1. General Instrument Repairs, Modifications and Teardowns
Action Status Issue Manufacturer Model Function
Repair Fixed No o/p General Resistance DAS-45 Precision ±10V Source #1
Repair Ongoing Erratic o/p General Resistance DAS-45 Precision ±10V Source #2
Repair Ongoing Erratic o/p General Resistance DAS-56A Precision ±10V & ±50mA Source
Repair Fixed Bad IEC filter Hewlett Packard HP4952A Protocol Analyser
Repair Ongoing Bad delay ctrls Hewlett Packard HP8082A 250MHz Pulse Generator #1
Repair Fixed Noisy controls Hewlett Packard HP8082A 250MHz Pulse Generator #2
Repair TBD Only ±1.7kV Leysop M5000 ±2.5kV EO Differential Amplifier
Refurb Done Stiff crocs Peak DCA50 Component Analyser
Repair Ongoing Bad ext i/ps Pulsetek 233 Dual 50MHz Pulse Gen
Repair Ongoing Bad controls Pulsetek 240 Single 50MHz Pulse Gen w. adj Tr & Tf
Repair Fixed Abused by owner Texscan RA5030 0-1GHz 0-50dB Variable Step Attenuator
Repair Ongoing Abandoned Time Electronics 2003S dc Voltage Calibrator 0.02%
Repair Fixed Erratic Time Electronics 9822 Multi-Calibrator 30ppm
Repair Fixed Missing dowel Thorlabs Si254 Shear Plate Assembly
Mod Done Bad power lamp Wavetek 154 Programmable Waveform Generator
GENERAL RESISTANCE
Judging by some comments I've seen in discussion groups, General Resistance instruments seem to be held in high regard but I am less impressed. The ones I have are all problematic and the deeper I've gone into their screened boxes the more hidden horrors I've discovered. This is particularly true of the
DAS-46, see Teardown below.
General Resistance Dial-A-Source DAS-45, constant voltage (CV) ±1µV to ±10V, 0.015% 5ppm
I have two of these and a better spec DAS-56A. They all seem to share a similar design regardless of enclosure, consisting of a bank of thumbwheels with precision resistors nearby, and a mainboard with a large packed daughterboard with a couple of LM308 op amps. The daughterboard is attached to the main board using pins in individual sockets. There are identified trim pots on the main pcb, but twiddling these I found it impossible to calibrate finely enough to ensure the LSD is stable, however 4 digits is enough for me. The adjustment pots (light blue in the photos below) affect each other so calibration was largely about compromise to ensure all ranges were reasonably accurate.
Note - I attempted the above long before I came across a service manual, found here:
http://bee.mif.pg.gda.pl/ciasteczkowypotwor/General%20Resistance/General_Resistance_DAS_46_56_57_86_DAS_Series_Dial-A-Source_Operating_Instructions_wSchematics.pdf
I have yet to repeat the process (I tend to re-calibrate instruments when I'm about to use them, as they often get left long periods without use). I'll update this if I find I was doing something wrong.
General Resistance Dial-A-Source DAS-45 #1
(a) Initally worked for a few weeks then the output failed.
Fixed by replacing the 2N2905 PNP output transistor and adding some Arctic Silver TIM (Thermal Insulating Material) aka paste between it and its heatsink.
General Resistance Dial-A-Source DAS-45 #2
(a) This has an intermittent fault; the output is either correct or far too high with HF ripple superimposed, rather like op amp feedback incorrectly compensated, but cursory investigation turned up nothing obvious.
It may be worth re-examining the voltage reservoir cap ESR:
https://theloopytech.wordpress.com/2015/10/17/misbehaving-frequency-counter-racal-dana-1998/
General Resistance Dial-A-Source DAS-56A CC 50mA & CV ±1µV to ±10V, 0.015% 5ppm
(a) This was expensive, bought as brand new old stock, but has never worked properly. The output is erratic, sometimes it displays the selected value to 5 digits, other times it's wildly wrong. Flicking the polarity switch sometimes cures it, as does power cycling, but neither works all the time. and there is nothing obvious to suggest the failure mode.
I've had it fully apart once in an attempt to convert it from US 110Vac to UK 230Vac mains operation but had to abandon this due to poor insulation practices on the mains transformer hidden in a bare metal shielding box.
Other nice little gotchas are: uninsulated wires soldered to feedthrough bulkhead capacitors emerging with live mains on them from the inlet filter; uninsulated live mains on a bare strip connector screwed to the base of the bare metal box, and bare mains on the front panel mains toggle. Soldering on the KVD switches is also pretty diabolical, and the handled metal lid scrapes some of these when you remove or re-attach it.
Poor soldering is not the same as untidy looking circuitry, often beneficial to noise reduction. For good examples of this skilful art, look at Jim Williams' AN-47 Appendix F [J1].
I have found a general rule of thumb is, if the company isn't that well known and it looks professional on the outside, it probably isn't inside. Although the performance-lauded classic QUAD 33 from famous British manufacturer QUAD Electroacoustics managed to achieve both using loose rows of pushbuttons with so much vertical play all adjacent lettering was misaligned, in an amateurishly futuristic simplistic orange and grey box; looking inside, I found components thrown into their positions without a care.
HEWLETT PACKARD
HP4952A Protocol Analyser
When one of these appeared for £50 I bought for fun it as a memento of the era I worked in. Given the freely available PC-based apps now doing largely the same thing much more easily, there is otherwise little point having one. As if to prove the point, the one time I did try to use it, it let me down.
Whilst attempting to log temperatures on the Tek 109, see [Projects: Tek 109 1-Shot & LF Mod] and discovering my PC serial port refusing to communicate with my Pico TC-08 temperature datalogger, I thought I could have a bit of fun with the 4952A. This was short-lived when a few minutes after I powered it up, I was met with a puff of smoke and a ghastly stench that took several days to vent out of the lab. Weirdly the CRT was still working. Of course I shut it down immediately. Perhaps it was mocking me for trying to use it for what was a trivial loss of a signal later found almost instantly with a scope probe.
When the 109 mod was complete, I returned to investigate the 4952A. Inside it's crammed full of pcbs, see below. Nothing was immediately amiss, so I removed the top 2 pcbs and on the bottom pcb I spotted black gunk over a couple of MOSFETs located close to the main IEC connector on the rear panel.
My immediate thought was the MOSFETs were dead. It wasn't possible to test them in-circuit, so I unsoldered them but they both checked out fine and once I'd cleared the gunk off the pcb, there was no sign of damage there either.
Peering in more closely at the filter on the back of the IEC mains inlet, I noticed traces of black gunk on the metalwork below, and a tiny hint of it on the bottom of the filter itself. It dawned on me a capacitor inside it must have blown, melting and ejecting the black potting compound outward with considerable force (and smoke) to cover the nearest components, which were of course the MOSFETs.
No-one stocked the original IEC, fuses, switch and Schaffner filter combo but I did find a slightly different filter in the same arrangement. I mused any filter was better than none since the 4952 would rarely be used; I just wanted it working again, and indeed once retrofitted, it came back to life.
Pondering why it had exploded in the first place, I worried I might not be so lucky with other older instruments that were in use, convincing me to look at stabilising the lab mains voltage close to its nominal 230Vac instead of the 250Vac it now seems to run at, see [Variac Repair & Add Meters].
HP8082A 250MHz Pulse Generator #1
(a) External trigger N/S, suspect Q10/Q11 on A3 Rep Rate board, SM CCT P.92, Assy P.89, BOM P.81. LIBS4 P.109. Q10/Q11 = HP1853-0284 = SMT1114 = Si PNP 10Vce Ic30mA 225mW, NSN 5061-01-261-4649 in a microwave transistor shaped plastic package. Found just 2 on eBay, $25ea including shipping to the UK.
It seems odd to me they chose what looks like an RF part when it doesn't have to operate at very high frequencies. The package strikes me as inconvenient too, it's a surface mount device with a raised profile on both sides. Presumably to limit/match inductance, capacitance, HP flipped one of them but had to drill holes in the pcb to accept the raised surface on both sides.
Unsoldering two of their pins and testing them revealed Q11 was in fact the culprit - the DCA-75 thought it was an LED. The other one checked out fine with hFE (dc gain) 76. The two I bought had hFE 60 and 137. Given the circuit topology of a differential long tailed pair that constrains the input to the setting of the input level pot on the front panel, I thought it wise to match Hfe as close as I could to Q10, and replaced Q11 with the hFE 60 transistor.
Next, the all-important test and verification of the Ext Input. The HP8082A only goes down to 1kHz. I need to feed a 1Hz pulse from the Wavetek 154 [I40] into the Ext Trig for the HP8082A to generate a short pulse with controlled rise time in the order of nanoseconds. This pulse will be used to trigger my experimental Pockels cell avalanche driver (details to follow in the Q-switch section).
Right:
Top trace is 8082A
20ns 7.4V output.
Bottom trace is Wavetek 1.5V pulse into 8082A Ext Trig
Left:
The Wavetek (top) feeds into the 8082A Ext Trig.
Right:
Wavetek at arbitrary
255Hz proves there are no missing pulses
Although Q11 blew and not Q10, it bothers me the Ext Trig input can only handle up to ±6V when both the HP8082A and the Wavetek outputs can go up to ±10V if offsets are added.
Given the rarity and high price of the SMT1114, I would not want to inadvertently blow Q10 up. I didn't want to muck around with the internal pcbs so instead built a tiny external voltage limiting box:
HP8082A DIY Input Voltage Limiter
Initially I worked on the assumption I would allow a couple of volts into the trigger and selected a 1kΩ input resistor and a 2V0 zener diode to limit the input, forward fed by a fast diode, a known way of countering a zener diode's high shunt capacitance, but found this introduced ±400mV spikes at the transitions. Adding a 180pF capacitor took them out but seemed counter-productive.
Given the HP8082A ext trig input can work down to just ±200mV, I decided a better solution would be a couple of diodes back to back. I selected fast diodes to begin with but they also introduced noise and slower diodes were little better.
Putting the zener back in to experiment further, I inadvertently put it in back to front and found it produced a better result than ordinary diodes, introducing a 30ns rise time and 800ns fall time. The
Tek 571 curve tracer revealed its VF around 780mV at 5mA. I added a diode across it to counter negative spikes, a 1N4007 proving to have the best characteristics. This circuit worked well but the edges still overshot by about ±200mV. The HP8082A has a claimed input impedance of 50Ω on its Ext Trig although its schematic shows a 75Ω resistor to 0V, R10, in the input. Regardless, when I added a 50Ω terminator to the output of the limiter, the overshoots disappeared.
I tried it on the HP8082A and found the same result. Happy with the solution, I put it in a 2" long Pomona metal BNC box, but it stuck out from the instrument and I was worried I would catch it and damage the Ext Input BNC. As a final solution I chopped an eBay China 2" long aluminium box in two and made a 1" cube BNC box for the limiter that is short enough not to protrude excessively when in place. I can now run the Wavetek 154 full output voltage into the HP8082A Ext Input without fear of damage:
Add photo 701jH25 of 1" on 8082A
(b) 2022 - Many HP8082A front panel controls were erratic so I stripped the front panel and found the slide switches are actually constructed on the pcb and there is nothing to replace. Switch cleaner was the only remedy I could apply and amazingly it did fix most things, and an ultrasonic bath removed unsightly gunk from all knobs, but the delay switch and its rotary pot have limited control which I assume is down to another issue on the mainboard. I bought another 8082A for £50 as a cheap source of parts but found it actually works better, so now the original 8082A is the backup / parts supply, although both generators do work otherwise.
Bizarrely the variable rise and fall times don't appear to work correctly on either unit on the 2ns pulse width setting. I'm not sure if this is a fault or a design flaw.
Leysop M5000 OE High Voltage Differential Amplifier
At present the outputs won't go higher than ±1.7kV.
Leysop ignored my request for a service manual but in 2022 I found
their website now has a link to a user manual complete with
circuit diagrams, see below.
http://www.leysop.com/m5000.html
The ±2.5kV Outputs are fused at 50mA max output current.
Manson EP-925 linear PSU 3-15V 0-25A
This worked hard for a couple of decades but eventually the output voltage began to drop. Now it only reaches 9.7V regardless of output current. The design is not the best, both mechanically and electrically; the fan on the back blows onto the big transformer completely missing the big heatsink at the front. The following webpage points out the circuit sets itself up for failure and suggests improvements, complete wirh extensive schematics:
https://pa0fri.home.xs4all.nl/Diversen/EP925/ep925eng.htm
However it's a pig to modify or repair as the output transistors are embedded within a large heatsink behind the front panel and the man pcb is located on the back of the front panel, mking it difficult to access the components. For this reason I have so far off repairing it. In 2024 I picked up a used one on eBay for £60 which is possibly more than I paid for mine new.
Repair videofor one outputting too high a voltage (20Vdc):
https://www.youtube.com/watch?v=plA_8Erz3R4
Discussion and fixes for various issues:
https://www.vintage-radio.net/forum/showthread.php?t=154229
https://www.vintage-radio.net/forum/...t=Manson+ep925
https://www.vintage-radio.net/forum/...ghlight=Manson
https://www.vintage-radio.net/forum/...d.php?t=104212
Specification / Manual / catalogue here:
https://www.manson.com.hk/product/ep-925/
Peak DCA50 Component Analyser
This has worked well for at least a decade.
(a) The crocodile clip insulation hardened and the clips didn't always grasp component leads.
I stuck a wire wrap socket to it using double sided tape and attached the croc clips to its pins.
Eventually I replaced the croc insulation by buying cheap eBay croc clips, unsoldering the leads inside the DCS50 and swapping the hardened insulation, keeping the original gold-plated clips.
Pulsetek 233 Dual 50MHz Pulse Generator
(a) The EXT TRIG / EXT GATE input, LED and controls don't work.
Pulsetek 240 Single 50MHz Pulse Generator with adjustable rise and fall time
(a) Many controls are erratic. Switch cleaner fixed most of them.
Texscan RA5030 0-1GHz 0-50dB var atten repaired
Damaged due to being opened by somebody unfamiliar with the assembly construction.
Usually when I try to dismantle an instrument I first think how the external connectors might affect its construction. The last person who dismantled his clearly didn't do this.
It's tempting when noticing 3 screws at each end of its metallic tube, to think all that is needed to pull the unit apart is to undo them, with no consideration for the two BNC connectors located along its length. Of course any connections to these must first be disconnected.
It turned out the reason why this didn't work was precisely because the last owner didn't think about this. Result: the BNCs jammed and bent the EMI fingers on the intermediate stages.
Time Electronics 2003S dc-Calibrator 0.02%
Very old (pot dated 1977?), batteries heavily corroded & pcb also very tarnished & filthy.
A few screws & washers missing
Newer? version fixed here (excellent video teardown & repair):
https://www.ianjohnston.com/index.php/videos/144-video-blog-113-2003S
Never occurred to me to use ultrasonic cleaner on pcb - did so for mine too.
He also ffers an excellent 20-bit DAC-based 0-10Vdc cal for £325 in 2023 (veery good value but a bit too expensive for me). Hopefully this much cheaper 2003S can be repaired. Not as good a spec, but enough for my needs.
Time Electronics 9822 Multi-Calibrator 30ppm
Erratic controls turned out to be the Motorola 6821 again:
Thorlabs shear plate
One steel dowel was missing from one of the three pairs that align with ball bearings on the plate. Replaced the missing dowel. The dowel I used is the same diameter but very slightly longer than the original one used by Thorlabs.
Close inspection of the original dowels reveals they have been glued into position but don't all lay flat on the metalwork. This is actually not a problem because the ingenious design relies on the shear plate ball bearings aligning as long as the distance between each pair of dowels is precise. Disc magnets below the dowels hold the steel balls and thus the plate, in position on the non-magnetic aluminium assembly.
https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=2970
Wavetek 154 Programmable Waveform Generator [I40]
(a) The power light broke but it's a mains voltage sealed incandescent lamp on the front panel. A quick and dirty fix that didn't involve opening the unit, was to run a short BNC cable from the AUX output connector on the back to the front of the unit, mate a cheap BNC to a RCA phono adaptor and a cheap phono plug inside, with a 10kΩ resistor and a low current LED poking out through the lead hole.
An added advantage is the LED doubles up as a frequency indicator.
TBD [I]
(a)
HOME BUILT EQUIPMENT
8 decade resistance box, 1%, 2W, 100mΩ to 11MΩ
Typically these are expensive to buy and most have fewer decades, although usually the tolerance is better, often 0.1% and lower wattage. I included a 0.1% range with higher wattage. Speaking of which, I considered the 6 decade Clarostat 225 Watt high power resistor box, then discovered from its datasheet (see below) that the manufacturer sneakily uses fewer resistors, having worked out a series/parallel circuit that uses less parts. This explains why they rate it 'up to' 225W rather than absolute. As far as I can tell, competitive manufacturer Winslow offers a cheaper but otherwise identical model 336. My own box has nowhere that power of course, but it does have 8 decades with a full resistor complement.
See [Projects: Eight Decade Resistance Box].
Twin power lead extender for Tek P6201 900MHz FET probes
See [Projects 2: Tek P6201 Probe Adaptor].
DIY Tek 067-0616-00 7k series plugin flexible extender
Based on aftermarket pcbs sold by John Griesson, who is also Ecosensory.com:
http://ecosensory.com/tek/tek_7K_flex_sales_blurb.txt
text