://w140.com/tekwiki/wiki/S-52
Rise time <25ps
Amplitude 200mV into 50Ω (SMA)
Output impedance 50Ω reverse terminated
Sequencing of internal events digital state-machine
Pre-trigger lead time 85ns
Pre-trigger to pulse jitter <10ps +1V into 50Ω (BSM)
Pulse repetition rate 120kHz
S-53 TRIGGER RECOGNISER HEAD
http://w140.com/tekwiki/wiki/S-53
Bandwidth: 1 GHz
Input impedance: 50Ω BNC connector
Sensitivity: 10mV to 1Vpp
Jitter: ≤15ps (from I/P to trigger O/P
Delay: 15ns (from I/P to trigger o/p
Trigger O/P: 1V into 50Ω with 600mV/ns slew
The S-53 Trigger Recognizer head effectively turns the 7S12 into a sampling oscilloscope.
S-54 PULSE GENERATOR HEAD FOR THE 7S12
https://w140.com/tekwiki/wiki/S-54
Output pulse rise time <1ns
Output pulse Pulse duration 25µs ±2µs
Amplitude 200mV into 50Ω (BSM)
Pre-trigger output rise time 5ns
Pre-trigger time 120ns – 1µs before main pulse
Pre-trigger duration 20ns
Jitter <100ps @ 120ns, <1ns @ 1µs
7T11 / 7T11A TIMEBASE
http://w140.com/tekwiki/wiki/7T11
The 7T11 timebase is designed to control up to two 7S11 sampler plugins and goes down to 10ps/div. It has a trigger pulse output, 500mV into 50Ω. The 7T11 and later model 7T11A look identical externally but only the 7T11A works in a Tek 7854 mainframe. Below is a UK-based website with a full description and kit of parts to convert a 7T11 to a 7T11A. Once converted, it will still work in other scopes: http://www.perdrix.co.uk/7T11Conversion/index.htm
7S11 SAMPLER
http://w140.com/tekwiki/wiki/7S11
The 7S11 sampler accepts one sampling head. It can be controlled by the 7T11 or the 7S12 by sliding them together so that a thin strip of connectors on the left side of the 7S12 / 7T11 mate with a similar strip on the 7S11 right side.
A second 7S11 can be also added to the first 7S11.
The 7S12 is typically fitted with a <25ps S-52 pulse generator and a S-4 or S-6 sampling head, but the slower S-3 has useful attenuator tips that permit significantly higher input voltages.
In addition to the main pulse output, the S-52 also produces a pre-trigger pulse 85ns before the main pulse. The delay is generated with a digital counter, and jitter between the pre-trigger pulse and the main pulse is specified as less than 10ps.
The pre-trigger pulse uses a proprietary Tek BSM connector that is very difficult to source. Others have replaced theirs with SMA connectors and I was going this route too until I fortuitously came across a couple at an electronics surplus supplier in France.
This time I made certain I got a working 7S12 which by fortune came with both a S-52 25ps rise time 200mV pulse generator and an S-6 30ps sampling head as well as TDR connecting coax, see above.
I had read that due to lack of an internal 50Ω terminator the S-6 sampler is prone to ESD damage (it is good practice to leave a termination resistor on the unit when unused, to give some protection against ESD) so it was an added bonus to get a known good one. It was only now I considered a TDR in its own right might be of great benefit to the design of the board layout for my pockels driver. Subsequently I also bought an S-4 head and locally, a 7S11 for a second channel.
Note: An external trigger, perhaps utilising the 7M11 dual 75ns (175ps rise time) delay line, may be required for dual-channel synchronisation on the 7S12: when I hooked up the 7S12 and 7S11 I found the channels were temporally displaced. 2020 note: the only way I've got around this so far is by using two independent 7B10 timebases on a 7104, see [Projects: Tek 109 1-Shot & LF Mod].
http://bee.mif.pg.gda.pl/ciasteczkowypotwor/Tek/7854_ops_sm.pdf
The 7834 manual said it would accept the 7S12 but I found the best combination would be in a 7854 which is itself a combination of an analogue oscilloscope and a slow 500kHz sampler (see page 32 of Tek's 7k series 1983 catalogue [I31]), offering an averaging function and digital storage as well as measurement cursors. Its math capability also provided a neat solution to an issue I had held in the back in my head for some while - how to calculate the impedance of the unknown flashlamps I intended to use for the ruby laser. Kigre's Chris Hardy had given me some useful advice:
'Using a good hall-effect current probe (such as Pearsons), a high voltage probe and oscilloscope (Tek, etc) you can determine the lamp's impedance at any time in the pump pulse. The most convenient way is to connect the probes to the scope and the scope to a computer to calculate the impedance. You can also measure the input energy as the integration of lamp voltage and current over time (the pump pulse) using a scope and computer. Some fancy scopes can do this without a computer if they have advanced math capability'.
I found a 7854 for $250 complete with plugins, keyboard and the original manual:
Finally, a good 7104 appeared at $300 and I bought it along with a couple of 900MHz P6201 FET probes with no accessories from eBay France that both turned out to be faulty - see [Repairs]. At the time eBay USA auctions listed them at ridiculously high prices. By good fortune a year later I bought a couple of new ones on eBay USA at $70 apiece, in their original cases complete with their entire accessory set including all attenuators, earth leads and the usually absent end spring hook.
Regarding the P6201, worthy of note for low voltages is a DIY 100kHz - 1GHz DIY probe designed by UK RF engineer David Jewsbury, that was published in Elektor magazine, Issue 10 April 2004, downloadable here:
http://www.elektormagazine.com/magazine/elektor-200410/17818
http://elektrotanya.com/files/forum/2009/10/e04a036.pdf
In 2018 it was for sale on eBay.com as 'RF Active Probe + DC Out 100KHz - 1GHz', from seller janielectronics from Hungary, in a variant that violates the original design by omitting the ground wire from the end of the probe. Nonetheless it's a snip at the offered price of around $30.
Had I seen this after I got the faulty P6201s I might have gone this way instead.
BANDWIDTHS COMPARED
Once I had all of the oscilloscopes, I ran an experiment with my Signal Hound
USB-TG44A 10Hz - 4.4GHz Tracking Generator accompanied by a
USB-SA44 1Hz - 4.4GHz spectrum analyser, now succeeded by the model USB-SA44B:
http://signalhound.com/products/usb-tg44a
http://signalhound.com/products/usb-sa44b
What I discovered was very interesting.
Perhaps unsurprisingly, my 0.4GS/s 100MHz Rigol DS1102CD can show nothing faster than 100MHz (although stunningly it is capable of triggering off a single 2ns pulse, if not displaying it very accurately), but the fastest discernible frequency on all of the Tek analogue oscilloscopes is amazing.
For the 2455B I used the 50Ω input and coax,
For the 7000 series I used 50Ω coax into a 1GHz 7A29 input with a 1GHz 7B10 timebase:
SignalHound TG44A tracking generator @ -5dB:
100MHz Rigol DS1102CD: fastest locked sinewave: 100MHz
250MHz Tek 2455B: fastest locked sinewave: 500MHz
400MHz Tek 7834: fastest locked sinewave: 1.55GHz
400MHz Tek 7854: fastest locked sinewave: 1.70gHz
1GHz Tek 7104: fastest locked sinewave: 2.20GHz
In all of the shots below, the timebase is set to 200ps.
The tracking generator assures certainty of measured frequency.
Some while after I put these up I realised the trace on the 7854 below left is 2ns/6cycles = 3GHz, not 1.7GHz as it is incompatible with the 200ps setting of the 7B10 timebase. Instead it looks like it's been displayed at twice the 1.7GHz frequency minus an error of 400MHz (3.0GHz - (1.7GHz x 2) = -400MHz).
However although greatly attenuated, the signal is nonetheless locked by the trigger.
Tek 7854 Mainframe 7854 fastest lock 7104 fasted lock Tek 7104 Mainframe
Over the years of research whilst buying these instruments I eventually realised what I really wanted was a Tek 7250 6GHz transient digitiser (photo on right) but it:
a) costs twice as much as all of the above together,
b) is 19" wide x 36" deep x ~9" high & weighs 132lbs,
c) only has one input channel
That the 400MHz 7854 can see 1.70GHz is stunning. Equally impressive is the 7104. If it were not incredible enough its cheap eBay 7B10 timebase can actually get a stable trigger at 2.0GHz, just look at the amplitude of the vertical deflection: the tracking generator -10dBm output is 70.7mV and the trace (above right) has only been attenuated by 10mV.
Later I used an HP8620C + HP86222 RF sweep generator to observe the bandwidth characteristics:
Rigol DA1102CD bandwidth Tek 7834+7A29 bandwidth
I have an embarrassingly large collection of Tek 7000 series oscilloscopes and very little space left in my home, but no wish to part with any of them.
2018 UPDATE
AN ISOLATED CHANNEL DSO
Whilst investigating how the MCP GATE control signal worked on the PCM-401 [Personal LIBS: Prototype Integration], I inadvertently damaged it. It occurred to me connecting my Rigol scope probe ground clip to the GND pin on the HV MCP PSU could have been the culprit if there was an opto-isolated interface between the two, and I had inadvertently shorted it out.
Just the possibility of this being the case meant I would need a means of determining how to fix it without repeating the same mistake and further damaging it. I considered buying a cheap (but still money) eBay Chinese battery-powered DSO but remembering the Chinese inductance meter only lasted a year before it locked up forever in the middle of a calibration cycle, I thought this might be throwing money down the drain. Then I remembered the distant past when I was working on a stabilised navigation platform on a ship and how impressed I had been with the handheld Tektronix THS720 100MHz 1GS/s (500MS/s per channel) DSO we had at the time.
A little research revealed the THS720P variant with HV isolated inputs and power measurement firmware. Such a beast is invaluable for SMPSU and high voltage work, and would certainly find use in other parts of the project. It first appears in Tek's 1997/1998 catalogue, basic spec on page 92 (Tekwiki [I31]).
In 2019 there were many on eBay although most still up in the $600+ bracket unless damaged in some way.
Eventually I found a working THS720P in the UK for £260 ($330) with tatty but functional BNC connectors that was last calibrated late 2016 (i.e. recently expired end of 2017), and bought it.
See [Repairs: Other Tek Repairs] for details of a DIY replacement battery.
[I31], http://w140.com/tekwiki/wiki/Tektronix_Catalogs
Page 620 lists the basic THS720A model at $2295 and the THS720P at $2995, equivalent to $4665 in 2019:
[C31], https://www.dollartimes.com/inflation/inflation.php?amount=3000&year=1998
Tek THS720 Std Tek THS720P
2019 UPDATE
A PC-BASED DSO
After some time playing with the Tek 7854 math functions it dawned on me just how hard it would be to create mathematically derived functions, as it is tantamount to writing machine code. I can do it, but I don't have an IEEE interface and compatible software with which to enter or save the code. Instead I'm limited to the dedicated keyboard and hoping the backup battery won't fail. Even basic cursor movement is tediously based on a bright dot, the forerunner of the line cursors we enjoy today. It is also only capable of digitising one channel at a time and even then it's very noisy compared to a modern DSO. I bought the 7854 for its math capability but I have no regrets as now I use it exclusively with its high sensitivity GHz samplers and S-53 universal trigger recogniser or the equally useful TDR. I can't use them on the 7104 because they will burn out its MCP intensifier.
As for the maths, it occurred to me a better solution would be to buy the cheapest Pico Technology
PC-based DSO, the 2-channel 10MHz 100MS/s 8kpts 2204A (£98 without probes in 2019) that shares the same customisable math function software with the highest spec devices in the range.
Another useful feature is comprehensive support for SPI, I²C, etc serial protocols which I've found invaluable for debugging the slow peripheral ICs I use. It's also conveniently tiny, and easily pops into a pocket if I work away from home. Despite the 10MHz bandwidth I was pleasantly surprised it accurately resolved a 20MHz sine wave, doubtless down to the generous 100MS/s sampling rate.
Being a DSO it also has true single-shot capability unlike the 7854 whose 400MHz 7B87 one-shot capable timebase only works up to 50µs/div, so you can't capture short pulses. The 2204A's 100MS/s sampling rate of 100MS/s ÷ 5 = 20MHz (observed & confirmed) into [Q4c] Tr = 0.4/20MHz = 20ns rise time.
I was however disappointed by the amateur feel of the software supplied with it, which like my Pico Tech TC-08 thermocouple datalogger was buggy and difficult to use. I'm pleased to say the new PicoScope 7 software is a massive improvement, with significantly intuitive controls, and a cheap DSO I would never have considered before has become a surprisingly versatile tool.
https://www.picotech.com/oscilloscope/2000/picoscope-2000-specifications
https://en.wikipedia.org/wiki/PicoScope_(software)
https://www.youtube.com/watch?v=Wj_jW4XLWFw
PicoTech 2204A PicoScope 6 (REPLACE WITH 7)
2022 UPDATE
Around 2014 I abandoned my plan to upgrade my Rigol DS1102CD that I bought new in 2007. Now in 2022 another 7 years have passed and it's 14 years old, yet still my go-to first choice, an impressive
indicator of build quality. Understandably it now has a few issues: the switch on the menu universal selection knob is worn and misses every few clicks, the channels have a small dc offset that auto-cal won't shift, and the voltage measurements are not as accurate as they once were. The trigger is now erratic and only works properly after I have run auto cal which annoyingly takes several minutes.
A couple of years back I asked Rigol UK if they could calibrate it for me and was told this wasn't possible as it was an obsolete non-UK model. I love my Rigol and it's usually the scope I first turn to, but like my Seat Ibiza that I finally had to scrap after 27 faithful years, see [Home: The Need for Speed], it's getting very old and I now find myself worrying it will fail and I'll need a replacement DSO, but I will still keep the analogue scopes which will always have some advantages over digital.
Not to scale: Rigol DS1102CD Rigol DS2202A
I have no way of knowing if their latest offerings are as good, but I do know people still rave about Rigol quality. Apparently the name was deliberately derived from the word 'Regal', and is correctly pronounced this way. Given the astronomical price of modern DSOs from market leaders like Tek and HP, I found myself again looking at Rigol's offerings, which often have more features included at a lower price at the expense of a reduction in overall refinements and quality. Rigol also seems to have a habit of regularly cutting prices.
DSO HISTORY
As time has passed, oscilloscope price/performance has decreased exponentially, taking advantage of the high speed high integration route spearheaded by mass-market PC manufacturing, reducing hardware footprint and costs whilst maintaining profits. Long gone are bulky discrete components, replaced by fast ADCs coupled to vast gate arrays forming complete PCs on which ever more complex software runs. Manufacturers increased profits by adding optional plug-in memory cards and competitors followed suit, later replacing the physical hardware with a software licence key. With GHz multi-million gate array chips commonplace, mass production profit goals further drove minimisation of hardware, permitting a single board to cover all models in a range simply by disabling or reducing functionality. This opened a new market opportunity for expensive upgrades, further encouraging purchase of cheaper sub-range models.
Production took advantage of Asia's low cost labour market and further savings were made by combining design offices with foundries and the Chinese in particular began to offer lower priced higher spec instruments with increasing reliability. Rigol is one of those companies. Initially concealing the hidden upgrade process until their 50MHz DS1054Z was famously hacked to 100MHz, led to a further advantage: Tek's analogue scopes were designed with generous contingency, see my bandwidth tests above.
DSO bandwidth is constrained by sampling rate and initially no attempt was made to offer similar contingency, but in 2018 Rigol upped the ante with the MSO 5k series, and it caught my eye. Although marketed as 350MHz max, its 8GS/s theoretically gives a 400MHz bandwidth across 4 channels, and even higher with refined sampling of repetitive waveforms. Again, hacks exist to exploit the hidden full bandwidth of lower models.
HACKS
That said, I would rather have a full spec machine than a hacked one, no matter how tempting, particularly as not long after I bought my DS1102CD and put it into a Cal lab, it locked up during their pre-cal soak. With no UK distributor at the time, I had to ship it out to Rigol in China to get it repaired under warranty. I have to wonder if they would still honour the warranty if the same thing happened to a new model I'd hacked. The thought of bricking an expensive new DSO also worries me, although I imagine this is unlikely if all you do is enter a key.
REQUIREMENTS
I had already decided I wanted a 4-channel DSO ideally with a 1ns rise time and 350MHz bandwidth but these were beyond my budget and I returned to Tek and HP to look for a high spec second hand scope, thinking at the time there is a limit to how much I'd be prepared to pay for a new Chinese DSO vs an old US one offering better quality, customer service and support.
A comment in October 2021 on one of Squeaky Dave's [see footnote] teardowns suggests this may actually be inviting trouble: 'Keysight [HP] have recently been a topic in the EEVBlog forums because they now appear to be refusing to provide supply, repair or calibration service for individuals and only providing service for companies.' Is Tek the same? https://www.youtube.com/watch?v=xbqx25wT2Qoh
This is significant given how less reliable Tek's in particular, PC-based DSOs seem to be. A further downside is these DSOs usually have noticeably less sampling memory and any external storage medium is often a grindingly slow floppy disk rather than a modern Chinese USB port. Lastly, they're usually big and heavy and rarely have schematics. I found myself returning to Rigol, also without schematics, but I would only consider buying a new one of these so at least I'd have a warranty.
Based solely upon affordability I thought I'd compromise with a Rigol DS2202A 4-channel 1.75ns 200MHz 2GS/s 56Mpts, 50kWf/s (£540 see below) but having kept my eye on Rigol's DSO developments over the years and now armed with a fuller appreciation of the needs, I began to consider better models.
In 2007 I paid $1500 on eBay USA to import my 2-channel 100MHz 0.4GS/s DS1102CD MSO with LA. In 2022 that equates to ~$2080 or ~£1700 and should be the minimum I pay for my upgrade, but Rigol now has UK distributors, so no currency exchange fee or international shipping which has risen considerably, or 7% import and 20% VAT that have to be paid on both that and the instrument. Actually a big saving.
First, let's see what's available. In May 2022 Rigol's prices look amazing.
RIGOL RANGE:
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/
DS1102E 2-channel 100MHz 1GS/s £260 +20% VAT = £312
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-ds1000e-series/
DS1054Z 4-channel 50MHz 1GS/s £305 +20% VAT = £366
DS1074ZPlus 4-channel 70MHz 1GS/s £359 +20% VAT = £431
DS1104ZPlus 4-channel 100MHz 1GS/s £395 +20% VAT = £474
DS1202ZE 2-channel 200MHz 1GS/s £294 +20% VAT = £412
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-ds1000z-series/
DS2102A 2-channel 100MHz 2GS/s £432 +20% VAT = £518
DS2202A 2-channel 200MHz 2GS/s £450 +20% VAT = £540
DS2302A 2-channel 300MHz 2GS/s £540 +20% VAT = £648 (£1600 when first launched)
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-ds2000a-series/
The £1700 I paid for my DS1102CD in 2007 could easily buy one of their exotic MSO5000 8GS/s DSOs:
MSO5104 4-channel 100MHz 8GS/s £900 +20% VAT = £1080
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-mso5000-series/
If I were to buy a new DSO I would now want it to offer me a bandwidth improvement for the LHC project that is currently using the 7104 for measurements that are beyond the 100MHz DS1102CD. The 7104 has other limitations: in single shot dual channel mode it can only display one trace at a time in what has to be either alternate or chop mode, as I observed in [Projects: Fast PD Detector].
There will always be a need for the 7104 but a DSO is significantly easier to operate and incorporates measurement enhancements that greatly speed up development, not to mention digitised storage.
The new plan was not to replace the 7104, but to find the highest spec DSO I can afford in order to minimise 7104 usage, which will also extend the life of its MCP,
CHEAP CHEAT
A cheeky lower cost solution to 4 channels might be to buy use two 2-channel DSOs. Since the 3rd and 4th channels are likely to be hosting slower signals, I could pair up a Rigol DS2302A 2-channel 300MHz 2GS/s with my DS1102CD and get 1-channel good for a 4ns rise time. Unlike the MSO series, the Rigol DS2302A has an external trigger.
Or instead buy 2 x Rigol DS2302A £648 for £1296 and get 2 channels good for a 4ns rise time.
However it's not entirely practical (even without the physical mounting of 2 separate DSOs, my DS1102CD fan is very loud; if the DS2302A fan is the same, two fans would be a nightmare).
IT'S NOT WORTH IT
It's still a huge amount of money. If I have to pay over a grand, I might as well save up and get the MSO5354 that should cover rise times all the way up to 2ns before the 7104 takes over. I can't afford a MSO5354 at the moment but given how low prices have dropped on Rigol DSOs and the MSO5000 series' well advertised issues, Rigol might drop the prices before long and phase it out for something better.
Besides, I only need a basic MSO5354:
The AWG and LA are expensive add-on options but the LA trigger on my DS1102CD is rather simplistic (videos of the MSO series look little different) and I really don't like the way Rigol LA traces are crammed so close together vertically, and from the specs it appears a standalone AWG is better than a scope based one.
I also don't need the serial coms package as these interfaces are usually quite slow and my cheap 10MHz Picoscope 2204A can do that instead.
The next higher spec series is the DS7000 but lower ranges share the same parameters and prices are prohibitive.
COMPETITORS
There are countless DSO manufacturers but Rigol and Siglent stand out as direct low cost Chinese competitors to mainstream US Industrial giants Tektronix and HP/Agilent. Rhode & Shwarze offers series somewhere in-between but just as Siglent is more expensive than Rigol, R&S is more than Siglent, and lacking some of the extra features found on the others. Therefore for now I'm only looking at Rigol and Siglent whose quality far surpasses the likes of Hameg and Tenma, which are excluded for this reason (I'm convinced the name Siglent was chosen to draw in more trade when people Googling for Agilent mistyped the name; the same goes for Fnirsi vs Anritsu).
Below are Rigol and Siglent models I considered, including some just out of my reach:
RIGOL RANGE:
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/
Rigol DS2302A 300MHz 2GS/s 56Mpts/#chs 50k wfm/s
https://www.rigol-uk.co.uk/product/rigol-ds2302a-300mhz-digital-oscilloscope/
Rigol MSO5354 350MHz 8GS/s 200Mpts/#chs 500k Wfm/s
https://www.rigol-uk.co.uk/product/rigol-mso5354-350mhz-digital-oscilloscope/
https://www.rigol-uk.co.uk/jg/wp-content/uploads/2021/08/Rigol-MSO5000-Oscilloscope-Datasheet.pdf
Not to scale: Rigol DS2302A Rigol MSO5354
SIGLENT RANGE:
https://siglent.co.uk/products/digital-oscilloscopes
https://siglent.co.uk/products/digital-oscilloscopes/sds1000x-e-series/
Siglent SDS1104x-E 100MHz 1GS/s 14mpts/#chs 100k Wfm/s
https://siglent.co.uk/products/digital-oscilloscopes/sds2000x-plus-series/
Siglent SDS2204x+ 200MHz 2GS/s 200Mpts/#chs 120k Wfm/s
Siglent SDS2354x+ 350MHz 2GS/s 200Mpts/#chs 480k Wfm/s
https://siglent.co.uk/product/siglent-sds2354x-plus-4ch-350mhz-2gsa-s-super-phosphor-oscilloscope/
The 350MHz Siglent SDS2354x Plus can be upgraded to 500MHz, but the cost is prohibitive.
Not to scale: Siglent SDS1104x-E Siglent SDS2354x Plus
SILLY MONEY: OTHER RIGOL / SIGLENT MODELS
Rigol DS7034 350MHz 4-ch 10GS/s 500Mpts 600kWfm/s 10.1" LCD £3873 + VAT = £4648 @ 30/04/22
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-ds7000-series/
Siglent SDS5034X 350MHz 4-ch 5GS/s 250Mpts 480kWfm/s 10.1" LCD £2571 + VAT = £3085 @ 30/04/22
Not to scale: Rigol DS7034 Siglent SDS5034X
PRICES (LAST UPDATE 16/11/22)
Prices for the DSOs I've identified above.
Siglent.co.uk prices include LA and all serial bus trigger options
Rigol.co.uk prices exclude LA and all serial bus trigger options are a time-limited offer.
Rigol DS2202A 200MHz 2-ch 2GS/s £ 428 + 20% VAT = £514
Rigol DS2302A 300MHz 2-ch 2GS/s £ 508 + 20% VAT = £610
Rigol MSO5074 70MHz 4-ch 8Gs/s £ 713 + 20% VAT = £856 (2-ch MSO5072 is £875)
Rigol MSO5354 350MHz 4-ch 8GS/s £1821 + 20% VAT = £2185
Siglent SDS2104X Plus 200MHz 4-ch 2GS/s £ 849 + 20% VAT = £1019
Siglent SDS2354x Plus 350MHz 4-ch 2GS/s £1780 + 20% VAT = £2136
Siglent SDS2354x Plus 500MHz 4-ch 2GS/s +£1148 + 20% VAT = £1378 UPGRADE (+£2136 = £3514)
Rigol DS2000 series prices
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-ds2000a-series/
Rigol MSO5000 prices
https://www.rigol-uk.co.uk/products/digital-oscilloscopes/rigol-mso5000-series/
Siglent SDS2000X Plus series prices
https://www.siglent.co.uk/products/digital-oscilloscopes/sds2000x-plus-series/
Siglent SDS2354x Plus 350MHz to 500MHz upgrade price:
https://telonic.co.uk/product/siglent-sds2354x-plus-bandwidth-upgrade-350mhz-to-500mhz-only-2-chs/
SAMPLE RATE MATTERS
The MSO5354 sample rate far surpasses that of the Siglent SDS2354x Plus.
Page 79 of the Rigol MSO5000 series manual states:
The maximum real-time sample rate in the single-channel mode of the oscilloscope is 8GS/s.
Single-channel mode: only one of the four channels (CH1/CH2/CH3/CH4) is enabled.
The maximum real-time sample rate in the half-channel mode is 4GS/s.
Half-channel mode: either CH1 or CH2 is enabled; and either CH3 or CH4 is enabled.
The maximum real-time sample rate in the all-channel mode is 2GS/s.
All-channel mode: CH1/CH2 are both enabled or CH3/CH4 are both enabled.
(I assume the latter is a typo, as All-channel mode implies all 4 channels are enabled)
IOW: 1 channel = 8GS/s, 2 channels = 4GS/s each, 3 or 4 channels = 2GS/s each.
https://www.rigol-uk.co.uk/pdf/Rigol-MSO5000-User-Guide.pdf
Page 65 of the Siglent SDS2000x Plus series manual explains Nyquist sampling but only says the maximum sample rate is 2GS/S. The manual doesn't explain multiple channel sample rates.
However 1 or 2 channels = 2GS/s, and 3 or 4 channels = 1GS/s each.
https://siglentna.com/resources/documents/digital-oscilloscopes/#sds2000xp
THE 350MHz MSO5354 CAN SEE 500MHz
This video shows a 70MHz MSO5072 hacked to the 350MHz full upgrade bandwidth, displaying a 500MHz
sine wave within the 3dB limit, dropping below when a second channel is turned on (still maintaining 3dB at 420MHz), yet adding the remaining channels has surprisingly minimal degredation:
Title: '#838 Rigol MSO5072 upgraded to MSO5354'
https://www.youtube.com/watch?v=eaoHYWYLRV0
I'd like to know how high it goes before the waveform gets deformed. I assume 500MHz was the highest the reviewer could go. The near perfect 500MHz sine wave suggests it could go higher still.
06/11/22 Update - it can see 1GHz !!!
https://www.youtube.com/watch?v=90VVZww2h3A
Well that is just crazy cool. Low amplitude as you'd expect, but LOCKED and a solid sinewave.
However questions remain: what about a square wave, and what about fast pulse response?
The initial post on 02/11/22 from the same poster, but I don't speak Portuguese:
He's got the onboard spectrum analyser locked on 1.9GHz!
https://www.youtube.com/watch?v=A9AU4Ovo6Xg
20/03/22 Update
A Spanish video with analysis examples of real electronics applications:
https://www.youtube.com/watch?v=0eSXdp4XK-g
BANDWIDTH VS PULSE RISE TIME
High sample rate is beneficial to single pulse BUT what matters more for digital signals with fast rise times is the DSO itself has a faster rise time: https://www.tek.com/en/blog/which-oscilloscope-to-buy
'An oscilloscope's rise time should be < 1/5 x fastest rise time of the signal.'
'A 4ns rise time needs a scope with faster than 800ps rise time.'
I'm likely to be looking at laser Q-switch pulses from 4ns to 20ns. 800ps translates to a 435MHz scope.
I concluded I should get the fastest scope I can afford and abandoned bandwidths below 350MHz.
The 350MHz Rigol 8GS/s MSO5354 is presently beyond my budget, but its perceived ability to capture much higher bandwidths suggests it is one to seriously consider. In a similar price range, the 350MHz Siglent SDS2354x Plus seemed a good competitor.
Next I compared the 350MHz Rigol MSO5354 and Siglent SDS2354x Plus in greater detail:
SAMPLING RATE VS PULSE RISE TIME
DSOs are always a compromise because they never take a full snapshot of a signal, instead guessing what is in the gaps between sample snapshots. Analogue scopes are better for single-shot if the entire trace can be captured on the screen, but suffer the same issue with repetitive waveforms when the electromagnetic coils are returning the blanked-out dot back to the left of the CRT, effectively also the equivalent of a sampling operation.
The sample rate needs to be sufficient to capture not just the outline of the signal in question, but also any smaller signals imposed upon it, that by definition must be a higher frequency, necessitating a higher sample rate.
At the top of this page I listed Nyquist [Q4] rule of thumb equations to determine the minimum required sampling rate.
Below, based on the frequency handling revealed by the reviews, I have calculated the sampling rate for pulse rise time using [Q4d] 3.5x as the minimum and [Q4e] 5x as the optimum, and identified the theoretical points at which the 350MHz Rigol MSO5354 and Siglent SDS2354x Plus DSOs should be able to accurately display these signals.
The 350MHz Rigol MSO5354 came out overall winner largely based on the review above that reveals its ability to accurately display a 500MHz sine wave as a direct result of its high sample rate. In reality 8GS/s only applies when just one channel is used but this is still impressive. Unsurprisingly the more channels in use, the slower it samples: 2 channels both run at 4GS/s and 4 channels each run at 2GS/s.
First we'll start off with silly numbers from HP's example equation in [Q4d].
Rise time = 100ps, BW = 0.4/100ps = 4.0GHz x1.4 = 5.6GHz
OPTIMUM: GENERAL 5x CALC:
5.6GHz x5 Nyquist = 28.0GS/s x 4ch = 112GS/s -
5.6GHz x5 Nyquist = 28.0GS/s x 3ch = 84GS/s -
5.6GHz x5 Nyquist = 28.0GS/s x 2ch = 56GS/s -
5.6GHz x5 Nyquist = 28.0GS/s x 1ch = 28GS/s -
MINIMUM: HP 3.5x CALC:
5.6GHz x2.5 Nyquist = 14.0GS/s x 4ch = 56GS/s -
5.6GHz x2.5 Nyquist = 14.0GS/s x 3ch = 42GS/s -
5.6GHz x2.5 Nyquist = 14.0GS/s x 2ch = 28GS/s -
5.6GHz x2.5 Nyquist = 14.0GS/s x 1ch = 14GS/s *7854+S4
* The 7854 + 14.5GHz S4 sampler would see this, but only if repetitive as it samples at 500kHz [I31].
These rise times are way above the Rigol MSO5354 and Siglent SDS2354x Plus capabilities, so we will pass on the next few iterations:
Rise time = 200ps BW = 0.4/200ps = 2.0GHz x 1.4 = 2800MHz required scope bandwidth
Rise time = 300ps BW = 0.4/300ps = 1.3GHz x 1.4 = 1867MHz
Rise time = 400ps BW = 0.4/400ps = 1.0GHz x 1.4 = 1400MHz
Rise time = 500ps BW = 0.4/500ps = 800MHz x 1.4 = 1120MHz
Rise time = 600ps BW = 0.4/600ps = 667MHz x 1.4 = 933MHz
Rise time = 700ps BW = 0.4/700ps = 571MHz x 1.4 = 800MHz
Rise time = 800ps BW = 0.4/800ps = 500MHz x 1.4 = 700MHz
Page 20 of the MSO 5k series datasheet says it can trigger off a minimum rise time of 800ps, but it won't achieve a 3% error rate on the measurement.
Rise time = 900ps BW = 0.4/900ps = 444MHz x 1.4 = 622MHz
Rise time = 1.0ns BW = 0.4/1.0ns = 400MHz x 1.4 = 560MHz
Rise time = 1.1ns BW = 0.4/1.1ns = 364MHz x 1.4 = 509MHz the MSO5354 should be in range now.
Rise time = 1.2ns BW = 0.4/1.2ns = 333MHz x 1.4 = 467MHz
Rise time = 1.3ns BW = 0.4/1.3ns = 308MHz x 1.4 = 431MHz
Rise time = 1.4ns BW = 0.4/1.4ns = 286MHz x 1.4 = 400MHz
Let's find the minimum rise time (HP 3.5x CALC) that can be accurately displayed on 3 or 4 channels
Rigol MSO5354 @ 3 ch or 4 ch = 2GS/s:
3-ch BW = 0.4/2.0ns = 200MHz x1.4 = 280MHz x2.5 Nyquist = 700MS/s x 3ch = 2.1GS/s -
3-ch BW = 0.4/2.1ns = 190MHz x1.4 = 267MHz x2.5 Nyquist = 667MS/s x 3ch = 2.0GS/s MSO5354
Siglent SDS2354xP @ 3 ch or 4 ch = 1GS/s:
3-ch BW = 0.4/4.1ns = 97MHz x1.4 = 137MHz x2.5 Nyquist = 341MS/s x 3ch = 1.02GS/s
3-ch BW = 0.4/4.2ns = 95MHz x1.4 = 133MHz x2.5 Nyquist = 333MS/s x 3ch = 1.00GS/s SDS2354xP
Across 4 channels, the Rigol will display a fairly accurate 2.0ns rise time vs 4.2ns for the Siglent.
REVIEW VIDEOS
Below are reviews of comparable Rigol and Siglent DSOs that give a feel for what they could do in the real world away from their manufacturers' glossy brochures and specifications:
MAXIMUM BANDWIDTH - RIGOL vs SIGLENT
This excellent test compares 1-ch bandwidth on the 100MHz 1GS/s Rigol DS1104Z Plus & Siglent SDS1104X-E
Title: 'Rigol vs Siglent 4 channel scopes review, test'
https://www.youtube.com/watch?v=wZQj7TnWVNg
Summary:
AUTO to trace update delay: Rigol DS1104ZPlus 8 secs vs Siglent SDS1104X-E 3 Secs
-3dB bandwidth: Rigol DS1104ZPlus ~ 139MHz vs Siglent SDS1104X-E ~ 117MHz.
Max displayed sine wave freq: Rigol DS1104ZPlus ~ 400MHz vs Siglent SDS1104X-E ~ 186MHz.
The Rigol also correctly measures 400MHz.
BOOT TIME
Rigol MSO5354 (350MHz 4 ch) takes well over a minute to boot.
https://www.youtube.com/watch?v=MRzUZsCxGdo/
Siglent SDS2354X Plus (350MHz 4 ch) takes 43s to boot.
https://www.eevblog.com/forum/testgear/siglent-sds2000x-plus-coming/150/
HACKING
All Rigol MSO5000 speed versions can be hacked to the full 350MHz bandwidth.
The Siglent 350MHz can be hacked to 500MHz. Bandwidth control appears to be down to an RF amp chip with a serial interface controlled selected bandwidth limiter. Both have 1k x 600 screens and low AWG Vout of ±2.5V. See Squeaky Dave's teardown:
'EEVblog #1309 - Siglent SDS2000X plus Scope Teardown+Hack'
https://www.youtube.com/watch?v=HxzQS-Bn2R0/
THE RIGOL MSO5000 SERIES HAS A LOT OF ISSUES
This Nov 2018 video shows Squeaky Dave proudly showing off the newly launched Rigol MSO and then discovering it locking up completely. As he says, 'Never seen that before on a DSO'. Hopefully by now that embarrassing bug has been fixed:
'EEVblog #1146 - New Rigol MSO5000 Oscilloscope'
https://www.youtube.com/watch?v=P5faiEUXbGg
It wasn't very good at serial comms either, another Squeaky Dave special. Let's hope that is fixed too:
'Rigol MSO5000 Bug Bonanza'
https://www.youtube.com/watch?v=UDGsZcAWgL8
This EXCELLENT comparison of the Rigol MSO5000 vs Siglent SDS2000X reveals many shortcomings of the MSO5000 that in most cases are trounced by the SDS2000X Plus, which also has a few issues pointed out.
Both accept 1MΩ I/Ps, but the Siglent also accepts 50Ω I/Ps. The poster admits to being a past 'Rigol fanboy' now instead preferring the Siglent. I too much prefer the Siglent, but wish it matched the MSO5000's sample rate. Perhaps there is a chance Siglent will rise to the challenge?
A comment in the following video says: 'The Siglent is 2x2 channel oscilloscope (grouped ch1+2 and ch3+4), which means that the sample rate is 1Gs/s at minimum for 3 and 4 channel use and 2Gs/s if 1 or 2 channels are used.'.
'Comparing the Rigol MSO5000 vs Siglent SDS2000X Plus'
https://www.youtube.com/watch?v=93QUNt1z6Gw
Live mode: apparently the Rigol isn't very good at this, or at producing a compatible HDMI video output:
'Siglent SDS2000X Plus vs Rigol DS4024 on live update'
https://www.youtube.com/watch?v=IZcfGfV3xu0
FINALLY, HERE'S SOMETHING REALLY COOL ABOUT THE RIGOL MSO5000 !!
'Running Doom on the MSO5000 series MSO5074'
THE SIGLENT ALSO HAS A FEW ISSUES
Squeaky Dave found something interesting about the way the Siglent displays its trace:
You have to zoom out before triggering on a trace, to be able to zoom in & see more, or press HISTORY VIEW. It has 200Mpts but at 10µs TB you only get 100kpts. Only at 10ms TB do you get 200Mpts. History does display more: 2GS/s @ 1µs/div with 10 divs on screen = 2G / 20kpts per screen = 100k samples across 10µs = 500ps = 2GHz. 200M/100k samples = 1987 frames. ('LeCroy does the same').
'EEVblog #1312 - Siglent Oscilloscopes CRIPPLING History Mode!'
SUMMARY
Based upon everything above and in particular the video reviews, I've collated the features that differ between the Rigol MSO5354 and Siglent SDS2354X Plus, and logged comments relatable to my intended use.
Both the Rigol MSO5354 and Siglent SDS2354X Plus have a 3 year warranty.
BASIC FEATURES MY COMMENTS
RIGOL MSO5354
Maximum Input Voltage CAT I 300Vrms, 400Vpk Same
Vertical scale 500μV/div – 10V/div Same
Resolution 8-bits Worse (repeating waveforms likely no worse)
Rise time ≤1ns Same
Timebase selection 1ns/div - 1000s/div Same
Screen resolution 1024 x 600 (WVGA) Same
Dimensions 367mm × 200mm × 130mm 9,542,000 cu mm (9% bigger)
SIGLENT SDS2354X PLUS
Max Input Voltage 1MΩ: 400Vpk (dc + ac), dc~10kHz Same
50Ω: 5Vrms, ±10V Peak Useful
Vertical scale 1MΩ: 500μV/div – 10V/div Same
50Ω: 500μV/div – 1V/div Useful
Resolution 8-bits, 10-bit mode @ 100MHz Better with limitations(Rigol can avg too)
Rise time 1ns @ 50Ω, 800ps @ 500MHz May be the same
TB selection ≤350MHz 1ns/div - 1000s/div Same
TB selection 500MHz 500ps/div – 1000s/div Better but at a very high price
Screen resolution 1024 x 600 (WVGA) Same
Dimensions 352mm × 224mm × 111mm 8,752,128 cu mm (9% smaller)
RIGOL MSO5354 CONS MY COMMENTS
Only 1MΩ I/P impedance I can easily add a 50Ω terminator
Noisy [8->5 bits] less of an issue for digital
Timebase Accuracy ±10 ppm; ±10 ppm/year Siglent is far better (DS1102CD = 100ppm)
No external trigger 4 channels is enough
9" LCD Better than my 5.7" DS1102CD
IP experience is buggy I don't need IP
External mouse option I'm unlikely to use it
Sluggish LCD touch & drag I usually try to avoid touch screens
Can hang BAD
Dim button LEDs BAD
Fiddly menu select button overshoots BAD
No hex search in I2C etc I'll use the 10MHz Pico 2204A instead
LA trace is always shown 1ns after trig point I don't need a LA
Probe x1-x10 switch loose I already have better probes
V menu hides trace cutting trace down 20% in size BAD
Soft buttons misaligned with their text BAD
Black case I'd rather have a white case
It looks less professional I'll live with it
Horizontal scale is 14 divs not 10 My DS1102CD is no better with 12 divs
video I/F is patchy I won't use it
RIGOL MSO5354 PROS
8GS/s max (1 ch=8GS/s, 2 ch=4GS/s, 3/4 ch=2GS/s) It's potentially a 4-channel 400MHz DSO
500ps glitch pulse trigger in Peak Detect mode Yes please
More buttons than touch Yes please
Serial data search but only on analogue I won't use the LA
Live trace capture is normal Good
Can zoom into trace on default trace capture Good
Slightly cheaper Always nice
I'm familiar with Rigol I'm used to their issues
SIGLENT SDS2354X PLUS CONS
2GS/s max (1/2 ch=2GS/s but 3-4 ch=1GS/s (=200MHz scope) It's clearly not a 4-channel 350MHz DSO
Fastest pulse width trigger is 1ns I'd like a faster trigger
More touch than buttons I don't like touch screens
No serial data search I probably won't need it
Live trace capture only works on roll I've used roll once in 14 years
Need history mode on to zoom into trace Fiddly
I'm unfamiliar with Siglent Will I miss the Rigol?
SIGLENT SDS2354X PLUS PROS
Both 1MΩ & 50Ω I/P impedance selectable Desirable but not essential
Lower noise [8->6/7 bits] Always desirable
Timebase accuracy ±1ppm 1st Yr aging; ±3.5ppm 10 Yr aging Outstanding
External trigger Useful but with 4 ch I doubt I'd need it
Much nicer HMI & ergonomics I'd like this but I can live without it
10" LCD I'd like this but I can live with 9"
Good IP-based remote control I'll never use it
Both external mouse AND keyboard option I've no space for a keyboard
Bright button LEDs I'd like this but I can live without it
More math functions The Rigol has enough
Probe readout My probes won't be compatible
Vmenu compresses trace else 20% wider LCD for trace I'd like this but I can live without it
White case I'd like this but I can live without it
Horizontal scale is 10 divs why isn't this a choice on all DSOs?
It looks professional I'd like this but I can live without it
CONCLUSION
The Siglent's low noise levels coupled with an impressive timebase accuracy of ±1ppm 1st year aging and ±3.5ppm 10 year aging together with refined MMI clearly reveals it is a higher quality instrument.
I'd love the Rigol to be this good but the 8GS/s compromise works better for me. Its apparent ability to accurately capture a higher bandwidth suggests it should also capture faster rise time pulses.
It will become less accurate sooner than the Siglent; I will not be able to calibrate its timebase but if I doubt its accuracy I need only verify it using my Signal Hound SA+TG, and a Ballantine 6130A I recently acquired, see [Calibration]. The Siglent has a lot going for it but at the end of the day the boxes the Rigol ticks for me are beyond the Siglent, and I can live with its idiosyncrasies.
DOCUMENTATION
Rigol MSO 5k Series User Guide:
https://www.rigol-uk.co.uk/pdf/Rigol-MSO5000-User-Guide.pdf
Rigol MSO 5k Series datasheet (specification at bottom):
https://www.rigol-uk.co.uk/jg/wp-content/uploads/2021/08/Rigol-MSO5000-Oscilloscope-Datasheet.pdf
Rigol's MSO5354 webpage:
https://www.rigol-uk.co.uk/product/rigol-mso5354-350mhz-digital-oscilloscope/
Siglent SDS 2k x Plus Series datasheet:
https://siglentna.com/wp-content/uploads/dlm_uploads/2021/03/SDS2000X-Plus_Datasheet_DS0102XP_E01B.pdf
Siglent - ALL docs:
https://siglentna.com/resources/documents/digital-oscilloscopes/#sds2000xp
UK distributor Telonic's Siglent-SDS2354x Plus webpage with specs (more readable than Siglent's below):
https://telonic.co.uk/product/siglent-sds2354x-plus-4ch-350mhz-2gsa-s-super-phosphor-oscilloscope/
Siglent's SDS2354x Plus webpage:
https://siglent.co.uk/product/siglent-sds2354x-plus-4ch-350mhz-2gsa-s-super-phosphor-oscilloscope/
2023 UPDATE
Despite all of the above, in 2023 I raided my savings and bought a barely used 3 year old Rigol MSO8104 that had been upgraded to 2GHz (50Ω) and four RP3500A 500MHz probes, together with the LA pod hardware
and two new RP6150A 1.5GHz probes included for free, for 1/3 the new price, from an established company.
It had finally dawned on me a 350MHz DSO would not be fast enough, as my analogue 400MHz Tek 7k mainframes actually display signals well beyond 1GHz. I had completely ignored this range due to its astronomical price. Still regretting many years later my decision not to attempt to buy the original PCM-403 with installation floppies, I decided this too was an opportunity unlikely to be repeated.
Sharing the MSO 5k architecture, the 4 channels are divided into two pairs: Ch1 & CH2 or Ch3 & Ch4 with each pair running at 10GSs. The 2GHz bandwidth applies to 1 channel in one pair (10GSs) or 1 channel in each pair (a 'half channel'). If both channels in 1 pair are used (5GSs) or all 4 channels are used (2.5GSs) it drops to 1GHz per channel but as observed in my tests below, in real terms these are highly conservative limits.
WHAT I DON'T LIKE (remember, this is a DSO costing thousands when new):
There is a whole second of lag between rotating a channel offset knob and the channel actually moving up or down the screen, regardless of the number of channels displayed. Rigol boasts a 400k/s display update
rate but the ordinary scope controls run at a fraction of this.
The auto acceleration feature on the common command rotation knob (the screen intensity knob by default) is appallingly bad and when using it to move cursors, they fly all over the place at the slightest touch meaning constantly having to re-compensate when trying to slowly pan across the screen.
Why stop at 200ps/div? 100ps/div is entirely achievable with a repeating waveform sampled at 10GS/s.
The MATH feature that forms equations by combining the results of up to 4 channels, including MATH channels themselves, to produce a new math waveform is unnecesssarily complex recursion. Why didn't they simply provide single line equation entry like on the Pico Tech 2204A?
There is no provision to include integers in MATH equations. A workaround is to synthesise it using a dc level on one of the Generators and feed that in instead, but that wastes one of the only 4 available operators.
The offset knob rotates in the opposite direction for MATH waveforms.
When a MATH waveform first appears it is often outside the display area, but there is no way to easily bring it onscreen. The touchscreen doesn't do this. You have to guess the likely MATH result and enter an offset value on the manual keypad.
The MATH function is significantly lower resolution than the channel sampling rate and struggles to display anything above about 20kHz; a 100kHz signal results in a MATH waveform of only ~50 points across one full sine cycle. It's also very noisy. What should be a pure 10kHz sinewave coming out is 10% noise. If MATH can't handle something as simple as V/R, what is the point?
The storage method is abismal, with a virtual keyboard so small it seems to have been designed for a four year old's fingers. There is no magnify feature. It is incredibly easy to inadvertently exit the character entry box and be forced to start over. This also applies to creating and selecting storage folders. Worse, storage doesn't happen when you click Ok. You have to then explicitly click SAVE. Even then, the vaguely labelled controls don't do what you'd expect. The whole thing needs a complete revamp.
IN SUMMARY
All DSOs have issues and it's a shame Rigol could not have taken the time to resolve the simple issues above because in general the MSO8k is very useable and in many respects, outstandingly good. Its bandwidth exceeds even the 7104 and has revealed hidden faults, for instance the slow rising edge I observed measuring the 109 mod output (see bottom of [Projects: Tek 109 LF & 1-shot mod]) turned out to be a 7104 mainframe issue.
At 10GSs with 500Mps storage, a rise time of ~175ps and glitch detection down to 400ps, it meets all of my present and hopefully future needs. Bandwidth is astonishing: the measurements below were on one channel at the 500M max memory depth and max sampling rates. The blue trace is the FFT math line and its magnification is directly linked to the timebase setting.
My 2GHz-enabled MSO8k can display a 2.6GHz (avg) sinewave within 3dB at 200ps/div:
Beyond 3dB displayed loss the channel histogram can accurately display and measure a sinewave up to 3.7GHz (avg) at 200ps/div (the FTT feature is invalid at 200ps/div):
The FFT Math feature accurately measured (within 0.6%: 4.0749GHz) a sinewave up to 4.1GHz at 2ns/div
(the channel histogram GHz measurement is inaccurate at 2ns/div; the displayed waveform has amplitude-modulated distortion but is still recogniseable as a sinewave). As with most of these screenshots, the blue FFT trace is wildy expanded to just a handful of points on its line:
The MSO8k is not perfect and like the MSO5k it sadly also locks up occasionally. Its erratic manual cursor control and clumsy storage mechanism are unforgiveable, but it's probably the last DSO I'll buy:
It's a lot quieter than my old DS1102CD; its 2GHz+ bandwidth, 10GSs sampling, 500MB memory, 10" screen, USB I/O and tiny footprint far outweigh the equivalent bang for buck I can get for a modern Tek or HP.
Footnote:
I call him Squeaky Dave but I have the utmost respect for him; it's just his voice gets me every time...