Valhalla 2724A Programmable Resistance Standard

[branadic]

Part [1]

I recently bought a Valhalla 2724A, that was offered as parts/repair for a reasonable low price. Turned out it has the CPR (cardinal resistance reference) option installed, the display is as good as new and the GPIB option (TL-1) is installed too and it comes with Option RX-3, the rack ears. 
Unfortunately, the calibration key is missing, but thanks to the Valhalla 2720GS repair and upgrade plus the support by Kosmic I have a US style copy of the original key and managed to get a EU style copy of that done too.

The unit works, but boots up directly with a flashing * without anything connected to it. The manual tells us: "This indicates that the applied test current exceeds the maximum allowed for the resistance selected. If connected to an auto-ranging ohmmeter, this will cause the ohmmeter to increase its range. If using another type of current input, the level must be decreased and/or the resistance value decreased."
Okay, that something we can work with.

The unit is build fairly simple, so it shouldn't be too hard to fix it, right!? The date codes on the chips range from '84 to '86, so the caps in there are almost 40 years old. Yes, still the original ones. A quick check at Mouser revealed that almost all of the original part numbers are still available today and for the price of 60 € all electrolytics and tantalum caps can be replaced (Mouser parts list attached), I'm sure worth the money. *Replacement caps ordered*

-branadic-

Source of images: messgeraetemarkt

[branadic]

Part [2]

In case you need it, attached is the EEPROM hex file (M5L2764k).

-branadic-

[Kosmic]

Haaa lucky you, look like a nice unit. 

The datasheet for those interested: Valhalla 2724a Datasheet

[branadic]

Part [3]

The good news is, the caps arrived and were replaced, with the flashing * disappearing for the moment afterwards. The bads news is, someone has been in the unit before and messed with C109, lifting traces and the associated Vias and has maybe replaced two other components as well.
Unfortunately, I haven't noticed that before, but traces around C109 have been fixed now.
However, the unit does not produce any proper resistances just yet. So I have to dig into it a bit deaper and check some voltages next.

Edit: The missing page 1/7 of the manual is available at ko4bb.com

-branadic-

Source of images: messgeraetemarkt

[Doctorandus_P]

And the manual:

https://valhallascientific.com/wp-content/uploads/2724A_Manual.pdf

Chapter 5 is about theory of operation, but it does not have a schematic.

Service manual with schematics is at:
https://www.opweb.de/en/manufacturer/valhalla-scientific-inc/2724a

I'm aware of how generic resistance multipliers work, but that they can get this to 6 digits of resolution / accuracy is out of another world. And this thing looks like it's 40 years old.

This reminds me of the Saga of Biorn. He wanted to go to Valhalla too.
https://www.youtube.com/results?search_query=saga+of+biorn

[branadic]

The service manual you've linked to is missing sheet 1/7 but again contains 2x sheet 2/7.
Attached it the corrected manual, that now contains the missing part of the schematic.

-branadic-

[branadic]

After a while the flashing * reappeared.
During some pocking of the unit with the thermal camera I've decided to replaced the +18 V (IC602), -18 V (IC603) and +5 V (IC605) voltage regulators in the power supply section with ST L7xxx and 78L05 (IC105).
I've found residue of some sort underneath the heatsink of IC603 that was furthermore touching a via underneath it, maybe creating some leakage? Some Kapton tape is now covering the via.
Having done that the unit boots, the flashing * has vanished from the display and the unit is outputting resistance values for the very first time that are plausible measuring with some handheld multimeter in 2-wire mode. *progress*
IC605 is still getting a bit too warm (~65 °C) and I have to figure out why. Maybe that is normal behavior as it creates +5 V from +18 V, but that is for the future me.

-branadic-

[TheDefpom]

It’s all your fault….

I’m now the owner of a 2724A… it’s broken, apparently, which is exactly how I like to buy test gear.

I plan to try powering it up for the first time and doing some video on it this weekend.

[branadic]

I don't feel guilty

So you can verify the temperatures of the regulators with a thermal cam for me, that is a good deal.

-branadic-

[TheDefpom]

Might be an idea to edit your post above about the HEX file only being for the CPR version.

I read in the manual (I was only skimming the diagrams and parts lists, so it doesn't count) that the parts list has different ROM versions for the CPR and non CPR version, mine is the non CPR version.

[TheDefpom]

Well I couldn't help myself and I started looking at it today... lots of magic smoke, yay !

Once I fixed the smoke generator, the V regs after about 5 minutes of being powered up are:

IC602 - 49 degrees
IC603 - 38 degrees
IC605 - 44 degrees

The processor is about 60 degrees.

[TheDefpom]

Here is the EPROM data for the standard (NON CPR) version that I have, Version FB dated 28th August 1989.

[TheDefpom]

Branadic, how stable is your resistance once set, for instance, is it drifting up and down when you have it set at things like 1 Ohm, or 10 Ohm, my unit seems that the last 2 or 3 decimal places are guesses.

The most stable mine gets is at 1K.

I tried it on my Datron 1082 and that REALLY doesn't like the 2724A for certain resistances, for instance setting 100K gives my ~120K and anything above that just gets garbage, 1K is pretty good but drifts 10 or 20 counts.

My Advantest R6581T likes it a lot more, it works right up to 1G, but the drift is quite obvious making the last 3-4 digits basically meaningless.

I tried different cables with no change, I have some nice Pomona cables and some I made which are low leakage, both gave the same result, so I am basically wondering if I have more issues with my unit apart from the smoke generator it came with.

I am really not sure what to expect from it, as I cannot find any videos etc showing the usage and how the multimeter is reading the resistance.

This could just be a sign of dirty relays, or there is some kind of control loop problem.

It is more stable in fast mode, slow mode drifts a lot more, but I am also getting a resistance offset between fast and slow mode too, about 300 counts, so I am not sure that is right either, again this might point to a relay issue.

[TUMEMBER]

My Advantest R6581T likes it a lot more, it works right up to 1G, but the drift is quite obvious making the last 3-4 digits basically meaningless.



I am really not sure what to expect from it, as I cannot find any videos etc showing the usage and how the multimeter is reading the resistance.

I don't know if Datron uses "voltage polarity reversal" in four-wire resistance measurement mode. This is supposed to prevent EMF errors. However, it also causes "indication variability" in some cases when using electronic calibrators—changing the polarity for half a measurement cycle destabilizes the measured value. The 50% + and 50% - periods are like a square wave signal. For physical resistances, this doesn't matter, but for electronic ones, it's a problem. Connect an oscilloscope to the connected Datron and Valhalla and observe the waveform for a "square wave" at the signal input.

[Dr. Frank]

--

I tried it on my Datron 1082 and that REALLY doesn't like the 2724A for certain resistances, for instance setting 100K gives my ~120K and anything above that just gets garbage, 1K is pretty good but drifts 10 or 20 counts.

My Advantest R6581T likes it a lot more, it works right up to 1G, but the drift is quite obvious making the last 3-4 digits basically meaningless.

Hello Defpom,
all resistance calibrators which can output random resistance values, in contrast to only fixed ones, will synthesize the requested values.
That means, they consist of an active circuit to simulate the resistance.

This always makes trouble with DMMs which feature the Offset Compensation method, let it be with on/off or reversing current source.

I have a DATRON 1271 which switches on/off, and this as well does not work at all with my Datron 9000 calibrator.
My HP3458a also does not work with that synthesized method.

I always switch off the Offset Compensation in these cases, and get absolute stable readings.

Why this not the case with your instruments, I can't understand.

Frank

[branadic]

I haven't made extensive measurements by now, still an open task, but that is for another day.

-branadic-

[IanJ]

ChatGPT's opinion on the synthesized resistance method and why some DMM's have trouble:-

Why Some DMMs Have Trouble Reading the 2724A?

1. Synthetic Resistance Isn’t Passive
The 2724A is not a real resistor — it's an active circuit that emulates resistance using op-amps and feedback.

Most DMMs assume they’re measuring a passive component. They apply a known current and measure voltage (or vice versa), expecting no active behavior from the DUT.

But the 2724A can behave unexpectedly to the DMM’s test current because it actively controls the voltage/current across its terminals.

2. Response to DMM’s Internal Resistance Test Current
Some DMMs use a very small or pulsed test current to protect sensitive components. The 2724A may not respond properly to this kind of excitation.

If the DMM applies too small a stimulus, the 2724A's control loop may not activate or stabilize properly, leading to incorrect readings or out-of-range errors.

3. Bandwidth and Settling Time
The 2724A’s internal loop may have a settling time or bandwidth limitation.

Fast auto-ranging or fast-sampling DMMs (e.g., 6.5-digit bench meters) may not wait long enough for the 2724A to settle, or may interpret the dynamic behavior as noise or instability.

4. Polarity and Terminal Confusion
The 2724A uses Kelvin 4-wire connections, while many handheld DMMs use only 2-wire resistance measurement.

Using the wrong terminals or combining 2-wire with a 4-wire source can result in partial or incorrect resistance readings.

5. Output Impedance and Loop Interaction
Because the 2724A controls current and voltage simultaneously, it presents a non-standard impedance at its terminals.

The DMM and 2724A may "fight" each other — the DMM applies a current expecting a voltage response, but the 2724A is also controlling the voltage based on internal feedback.

DMMs That Usually Work Well:
Bench DMMs with proper 4-wire resistance mode (e.g., HP 34401A, Keithley 2000 series, Fluke 8846A).

Meters that allow for manual range selection and 4-wire sourcing tend to behave better.

Multimeters designed for characterizing active components are more tolerant of synthetic sources.

[TUMEMBER]

ChatGPT's opinion on the synthesized resistance method and why some DMM's have trouble:-



The DMM and 2724A may "fight" each other — the DMM applies a current expecting a voltage response, but the 2724A is also controlling the voltage based on internal feedback.

DMMs That Usually Work Well:
Bench DMMs with proper 4-wire resistance mode (e.g., HP 34401A, Keithley 2000 series, Fluke 8846A).

Meters that allow for manual range selection and 4-wire sourcing tend to behave better.

Multimeters designed for characterizing active components are more tolerant of synthetic sources.

Maybe this AI chat should stop confabulating and read the Keithley 2700 manual. I know it's a thick tome, almost 500 pages, and he could have messed up the "bits," so he didn't. The manual clearly describes a resistance measurement technique "with signal interleaving and polarity reversal." This was practically verified after connecting an INMEL 60 electronic resistance simulator. As a result, I had to redo a lot of documentation before the external audit—document falsification is commonplace in laboratories supervised by national ILAC administrators. It was necessary to enter single-measure resistance standards as a reference standard.

[Doctorandus_P]

Maybe this AI chat should stop confabulating and ...

But it does manage so put in some advertisements. That's probably the most important optimization of those AI things: Manipulating people without letting them realize they're being manipulated.

My own guess is that the meters below would also be pretty capable.
(Disclosure: No affiliation, I just selected some random benchtop DMM's >> EUR 1000)

• Hioki DM7276-02 (<- Oops, that's a volt meter only).
• Hioki RM3545A-1 (<- That is a Resistance Meter).
• R&S UDS600
• Siglent SDM4065A-SC
• Gw Instek GBM-3300
• Teledyne LeCroy T3DMM6-5(Hint: Compare with Siglent)
• Additel ADT286 - Thermometer would also fit the bill with it's USD11k price tag (and it does RTD's)

[TheDefpom]

Repair videos for my unit in case it helps anyone (there are 2 parts, Part 2 will be added later or linked in the video description of Part 1)

Part 1:



Part 2:

[Le_Bassiste]

ChatGPT's opinion on the synthesized resistance method and why some DMM's have trouble:-

Why Some DMMs Have Trouble Reading the 2724A?
...

i never found a single DMM having a problem with measuring a non-simulated resistor. the simulator should be "blamed" for not properly simulating the "real thing", rather than expecting a DMM to do something that it isn't intended to be used for: DMMs shall properly measure resistance, not simulations of resistance.  hence, shouldn't the question rather be "Why is the 2724A having trouble to present a resistance to some DMMs?"?
don't get me wrong, the 2724A is probably a fine instrument in its own scope of use (i.e., the Ohms-measurement strategy it was intended to be used with). thus, just willy-nilly throwing all kinds of different  strategies at it and then assuming the DMM has a problem is just obfuscating the view on the root cause.
just my 2c

[IanJ]

 I think you are misreading my use of the word “trouble” as if I was trying to say it’s a DMM issue……..That’s not the case at all.
Perhaps a phrasing, ambiguity, implied subject responsibility issue. I apologise.

Ian

[TERRA Operative]

My turn to acquire on of these, with the CPR option.

Does anyone have any information on the Option M24 Maintenance Kit? I see a parts list in the service manual, but if anyone has pictures or other information, that would be great.

[TheDefpom]

I managed to find a M24 switch box (model number is 2724), all it does is make it easier to do a calibration by allowing you to turn switches instead of manually moving cables around, is it useful, yes, is it REALLY needed, no.

I have shown it in a video below, I don't show its use, but you can see it sitting there:
 

I THINK I repaired the switch box in a live stream, I have been looking through my videos trying to find it but haven't been able to so far, I will let you know if I find it.

[TERRA Operative]

If you don't mind taking some photos inside and out, that would be greatly appreciated!

I have a rotary switch of the same type used in the box on the way and am organising the purchase of some low EMF binding posts too, so it'd be great to see the details so I can get my reproduction 'right'.