7. Calibration
This section covers calibration instruments acquired to ensure accuracy of measurements, in particular for the LHC project.
VOLTAGE, CURRENT & RESISTANCE CALIBRATION
GENERAL RESISTANCE 'DAS' SERIES PRECISION VOLTAGE AND CURRENT SOURCES
I soon found myself grumbling at continuing issues with the General Resistance Dial-a-Source DAS-45 & DAS-56 precision dc V/I sources I bought in 2013/14/15, that at the time seemed a sensible choice:
General Resistance Dial A Source DAS-45 ±1µV to ±10V 0.015% 5ppm ±30mA
General Resistance Dial A Source DAS-56A ±1µV to ±10V 0.008% 2ppm & ±10nA to ±50mA
See [Repairs: General Repairs] for more details.
TIME ELECTRONICS 2003S PRECISION VOLTAGE SOURCE
The Time Electronics 2003S was my answer to the troublesome DAS V sources although the first one I bought in non-working state turned out to be far more ancient and damaged than I realised.
Time Electronics 2003S ±10nV to ±10V; <100mV ±0.05%, >100mV ±0.02%, 30ppm, output to 30mA
See [Repairs: General Repairs] for more details.
TIME ELECTRONICS 9822 MULTI-CALIBRATOR
In 2020 I came across a vintage UK Time Electronics 9822, 6.5 digit multi-calibrator on eBay UK that seemed a much better alternative. Looking back at all previous auctions logged on the net, I reckoned a bid of £410, the maximum I could afford, stood a good chance of winning it, especially as it was listed as collection only. Sadly I was proven wrong when it went for £532.
9821 on 21/04/15 sold for £270 (ebay auction)
9821 on 02/02/18 sold for £380 (ebay BIN)
5025 on 06/04/18 sold for £460 (ebay auction) The 5025 is a much newer multi-calibrator
9822 on 28/04/19 sold for £280 (Polish auction)
9822 on 03/09/20 sold for £532 (ebay auction - the one I missed)
This instrument covers 0-1kV ac or dc, 0-10A ac or dc and fixed resistances of 10Ω, 100Ω, 1kΩ, 10kΩ, 100kΩ, 1MΩ and 10MΩ. Ac can be any frequency from 15Hz to 20kHz as well as sine, square wave, sawtooth or rhomboid. There are three versions: the 9823 is the most accurate at 10ppm, followed by the 9822 at 30ppm and finally the 9821 at 100ppm.
HV is classed as anything above 40V and when running ac, only a sinewave is available from 40Hz to 400Hz. Other waveforms e.g. sawtooth and squarewave are available, but they don't go above 40V.
At the time I was only looking for a reliable dc source but after the auction finished I realised I had missed an opportunity: it isn't economically viable for me to have all of my instruments professionally calibrated. I checked a local company and found they would calibrate a 982x for an affordable £120 and realised I could then use it to calibrate the remaining instruments myself.
Something to be wary of: calibration companies often just check if an instrument is within its specification and if it is they then certify it as such without doing any more work.
Be sure to check the company is actually going to calibrate your equipment to the highest resolution it can meet.
The 9822 stood out as it was clearly a modern microprocessor-based design rather than purely analogue that might involve a lot of component replacement as it aged.
Even so, normally when selecting instruments I try to buy ones with schematics and I found these are rare for calibrators, perhaps due to their very purpose since they are usually very expensive. The 982x again proved to be an exception: whilst researching the 9822 I came across the xdevs website describing the repair of a 9823, with schematics and code dumps provided: https://xdevs.com/fix/te9823/
This largely cemented my plan to acquire a 982x but by the summer of 2022 none had come up on eBay. However in past conversation I had found a seller with a faulty one that I now bought for £500, marginally less than the one I missed. The fault was strange: any programmed voltage was too low and its polarity repeatedly swapped, and any selected HV ac voltage was always clipped to a square wave.
I thought it was going to be an absolute pig to debug and fix as the schematics I'd downloaded from xdevs appeared incomplete. In an attempt to familiarise myself with it I decided to first try his fix, also suggested by the seller, which was to replace the 1.5MHz 68A21 PIA chips wihich have very old manufacture date codes: 8550 & 8608 (i.e. 1986 week 08).
To my utter surprise and delight, this instantly fixed all problems. I emailed the author Illya Tsemenko to update him (he's since donated his 9823) and his response was he has found other instances where 68A21s have failed. I'm wary of believing this as Motorola is a pretty solid manufacturer. I'm more inclined to think there may be an issue with the design of the 9822.
Aside from this, it's good news all round as it seems to work perfectly. A down side is the interface is a bit confusing as voltage and waveform selection share the same LED display and so are only displayed independently. If you make a mistake it displays OP ERROR and the only way to exit from this is to turn the thing off, wait a little while, re-boot and start over.
May 2024:
An assumed capacitor in the IEC input filter gave up the ghost going 'Pffft!' whilst emitting a foul potting compound stench, necessitating its replacement, Schaffner FN372-4/21. The only place I could find one was Digikey.co.uk who wanted £18 + £12 post which I thought was extortionate. It was only when it arrived 2 days later I realised the high postage was because it was posted directly from the USA.
I also replaced the Papst 8560N (outlet) fan and filter that I'd bought for it when I first got it as it was unpleasantly noisy. Unfortunately it's still noisy (par for the course with Papst) but at least the bearings are tighter on the new one.
This repair turned out to be more scary than expected because a contact snapped off the PSU floating chassis connector that leads to the fan and I had to retrofit one from an unused pin (new connectors are at least £80: £62.13+post+vat LMRS24DP | Weald Electronics | FC Lane Electronics). Having also observed a break in one of the wire insulator sleeves, I conclude the fan was replaced before and the heavy handed repairer had both severed the insulation as well wiggled the contact so much that it parted company when I came to remove the wire again.
The same day I discovered a higher speed 2MHz 68B21P and bought a couple from eBay China for $5. I'm hoping the '0726' stamped on them is the manufacture date code for 2007 week 26, which should extend the 9822 life beyond the 68A21Ps I fitted as, they don't seem to have been made after the 90s.
I intend to replace the HV standard bananas with HV safety types and then get it calibrated.
XG-ecu T48 IC PROGRAMMER
I bought a cheap Chinese IC programmer to take copies of the 9822 code Eprom, ac waveforms Eprom, and calibration EEProm, and I can now do the same for other instruments such as the Tek 7854 mainframe.
XG-ecu T48 USB2.0 480MHz universal programmer ZIF-40, http://www.xgecu.com/en/
The T48 is the 2022 update to their popular TL866II programmer with support for 20k+ devices although as with all programmers, many of these are just repeats from different manufacturers:
http://www.xgecu.com/MiniPro/T48_List.txt
I have many older devices it doesn't support, but having long ago moved on from separate Eproms & µCs to PIC µCs with onboard ICP (In-circuit Programmable) code space, these are unlikely to be used again. Even in that era I avoided the purchase of a prohibitively expensive standalone programmer by building my ISA card based [WCS: Writeable control Store]. For these reasons I never got around to buying a programmer until now.
My 9822 memory chip dumps are below (Wix won't upload HEX or BIN files so they are both in these RARs):
Next I shall replace the irritatingly noisy and clearly worn out Papst 8560N fan (a new one was £73 from eBay USA; be wary of Chinese companies selling 'refurbished' ones), and swap the HV output standard 4mm banana sockets with modern safety alternatives.
CALIBRATION SHUNT RESISTOR
I have an Amrel PPS-1202 dual 0-16V 0-4A programmable power supply with a resolution of 5mV.
I only paid £210 for the entire PSU including shipping. Given the cost of a yearly professional calibration is fairly similar, I've been investigating alternative options.
COMMERCIAL PRECISION CALIBRATION SHUNTS
Calibration is simple if you can find an affordable ±0.001% 10A 100mΩ, 20ppm, 10W resistive shunt. This is of course in the land of calibration standards costing $1000s:
'I've seen versions with 0.001% with matching price $6500'
https://diysolarforum.com/threads/shunts-can-it-be-to-small-or-too-big.13924/
Dropping an order of magnitude to ±0.01%, in 2024 the Burster 1240 series costs €315+ 19% VAT = €375 + shipping (probably €30, total est. Feb 2024 €405/$440/£345): https://www.calplus.de/burster-1240.html
but this is still far too high.
0.5% 100mΩ 10W resistors are also available (Rapid Electronics: Isabellenhutte PBV-R1-F1-0.5 0R1 ±0.5%
4-wire Precision Resistor £31). Again not cheap (Burster is 10ppm/K, Isabellenhütte is 30ppm/K) although drift necessitating repeated calibration has been reported, see comment by tszaboo:
But this is moot: 18V max output x 0.5% is 90mV, far exceeding the 5mV resolution. OTOH the specified ±0.001% at 18V is ±180µV, which seems equally excessive being 28 times less than 5mV, and 4A x 0.001% is 4µA, which again seems OTT. If it was an ac signal I could understand, but this is simple dc.
CAN I MAKE ONE MYSELF?
The following table and summary is from:
https://www.arrow.com/en/research-and-events/articles/shunt-resistors
Thermal ElectroMotive Force (EMF) of various metals vs Copper, µV/°C and
Thermal Coefficient of Resistance (TCR) in ppm/C across various temperature ranges
Room temp
Metal/Alloy µV/C -55°C to +25°C 0°C to +25°C +25°C to +60°C +25°C to +125°C
Aluminium -4.0
Cupron -45.0
Evanohm +2.0 +5.0 +2.5 -2.5 -5.0
Gold +0.2
Manganin -3.0 +50.0 +10.0 -5.0 -80.0
Nickel -22.0
Silver -0.2
Zeranin -1.3 +20.0 ±2.5 ±5.0 +10
'Based on thermal EMF, TCR, and cost trade off, manganin is the favorite choice for shunts with exposed blades. Shunts with exposed parallel wires are made with zeranin, a relative of Manganin with a lower temperature coefficient. Shunts enclosed in heat sinks are generally made with Evanohm, which has a near-zero temperature coefficient with high sensitivity to strain.'
Manganin:
https://en.wikipedia.org/wiki/Manganin
Zeranin:
https://www.isabellenhuette.de/en/precision-alloys/products/zeraninr
Evanohm:
Manganin is more readily available than Zeranin or Evanohm.
The following table indicates #12 AWG ~2mm diameter wire will carry 10 Amps:
https://www.jst.fr/doc/jst/pdf/current_rating.pdf
HOW MUCH MANGANIN DO I NEED?
The diameter and length of wire determines its resistance. The larger the wire diameter the lower the resistance, necessitating a longer wire that is easier to cut more precisely.
On eBay, Manganin wire thicker than 0.3mm diameter seems to go for very high prices. The following UK distributor sells 1m lengths of 2mm diameter Manganin wire for £7.19 (scroll to the bottom):
https://www.wires.co.uk/acatalog/manganin.html
Length can be calculated from L = RA/p, where
L is the length in metres,
R is its desired resistance, 100mΩ
A is the area of the wire cross-section in m²
ρ is manganin resistivity constant, 482x10⁻⁹Ω⋅m
d is the wire diameter, 2mm
L = RA/p
A = area = π x (d/2)², for 2mm A = π x 1x10^-6 = π x 10^-6 m²
So L = (1 x 10*-1Ω x π x 10^-6) / 482x10^-9Ω⋅m = 0.651783m
The following calculator can be also used (although resistance is rounded to a low precision):
https://www.basictables.com/electronics/resistor/wire-resistance
L, calculated length = 0.65178m or 651.78mm
Let's try rounding it up to make it an easier measurement to cut:
Rounding up to 652mm / 651.78 = 0.0337% x 18V is 6.074mV x5 = 30.37mV which is too high.
Rounding up to 651.8mm / 651.78 = 0.00307%
18V at 0.003% would give 552µV x5 = 2.762mV which is acceptable.
Manually re-calculating resistance for 651.78mm length of manganin:
R = ρL/A = (482x10⁻⁹Ω⋅m x 0.65178)/ π x 10^-6 m² = 99.99958448mΩ.
HOW TO MANUALLY CUT IT ACCURATELY?
However 0.00307% x 651.78 is still only ±0.02mm which is a precision beyond my abilities to cut.
Several 2mm wires in parallel increases cutting precision because twice the length is needed to create the same resistance. Two 2mm diameter wires would need to be 2m long each to achieve the same 100mΩ.
The cost quadruples to £7x4=£28 but the precision only doubles to ±0.04mm which is still out of reach.
Can we relax the tolerance?
18V at 0.05% gives 9mV, unacceptable.
18V at 0.005% gives 900µV which is 18% of 5mV which I'd be happy with.
18V at 0.008% gives 1.44mV which is 30% of 5mV and which I'd still be ok with.
18V at 0.010% gives 1.80mV which is 36% of 5mV and I'd accept that too: PSU resolution is 5mV, not 1mV.
From the previously calculated value for a single 2mm wire, ±0.01% of 651.78mm = ±0.065178mm
The wire would need to be cut within the range 651.715 and 651.85mm. Still beyond my abilities.
How about wires in parallel as well as the reduction in tolerance?
With two 2mm wires in parallel the tolerance is ±0.13mm, manganin cost is £7.19 x 4 = £29
With three 2mm wires in parallel the tolerance is ±0.20mm, manganin cost is £7.19 x 9 = £65
With four 2mm wires in parallel the tolerance is ±0.26mm, manganin cost is £7.19 x 16 = £115
±0.26mm might be achievable but £115 is a lot to lose if I cut it wrong. Could I make an adjustable high current contact?
Precisely measuring long distances is also difficult. The only practical route would be a bobbin. Will the accompanying inductance be an issue for calibration?
Manganin is also available in sheet form. Would this be a better route? Unfortunately I've seen no UK distributors with small quantity prices. Potential sources are India and China. More research needed.
IS THERE A MUCH CHEAPER ALTERNATIVE APPROACH?
Maybe I can junk this whole idea and just use a 1% 100mΩ aly resistor I have and fudge the cal.
Amrel says a 5.5 digit dmm is all that's needed to read off the shunt voltage for its cal feedback entry. I can calibrate the psu with this resistor on a regulated TEC at say 20°C, then use my 5.5 digit [Other Instruments: HP3478A dmm] to see how far out the voltage is, then repeat the cal procedure N times tweaking the cal feedback value until I get the PSU output spot-on. It would take a few iterations but would cost nothing. The dmm will already be calibrated using my professionally calibrated Time electronics 9822 multi-calibrator, the reason I bought it.
Assuming the maths is linear, it might be possible to shorten this process by measuring the resistor with the dmm to determine its deviation from 100mΩ and adjusting the cal feedback value by the ratio.
However the first move will be to look inside the PSU and see if there is an EEP, especially one in a socket, because I'll take it out and save the contents with the T48 programmer before fiddling with cal.
Amrel PPS-1200 series:
https://www.cyfronika.com.pl/dok/pps.pdf
PPS Manual with spec on pages 49,50 [44,45] and calibration procedure on page 37 [33]:
OSCILLOSCOPE VOLTAGE & CURRENT CALIBRATION
TEK 067-0502-01 TEST FIXTURE
This can only drive a 1MΩ load, but provides all typical oscilloscope voltage div settings:
Tektronix 067-0502-01 Test Fixture 100µV to 100V ±0.25% into 1MΩ + 5mA calibrator)
DIY RUBIDIUM FREQUENCY STANDARD
Late 2022 I bought a used Datum Efratom LPRO-101 10MHz Rubidium Frequency Standard for $200, after watching the following video: https://www.youtube.com/watch?v=zW5ffFuEQsw
I put it in a box with a PSU, see [Projects 2: DIY Rubidium Oscillator]
I'll use this as the reference frequency for all of my test equipment that accepts an external source, and as an absolute frequency reference for all precision timing.
RACAL-DANA 1998 FREQUENCY COUNTER
Shortly after I bought the Rubidium oscillator I realised the Signal Hound spectrum analyser I've been using as my frequency counter only displays MHz down to the LSD and I needed more accuracy to assist instrument calibration with the new oscillator. Such accuracy is usually very expensive but by chance I found a 1986-era 10.5-digit mil-spec Racal-Dana 1998 1.3GHz frequency counter with option 4E ultra high stability 5x10^-10 OCXO, tested and working on eBay UK for £260, and bought it. Option 4E is only bettered by the rare Option 04R, ironically a Rubidium oscillator offering 5x10^-11.
The following link to EEVBlog describes its TEC feature that adds another 2 digits of accuracy:
https://www.eevblog.com/forum/chat/frequency-counter-resolution/
Whilst doing some background research I discovered this excellent teardown by Gerry Sweeney:
https://gerrysweeney.com/racal-dana-1999-teardown-repair-and-calibration/
Towards the end, he uses a 150MHz Tek 2445A to fine tune the 1998 OCXO by comparing its 10MHz output with the 10MHz output of his Rubidium oscillator.
A more accurate method is to feed the signals into X and Y and create a lissajous circle:
https://www.testandmeasurementtips.com/using-scope-display-lissajous-patterns/
https://www.youtube.com/watch?v=Wq3v-WN_hjo
The March 2011 issue if VHF Communications magazine features the 1.3GHz Racal 1991 on its front cover together with an article giving details of how to convert it to 12GHz. I wonder if this could also be fitted to the 1998? https://worldradiohistory.com/Archive-DX/VHF-Communications/VHF-COMM.2011.3.pdf
BALLANTINE 6130A TIME MARK GENERATOR
This produces 5s to 100ns pulses & 50ns/10ns/5ns/2ns sinewave, similar in functionality to the Tek 2901: https://w140.com/tekwiki/wiki/2901
https://www.radiomuseum.org/r/ballantine_time_mark_generator_6130a_101_anusm_441.html
https://bama.edebris.com/download/ballantine/6130A-time-mark-generator/6130A%20Schematics.pdf
Footnote:
I call him Squeaky Dave but I have the utmost respect for him. His voice gets me every time...
New text box
µ Ω ± ° ⌠ ⌡ ∫ │ ─ √ φ θ Θ ∂ δ ζ ξ ς λ ψ ω τ µ Ω ∆ Δ ∑ ∏ π Ξ ○ ≠ ³ ² ±