HP3458A SRAM replacement: copy or calibrate?

[View[+]Finder]

Why are nearly all of the posts about SRAM replacement concerned with “backing up the CAL data” so that it can be written into the new parts?

For the last few years I have been repairing and calibrating eBay gear to use in an educational program to help kids learn electronics. Our HP3458A is the lab standard and I became concerned what would happen if the SRAM batteries failed. All but a few of the posts described the process of extracting data and the expense of failing to do so and that didn’t jibe with what I had read in the HP Calibration manual. Based on my experience with repair and calibration, I find the 3458A to be the easiest of all to calibrate—yes, I have precision standards as needed—it just takes some time. Most of the hard work is built into the calibration firmware by HP.

It will soon be time to replace the SRAM’s in my 3458A as the batteries are way past their expected life, so I thought “why not just replace and run a full CAL and SCAL? Or are there ‘factory’ items in the SRAM?” I’ve reviewed the GPIB info on reading the cal data and it looks like all the adjustments are written during calibration. Am I missing something?

Thanks to advice from ‘an expert’ (xdev), I learned that I was on the right track about calibration data and that there are just few factory items that are not related to accuracy. (Perhaps as included in CALSTR?)

Not everybody has proper standards, cabling and SCAL equipment (TVCs,SG,reference meter) to perform full calibration though. Fortunately the key items for me—DCV, DCA and Ohms—require only a short piece of solid copper wire (12-14ga) for a thermal short (Zero Ohms), a precision 10VDC source and a 10kOhm resistance standard. AC calibration needs more equipment—I’ll figure that out when needed.

Right now the 3458A is working fine and in calibration, so my plan is to wait until the SRAM becomes unreliable—SRAM’s have been obtained from a reliable source—and then I’ll remove the old ones, install sockets and the new SRAM’s and start the calibration process.

Yes, the Keysight Cal Lab will be glad to help, but I’ll try DIY first.

Documentation to follow, maybe a video . . .

[nidlaX]

Lucky you that you have access to a few thousand dollars worth of voltage and resistance standards. Those standards by themselves may cost nearly a thousand dollars to calibrate.

Calibrating the 3458A is not a matter of technical difficulty, but rather a matter of financial expenditure.

[Dr. Frank]

Hi,

I also copied the content, due to financial reasons.
If you let KS do the job, they do not replace the 3 nv-SRAMs any more, but the whole PCB for a bundle of additional bucks, that at least doubles the price of the calibration.

The calibration of 10V and 10k may be comfortably done DIY, if you have stable standards at home, and then may transfer their known values to the fresh cal-RAM by SCAL.
Frequency calibration might also be done DIY, as frequency standards are also commonly and cheaply available.

The only problem left is the AC calibration, requiring a stable synthesizer, and a set of precise thermal converters, both being very uncommon in a private lab.

That was the main reason for me for to copy (thanks to quarks  ).

Frank

[View[+]Finder]

True, the AC side of the DMM is more complicated, however the calibration process has its origin in the time of the introduction of the 3458A. Back then the measurement of RMS was analog and derived from heating a wire and measuring voltage with a thermistor to get a DC equivalent ‘RMS’. According to what I’ve read, digitized measures as in a modern scope or DMM will yield an accurate RMS voltage at various frequencies. My bench DVM is a Keysight 34465 that measures true RMS by some sort of heating process and I have a signal generator that is supposed to be able to generate a sine wave at a specific RMS voltage.

The 3458 calibration calls for AC at specified RMS voltage and frequencies, so my plan is to measure the sig gen with the 34465 in parallel with the 3458 at the various stages of the cal process.

Yes, money helps—the 34465 cost a bit, but worth every penny—but most of my gear came from eBay and repaired by hand. Over time I bootstrapped from 4.5 digits to 5.5, 6.5 finally to the 3458 also from eBay.

I had to learn a lot along the way and I’m by no standards an expert, so if there is a flaw in the thinking, let me know.

Thanks to all

[Dr. Frank]

True, the AC side of the DMM is more complicated, however the calibration process has its origin in the time of the introduction of the 3458A. Back then the measurement of RMS was analog and derived from heating a wire and measuring voltage with a thermistor to get a DC equivalent ‘RMS’. According to what I’ve read, digitized measures as in a modern scope or DMM will yield an accurate RMS voltage at various frequencies. My bench DVM is a Keysight 34465 that measures true RMS by some sort of heating process and I have a signal generator that is supposed to be able to generate a sine wave at a specific RMS voltage.

The 3458 calibration calls for AC at specified RMS voltage and frequencies, so my plan is to measure the sig gen with the 34465 in parallel with the 3458 at the various stages of the cal process.

Yes, money helps—the 34465 cost a bit, but worth every penny—but most of my gear came from eBay and repaired by hand. Over time I bootstrapped from 4.5 digits to 5.5, 6.5 finally to the 3458 also from eBay.

I had to learn a lot along the way and I’m by no standards an expert, so if there is a flaw in the thinking, let me know.

Thanks to all

Well, yes, at first you should definitely read the calibration manual of the 3458A, page 27 onwards.
You need to calibrate the AC mode at 2MHz and at 8MHz, at 3V, 1V and 100mV.

That is not possible with a 34465A, only Thermal Converters will do the job properly.

Even if the 3458A is quite old, it's still one of the best and most demanding instruments.

Frank

[trle]

Hi !
I look at the forum and see that you have a knowledge for resolve my problem. Hope that you have a will to help me !?
I have a oscilloskope HP 54201A that doesn't work well, actually measured values are not good and they are constantly  changing.
Exp. when measuring frequencies, it is uneven and varies, even though the measured frequency is accurate and constant ( professional fm transmitter).
Hope that you have a time and will to help me to solve this!?
 Best regards
 Rade
Šibenik
Croatia

[View[+]Finder]

One should read before trying of course.

I read the calibration manual several times before posting and the specs for the 34465 as well. In addition, I read several sources about thermal converters. Aren’t ‘DC equivalent’ and RMS the same measurement??

What does the thermal converter do other than calibrate the synthesizer’s RMS voltage at a chosen frequency? Why would a 34465 not be able to measure RMS to the same effect?

With all due respect, I’m not sure I agree with Dr Frank’s conclusion about the necessity for a TC other than when a 3225A synthesizer is used.

With thanks to all.

[View[+]Finder]

Unless, of course, the 34465 only goes to 300kHz for RMS. Oops! Double-check time.
Ok, time to check the RMS specs for my o’scope. Thanks

[View[+]Finder]

Here's my setup for the calibration of the AC portion of the 3458A (SCAL):
1) Siglent 2042X function generator; frequency 100kHz, 2MHz, 8MHz; RMS voltage 100mV, 1V, 3V
2) Rubidium 10MHz frequency standard as ext input to the 2042X
3) R&S HMO1212 oscilloscope to monitor frequency and adjust the RMS voltage on the 2042X

Procedure:
The 2042x is connected to the HMO1212 with a BNC cable and also to the 3458A with a BNC 'T-connector'.
The 2042X is a substitute for the HP3325A function generator and the 'scope takes the place of the thermal converter. In this case the measurement of RMS voltage (DC equivalent) can be obtained from the 'scope directly and the 2042X adjusted to match the required input voltage. This will eliminate the steps associated with measuring the voltage from the thermal converter in order to correct the RMS errors in the 3325A output.

Rationale:
The accuracy and precision of today's budget-priced electronic equipment sufficiently exceeds that of equipment available when the HP3458A was introduced in 1989 to justify a modification of the claibration procedures. While it is certain that many 3458A's in use by readers of this forum are relied upon to calibrate other equipment on their bench, one would doubt that any operate at metrology levels of environmental control. In short, the 3458A I have is an 'old workhorse' still able to perform all of the tasks needed at a level of precision and repeatability sufficient to calibrate other equipment in my 'lab'. Many 3458A's on eBay have been 'rode hard and put up wet' (to continue the 'horse' metaphor) and the stories of their rehabilitation and recovery are inspiring to say the least.

In the spirit of 'reuse before recycle' it is better to have a working 3458A on one's bench and given the limitations of the NVSRAM battery life, saving the cost of a Keysight calibration is an added bonus.

If anyone wants links to the specification pages for any of the equipment mentioned, please ask. I have read through the specs, but not done an exact comparison. My guess is that the procedure above might improve on the 'thermal converter' method by an order of magnitude, particularly considering the state of gear on eBay these days.

Comments appreciated . . .

[lukier]

Here's my setup for the calibration of the AC portion of the 3458A (SCAL):
1) Siglent 2042X function generator; frequency 100kHz, 2MHz, 8MHz; RMS voltage 100mV, 1V, 3V
2) Rubidium 10MHz frequency standard as ext input to the 2042X
3) R&S HMO1212 oscilloscope to monitor frequency and adjust the RMS voltage on the 2042X

Procedure:
The 2042x is connected to the HMO1212 with a BNC cable and also to the 3458A with a BNC 'T-connector'.
The 2042X is a substitute for the HP3325A function generator and the 'scope takes the place of the thermal converter. In this case the measurement of RMS voltage (DC equivalent) can be obtained from the 'scope directly and the 2042X adjusted to match the required input voltage. This will eliminate the steps associated with measuring the voltage from the thermal converter in order to correct the RMS errors in the 3325A output.

You realise that a scope has something like 3% accuracy on the vertical? 8 bit ADC. 3458A can get down to 0.01% according to the datasheet. If you can afford 3458A then spending $100 or so on a Fluke 540A shouldn't be a problem. Fluke 540A might just about cut the muster with careful operation when doing transfers. Here https://www.eevblog.com/forum/testgear/fluke-540b-thermal-transfer-standard/ users report that 0.01% is doable.

Edit: For frequencies above 1 MHz you will need additional plugins for Fluke 540B.

The accuracy and precision of today's budget-priced electronic equipment sufficiently exceeds that of equipment available when the HP3458A was introduced in 1989 to justify a modification of the claibration procedures.

This is precision analog, not CPUs or whatnot. 3458A has the best ADC on the planet, unbeatable since 1989. Even Keysight apparently struggles to make a modernized replacement. That speaks volumes.

[TiN]

The accuracy and precision of today's budget-priced electronic equipment sufficiently exceeds that of equipment available when the HP3458A was introduced in 1989

You are in for a surprise. One of the reasons why 3458A is still currently manufactured and supported product since it's release in 29 years ago is simply due to precision and accuracy of it is not beaten by any modern equipment in that price range. There are handful of meters than are on par, with much higher pricetag (8508A on resistance, 5790A/B on AC voltage/current, etc).

My guess is that the procedure above might improve on the 'thermal converter' method by an order of magnitude, particularly considering the state of gear on eBay these days.

Thermal converter method is what is still currently used by many metrology labs and NMIs, together with AC voltage JJA systems (for those who can afford them). Fluke 792A is current commercial standard for AC voltage measurement for highest accuracy, still based on thermal conversion method. And 57xx series calibrators as well. 

I'd suggest to research the literature more. Also HP Journal covering 3458A have plenty of juicy details.

[texaspyro]

Sorry,  I can't stop giggling at the thought of using a scope to calibrate a 3458A...     Maybe use an LM327 and some carbon comp resistors for the 10V / 10K cal standards? 

Also,  I doubt that we will ever see a DMM with significantly better overall performance than the 3458A and today's other highest end meters.   8.5 digits is approaching the physical limits of what silicon can achieve at room temperature.

[View[+]Finder]

Thans for the comments--glad to hear there is still humor in electronics, even if at someone else's expense.

However, several posts missed the point: my choice was between copying the contents of the NVSRAM (then loading on to the new chip) and just doing a full calibration on the 3458A once the new chips are installed. That's it. Yes, I could send it to Keysight for calibration, but how does that help someone who neither has the skill and gear to do an SRAM clone nor the funds to send to a cal lab?

Some readers (and perhaps a few posters) might have their 3458A in a controlled environment, fully caibrated to yield measurment to the full spec of a 'new' meter and up and running 24/7. That might be my goal, but falling short of perfection is not failure. Considering the comment "precision and accuracy," I would add "repeatability" as the feature that sets the 3458A apart. That is something I learned from a very smart guy on 3458A's who also was the source of my resistance calibration standards and a wealth of good advice.

For those who don't know, The major selling point of the 3458A back in 1989 was that it required only two standards to calibrate: 10VDC and 10kOhm. That was sufficient for all DC measures. The SCAL was needed only every 2 years and generally was done in a cal lab. The AC section of the 3458A--unlike my 34465A DMM--can measure RMS voltage over a range from 1Hz to 10MHz. As far as I have found in the manual, this capability is not a front panel setting, only GPIB, any help here? In any event, this measure is not on my 'must-haves' list. SCAL is needed to setup for any AC measure, hence my need for an alternative approach. BTW, the AC section is six and a half digits and who know what one should expect from an eBay thermal converter. Good luck measuring RMS with a 'heater and thermocouple' anywhere outside a temperature controlled, draft-free space.

Finally, the RMS voltage is being generated by the Siglent 2042X linked to a Rh 10MHz standard--far superior to the 3325A synthesizer--and the 10bit ADC o'scope is in the circuit to monitor and fine tune the RMS.

Here's my conclusion: disassembling the 3458A, removing the A5 board, unsoldering a few chips, adding sockets and new chips and putting it all back together will require a full calibration even if the SRAM cloning is successful. Add to this the uncertainty of 'component fondling' during assembly, thermal shock from the soldering and assistance from the family pet and a full calibration is the only option.

[TheSteve]

Removing the A5 board should not change calibration at all.
A device to read/write the nvram IC's is cheaper then the new nvrams. If you can afford a 3458A hopefully you can cough up the money for a programmer(less then 50 bucks).

[View[+]Finder]

I was wrong: the 3458A display shows AC voltage as RMS. If generator output is 3V RMS, that is what is on the 3458A; 3VAC.

To Steve's comment: it is not about the cost of the programmer (I probably have one, it's been a while); it's about "Why bother?" I don't regularly use chip reading and writing software and everytime I do it turns into a day's work--the software needs updating, I can't find the programmer, WTF do I do next . . .. That's on me.
 

I have a 10V standard and a 10kOhm standard as well, so I can get the 3458A calibrated for DC no problem. I don't see the need for the cloning operation to avoid the SCAL. Anyone who is replacing NVSRAM is probably working on a 3458A that is way overdue for an SCAL and using the data form the clone is not a substitute for calibration.

Here's a thought: if the 3458A is still working (as mine is) use it to calibrate the signal generator. Hmmm . . .

[dr.diesel]

I have a 10V standard and a 10kOhm standard as well, so I can get the 3458A calibrated for DC no problem.

Are they the appropriate standards for a 3458A cal?

There is only a handful of people with these laying around at home in calibration.

[View[+]Finder]

Good question. I'd say "not as good as it deserves" but good enough until I have to replace the NVSRAM. Then it gets a trip to Loveland for a real cal. I was very lucky and got a very clean and functional 3458A on eBay for under $3k, no errors other than needing a calibration.

The 10kOhm is Leeds and Northrup calibrated against a 3458A Op002 with a NIST cal cert. The 'actual' value is 10.00077kOhms over a 24hrs in a water bath. Temp was correlated with resistance--I have the papers on it from the calibrator.

My 10V is http://www.ianjohnston.com voltage standard first model. It was calibrated against a 3458A (I don't recall whether Op002) with 10.00000 at 23.1C. I have his new model on order.

The bench 34465A is in cal, but will need a calibration soon.

So to answer your question, I'm not anywhere near metrology standards, but with the tools available, I can work on ebay gear and repair and calibrate up to the standard when most of it was new 25 or so years ago.

Thanks

[Dr. Frank]

If you send your instrument to Loveland anyhow, this discussion about copying or not is quite useless.

But don't miss the fact, that KS will definitely replace the A5 board with a modern one, if you won't change the nvSRAMs before.
The old PCBs have the advantage, that you can add 128kB of reading memory for free, as KS will charge you additional ~ 500 bucks for that option.

If you measure your references by your 3458A, just before you change the SRAMs, this should be good enough for a close reconstruction of the two calibration constants.. you can just write them down before the re-calibration, and compare them afterwards, how big the difference might be.

In the end, it's up to you, the advices given were the least costly options, I think.
Frank

[wraper]

To Steve's comment: it is not about the cost of the programmer (I probably have one, it's been a while); it's about "Why bother?" I don't regularly use chip reading and writing software and everytime I do it turns into a day's work--the software needs updating, I can't find the programmer, WTF do I do next . . .. That's on me.

Do you realize you spent 10x more time on this tread than it would take to copy SRAM contents?

[macboy]

True, the AC side of the DMM is more complicated, however the calibration process has its origin in the time of the introduction of the 3458A. Back then the measurement of RMS was analog and derived from heating a wire and measuring voltage with a thermistor to get a DC equivalent ‘RMS’. According to what I’ve read, digitized measures as in a modern scope or DMM will yield an accurate RMS voltage at various frequencies. My bench DVM is a Keysight 34465 that measures true RMS by some sort of heating process and I have a signal generator that is supposed to be able to generate a sine wave at a specific RMS voltage.

The 3458 calibration calls for AC at specified RMS voltage and frequencies, so my plan is to measure the sig gen with the 34465 in parallel with the 3458 at the various stages of the cal process.

Yes, money helps—the 34465 cost a bit, but worth every penny—but most of my gear came from eBay and repaired by hand. Over time I bootstrapped from 4.5 digits to 5.5, 6.5 finally to the 3458 also from eBay.

I had to learn a lot along the way and I’m by no standards an expert, so if there is a flaw in the thinking, let me know.

Thanks to all

You really need to read the manuals and datasheets of your gear. Otherwise you have no idea what they are capable of.

Neither the 3458A nor the newer 34465A uses a thermal RMS converter. Actually, thermal RMS converters have always been quite uncommon, especially ones that are done right.  The 3458A has three choices of RMS measurement technique. One is using analog computation, which is incidentally by far the most common method of RMS measurement. This involves passing the AC signal through a log amp, taking an average, and passing through an anti-log amp. This is implicitly calculating the RMS value (the result is the same, but it isn't doing actual square, mean, and square-root calculations).  The 3458A's  second and generally preferred method is the synchronous sub sampling method which works well for repetitive signals up to 10 MHz. The third is the random sampling method, good for random signals.  The 34465A uses the sub-sampling methods that were designed for the 3458A. Judging solely by the datasheet, it doesn't seem to have any analog-calculating type of RMS converter, and definitely has no thermal converter.

The analog computation method is the most popular largely because Analog Devices has a set of easy to use and accurate ICs to perform this computation: AD636, AD637, etc. You can find these in many meters, both handheld and bench. If it does True RMS with more than about 40 or 50 kHz of bandwidth, the chances are very good that it uses one.

[View[+]Finder]

1) My purpose in starting this thread is to assist someone else who (for whatever reason) might want to calibrate their 3458A with a minimum amount of gear or replace the NVSRAM without having to clone the old one. The time I spent might help someone else spend their time doing something else.

2) Why would anyone start a thread like this without having first read the manuals? Really? More to the point, before even looking on eBay for a 3458A, I read the manuals and the 1989 Keynote article.

3) The goal is to have a working 3458A after the day arrives when the NVSRAM's are replaced. The bulk of the posts and YouTube vids left me with the impression that there was "factory-magic" in the CAL RAM that needed to be copied or the 3458A would become a brick. Thanks to this forum, I learned that was not true: a full calibration is all that is needed to put the DMM back online. No, I do not want to spend $1,000 for a new board at Keysight, hence the need to do the SRAM replacement first. I also read enough of the 3325A manual to conclude that, for the purpose at hand, it is just a sine wave signal generator. Then I read up on thermal converters--a thermocouple combined with a resistor--about as old-school analog as you can get.

Now, after trying the sig-gen/oscilloscope combo to do the SCAL, I learned a few things: first, it can be done; second, an analog thermal converter might have an advantage over the digital setup I used if the objective is to bring the 3458A fully up to spec; and finally, spending time helping others is never wasted.

[texaspyro]

Get a TL866 device programmer.  They are cheap, easy to use,  and very versatile and can do the NVRAM module.   My recommendation is to first dump the cal ram using the GPIB connector.  KE5FX's gpibkit package includes my cal ram dumper program.  It outputs a listing of the cal ram contents as a text file that lists all the current settings.  It also outputs a binary ROM image.  The GPIB dump is insurance against a glitch in desoldering the NVRAM corrupting the data.

Once you pull the NVRAM, use the TL866 to read it and save the contents to another file as an additional backup.  Program the new NVRAM chip with the old data.  Even  decades old SCAL values will be much more accurate than anything you can do at home.  The same applies for the 10V/10K artifact calibration unless you have a calibrated 732A/732B and SR104 standard resistor. You REALLY don't want to mess up a 3458A by using inferior calibration standards.

[TiN]

I would be far more concerned with ADC stability (on which topic I've posted many times), than with home calibrations, leave alone SCAL. As if your meter unstable, you are in for 1300$ ADC replacement the moment you send it to KS cal.

[wraper]

1) My purpose in starting this thread is to assist someone else who (for whatever reason) might want to calibrate their 3458A with a minimum amount of gear or replace the NVSRAM without having to clone the old one. The time I spent might help someone else spend their time doing something else.

Barely anyone has all equipment needed to calibrate 3458A. And if someone has equipment required, he/she doesn't need this tread in the first place. From what you wrote, you don't have equipment needed to properly calibrate 3458A.

Procedure:
The 2042x is connected to the HMO1212 with a BNC cable and also to the 3458A with a BNC 'T-connector'.
The 2042X is a substitute for the HP3325A function generator and the 'scope takes the place of the thermal converter. In this case the measurement of RMS voltage (DC equivalent) can be obtained from the 'scope directly and the 2042X adjusted to match the required input voltage. This will eliminate the steps associated with measuring the voltage from the thermal converter in order to correct the RMS errors in the 3325A output.

This procedure is not good even for cheap 6000 count multimeter.

that help someone who neither has the skill and gear to do an SRAM clone nor the funds to send to a cal lab?

It's relatively easy and equipment required costs <$50. Poor man's calibration is only valid if SRAM already lost it's data.

[wraper]

I would be far more concerned with ADC stability (on which topic I've posted many times), than with home calibrations, leave alone SCAL. As if your meter unstable, you are in for 1300$ ADC replacement the moment you send it to KS cal.

Yep, I'd be concerned what Keysight would say about your homebrew calibration. They may decide that it drifted that much that meter should be faulty.