Affordable Useful AC & DC references

[Majorassburn]

It was suggested by other forum participants that I start a new thread to discuss specs and features of AC & DC voltage references that I make and sell on eBay (as SQWARREL. Sample Link https://www.ebay.com/itm/285789548856) where I kept things brief and non-specific. So, here is my attempt to do so. I appreciate your input and suggestions as this thread (and my device improvements) develop.

My intent is to make and sell very low-priced but USEFUL references for those of us who are not yet ready to resort to costly lab calibration. We may want to be able to perform basic, "ballpark" function checks of our DMM's to see if they're even working correctly or reasonably holding calibration.  And, we can't justify spending $150 or more for higher quality devices like the DMMCheck+ and others.

AC Voltage Reference:  Here's a device that generates a relatively pure, NON-FLUCTUATING, adjustable sine wave from 0 to 6VACrms @ 100Hz (fixed frequency) for those who do not have a quality signal generator or alternative. (100Hz was chosen as a DMM-friendly "sweet spot" for avoiding EMI-influence in the working environment.)

Remember, you can't do credible AC comparisons or checks by using isolation/step-down transformers, variacs or by sticking your probes into a mains socket because the line voltage is always jumping around and the sine waves don't look anything like sine waves because of all the noise and distortion.

The AC references can be pre-calibrated (buyer's choice) to between 100mV and 6VACrms @ 100Hz +/- 0.5% to cover near-full-scale readings of most popular DMM's, requiring only a single 12VDC regulated power supply or battery because they have an on-PCB ICL7662 inverter for the -12V rail.

There can be several variations (and prices) for such a device including lower sine wave amplitudes which allow using single supply op-amps and 9V batteries, etc., etc., but I would rather focus on minimum performance SPECIFICATIONS for device USEFULNESS at this time.

DC Voltage Reference: Rather than the untrustworthy cheap references, many using crappy components and USED AD586 parts and phony calibration stickers, I have developed a straightforward 10VDC reference based upon the LT1236 chip.

This DC reference currently uses a 14VAC 0.5A wall-wart transformer or other suitable 14-16VAC power supply to the input jack leading to an on-PCB 15VDC regulated power supply that then feeds the LT1236 circuitry. The power input can also be 18-24VDC through the same input jack and that also offers autopolarity protection due to the full wave bridge rectifier at the input on-PCB.

The LT1236 is trimmed to exactly 10.0000VDC. How does this happen? I use our brand new SDM3055 5.5-digit DMM (new production, direct from Siglent) which is "comparison calibrated" to a new DMM6500 that I have access to. Then, the SDM3055 is the basis for "comparison calibrating" my Fluke 87V's.

The DC reference trimming is done with the SDM3055 by powering on both for 24-48 hours before final trimming to 10.0000 which is as good as it gets until we buy a DMM6500 or similar in the near future.

However, after trimming one of the 10V references 2 months ago and keeping it powered on, and after re-checking the SDM3055 vs DMM6500 readings weekly, the 10VDC reference STILL reads exactly 10.0000, spot on, if that's worth anything.

Now, I realize that this whole discussion doesn't meet any volt-nut standards but, remember why we're here in the first place, as I set forth in my introduction.

I simply would like to improve the SPECIFICATIONS and USEFULNESS of these devices while keeping them low-cost to give the average DMM owner an alternative to spending hundreds of dollars and more on devices that may be of higher quality and accuracy but are of no more usefulness.

Thanks for any insight and help that comes to mind.  A few pix attached for general reference...
 
Modify message

[schmitt trigger]

Does it have distortion specs?

[alm]

The LT1236 is trimmed to exactly 10.0000VDC. How does this happen? I use our brand new SDM3055 5.5-digit DMM (new production, direct from Siglent) which is "comparison calibrated" to a new DMM6500 that I have access to. Then, the SDM3055 is the basis for "comparison calibrating" my Fluke 87V's.

The DC reference trimming is done with the SDM3055 by powering on both for 24-48 hours before final trimming to 10.0000 which is as good as it gets until we buy a DMM6500 or similar in the near future.

However, after trimming one of the 10V references 2 months ago and keeping it powered on, and after re-checking the SDM3055 vs DMM6500 readings weekly, the 10VDC reference STILL reads exactly 10.0000, spot on, if that's worth anything.

I would try to translate this to an adjustment uncertainty specification and long term drift specification (though measuring on a 2 month time scale may require very accurate and stable equipment). Compare this to measuring the drift of continents on the time scale of a year vs a century. You need way more accurate equipment to measure low velocities by comparing the position at two points in time. You are not trying to measure drift, you are trying to set an upper bound like the drift over 2 months is less than 100 uV (which might translate in an upper bound of 600 uV/year). Compare this to a datasheet values.

For adjustment uncertainty. To me the "brand new SDM3055" says "suspect and potentially high drift". Generally voltage references start to settle in after some months of operation. And from what I read, Siglent may not be doing a very good job burning in their references before putting them in instruments. But with the weekly comparisons to the DMM6500 you should be able to get some measurement of stability over time. The uncertainty might look something like this:

• DMM6500 absolute uncertainty from datasheet (1 year or 90 days since calibration?)
• Uncertainty of transferring from DMM6500 to SDM3055 (noise, temperature fluctuations / coefficients for both meters, repeatability)
• 1 week stability of SDM3055 (you can estimate this from repeated comparisons with the DMM6500 taking into account the uncertainties of the DMM6500 and the comparison)
• Uncertainty of measuring the voltage reference with the SDM3055 (noise, temperature fluctuations / coefficients for DMM and reference, repeatability)

For the ones that are not normal distributions (like specs from a datasheet), you multiply them by the factor from The Guide to Uncertainties in Measurement (see page 13 for how to rectangular distributions, e.g min/max specs). Then you add them all up (linear is the most conservative), and multiply by a coverage factor k=2. This should give you a best estimate for the uncertainty of the reference just after adjustment. Probably not all components will contribute equally. Feel free to simplify and use conservative estimates for those.

Then using the same procedure at different points in time, you can calculate an upper bound to the drift based on the uncertainty of each measurements. I would show a table with uncertainty calculations so anyone who doubts the figures can verify.

Ideally you would build a small thermal chamber that would allow to to characterize the temperature coefficient of the circuits. Temperature and time are probably the major variables affecting the output of the standard. Humidity would be next, but is difficult to control and also difficult to measure to to the large time constants involved.

[bdunham7]

The LT1236 is trimmed to exactly 10.0000VDC. How does this happen? I use our brand new SDM3055 5.5-digit DMM (new production, direct from Siglent) which is "comparison calibrated" to a new DMM6500 that I have access to. Then, the SDM3055 is the basis for "comparison calibrating" my Fluke 87V's.

I typically think of the errors rather than the accuracy, so I avoid terms like 'spot-on'.  To me 'spot-on' just means you don't have enough resolution to see the errors.  It's like someone asking me to machine something with a dimension of "one inch".  So I ask "to what tolerance?".  They reply "I want it exactly one inch".   I spend a week setting up my most precise measuring tools and another week lapping the surfaces so that they are 1 inch apart to a precision of 0.00001", the best I can do.  Customer comes in to pick it up, pulls out a measuring tape and says "yep, looks like exactly one inch". 

That's not a true story, of course, but it is sort of based on one.  The question I have first is about your 'comparison calibrating".  I'll presume you are not adjusting anything but rather just using a stable voltage source to take measurements with the two meters in parallel and comparing the results.  Correct me if that is wrong.  If that is the way you are doing it, are you saying that your SDM3055 always matches the DMM6500 exactly to the last digit?  And then your Fluke 87V matches the SDM3055 to the last digit, each and every time?  If so, that is truly exceptional.  If not, my recommendation would be to start by logging and stating the errors in each of those steps. 

[Majorassburn]

The LT1236 is trimmed to exactly 10.0000VDC. How does this happen? I use our brand new SDM3055 5.5-digit DMM (new production, direct from Siglent) which is "comparison calibrated" to a new DMM6500 that I have access to. Then, the SDM3055 is the basis for "comparison calibrating" my Fluke 87V's.

I typically think of the errors rather than the accuracy, so I avoid terms like 'spot-on'.  To me 'spot-on' just means you don't have enough resolution to see the errors. 

-snip-

If that is the way you are doing it, are you saying that your SDM3055 always matches the DMM6500 exactly to the last digit?  And then your Fluke 87V matches the SDM3055 to the last digit, each and every time?  If so, that is truly exceptional.

Both good questions.  First part, the display of "10.0000" on the SDM3055 is the limit of its resolution. So, all further errors (and there must be some) are masked.  However, four zeroes is probably as good as it's going to get with a $25 reference with today's technology.

More importantly, four zeroes is certainly adequate enough for the purposes outlined in my intro although certainly not anywhere near good enough for verifying the calibration of a 34465A or DMM6500.

The second point, is also a good question.  So, far the "comparison calibration" method I referred to is simply attaching both the DMM6500 and the SDM3055 to the same 10V reference (after warmup) and reading the displays. No intervention or actual calibration. The SDM3055 agrees with the DMM6500 to four zeroes. I have NOT recorded any further errors or deviation of the 10V reference revealed by the DMM6500's extra resolution during these comparisons.

As for the Fluke 87V in the photo, it agrees with the SDM3055 only SOME of the time because it's running in Hi-Res mode where the offset fluctuates +/- 1LSD. It can read anywhere from 9.999 to 10.001 with 10.000 most of the time. That's not good enough but it is a curiosity.

I'll buy either a 34465A or DMM6500 if these reference devices prove salable. If not, I'll stick with the SDM3055 for awhile as long as I have access to the new DMM6500.

So, considering that these are affordable, low-cost references, I think accuracy to four zeroes is adequate for the purposes which they are designed to fulfil. Maybe I'm missing something but offering free periodic re-cals of the 10V & AC references once further detailed specs are developed along the lines that poster ALM suggests above, might keep buyers happy and provide the value that is intended for the prices paid.

What I'm getting at is developing a spec for example, the 10V reference, that says something like, trimmed to four zeroes, plus or minus 2 LSD's over a 6 month period? Maybe?

[bdunham7]

What I'm getting at is developing a spec for example, the 10V reference, that says something like, trimmed to four zeroes, plus or minus 2 LSD's over a 6 month period? Maybe?

2 LSDs means 20ppm in this case.  If your device can meet that spec over 6 months under specified conditions, that would make it pretty good.  If they are that good, then a DMM6500 would be a good investment because you would be able to clearly see the errors, instead of just looking at that last digit and wondering if it is going to flicker or not.  My main suggestion, agreeing with alm, would be that you work on characterizing a set of them--at least 10--over a period of at least six months and over at least a 10C temperature swing.  You should also at least record the RH.  My second suggestion is that you change the IC socket and use this:

https://www.mouser.com/ProductDetail/Mill-Max/110-13-308-41-001000?qs=WZeyYeqMOWeYjIS4tXLt7Q%3D%3D

I know your BOM is tight, but I think you may eventually have issues with the style of socket you are using now.

[Majorassburn]

What I'm getting at is developing a spec for example, the 10V reference, that says something like, trimmed to four zeroes, plus or minus 2 LSD's over a 6 month period? Maybe?

My second suggestion is that you change the IC socket and use this:
https://www.mouser.com/ProductDetail/Mill-Max/110-13-308-41-001000?qs=WZeyYeqMOWeYjIS4tXLt7Q%3D%3D
I know your BOM is tight, but I think you may eventually have issues with the style of socket you are using now.

HA! I already have those great sockets. Gold flashed, low insertion force, beryllium copper spring retention, low profile and great K effects,  However, I opted to use the cheap tin plated plastic sockets because of the similarity of the plating on the leads of the IC and the sockets themselves possibly reducing K effects.

Somewhere, I read about the Nobility of Metals stuff. I like the construction quality of the round gold sockets much better but what do you think about the dissimilarity of the IC pin plating to socket junction effects?

[shapirus]

If you're going to make at least 5-10 of these references, why not design an actual PCB and order some at a fab house?

It will likely not be more expensive than the breadboards I see in the pictures, and it will definitely require way less work to assemble and will be less error-prone, not to mention professional appearance (especially if you pay a little extra for black solder mask lol), which may improve sales.

If you design a PCB, you can add holes/slots for inserting (with or without soldering) a metal shield that, in combination with the ground plane on the other side of the board, will cover the reference chip and all the signal traces to reduce noise in the output signal.

[bdunham7]

Somewhere, I read about the Nobility of Metals stuff. I like the construction quality of the round gold sockets much better but what do you think about the dissimilarity of the IC pin plating to socket junction effects?

My guess is that thermal effects will be minimal, mostly because the temperatures will pretty quickly even out and also because when you add them up they'll mostly cancel.  But this is something you should determine experimentally.

[schmitt trigger]

Asking again; does it have distortion specs?

A “relative pure sinewave” is not a spec.

[Majorassburn]

Asking again; does it have distortion specs?

A “relative pure sinewave” is not a spec.

Sorry, I missed your question earlier....
1) Where did you see "relative pure sine wave" ?  I don't recall using that term in the AC Reference description.
2) Distortion is less than 10%. Depending on the designs I've used (phase shift, dual integrator, etc. and the type of components and amplitude designed for, it can be less than 1%. The newest designs have about 1-2%.

When designing these references for DMM checking, the frequency, stability and amplitude are much more important than a few percent distortion. These are fixed, low frequency sine wave generators and they aren't aimed for audio use. I have a line of  low distortion 1KHz-20Khz sine gens for that.

How about you guys with much more sophisticated measuring setups buy a few of these AC & DC references and actually put them to the test? I would make a special price of 50% of my cost to you and after you evaluate and report, I'll give you a free sample from the first production run. Partners anyone? 

[J-R]

Some thoughts:
- Have a look at the sites of Doug/Russ/Ian for ideas of the information that might be worth adding to your listings: https://voltagestandard.com | https://dmmcheckplus.com | https://www.ianjohnston.com/index.php/onlineshop/handheld-precision-digital-voltage-source-2-mini-detail
- A real PCB with a revision number seems critical.
- Ship some free units out for reviewers to check out, but talk to them first.  (mjlorton just came back after a year off and may be looking for stuff, dunno)
- Protect the trim pots or lock them in place with some adhesive or etc.
- Enclosure option?

[Kean]

Sorry, I missed your question earlier....
1) Where did you see "relative pure sine wave" ?  I don't recall using that term in the AC Reference description.

Your top post, third paragraph.

AC Voltage Reference:  Here's a device that generates a relatively pure, NON-FLUCTUATING, adjustable sine wave from 0 to 6VACrms @ 100Hz (fixed frequency) for those who do not have a quality signal generator or alternative.

[EC8010]

When designing these references for DMM checking, the frequency, stability and amplitude are much more important than a few percent distortion.

I'm afraid you're wrong there; you do need low distortion. If you look at mains voltage, it has flat tops (caused by capacitor input supplies) and a true-RMS meter gives a very different answer to a mean reading meter calibrated RMS of sine wave. Total harmonic distortion on mains with those flat tops is typically 3%.

[IanJ]

You need a proper Pcb if you are manufacturing to sell IMHO. Here's some points:

- Looks more professional, better branding.
- Better control/elimination of noise induced by crosstalk etc.
- Better control/tracking over batches & Cal certs using Serial No's / Pcb versions.
- Much faster to manufacturer.
- Your sales will go up.

Also,
- The test instruments you use to calibrate need to be an order of magnitude or better resolution/stability of the device you are trying to Cal, and have their own traceable Cal standard test certs.

Hope that helps.

Ian.

[Majorassburn]

Wow!  I want to say Thank You to all who have made suggestions and offered advice in this thread. This is a great forum and a great learning tool. I love it because of all you guys who take time to weigh in on topics like this.

As you know, I'm in the early stages of developing a few very low-cost products to see if they are even salable and useful to low-end DMM & Scope buyers (hobbyists, DIY, Homeowner, Students, Field Repair, etc.).  I saw that this explosively growing market had a gap of available, affordable (very low cost) but still useful "ballpark" references to do some function testing of their low cost devices.

These typical low-cost devices do not warrant professional calibration and, in most cases, have no calibration adjustments, anyway. So, for those low-end buyers who don't already have high-quality DMM's, etc. to compare to, an inexpensive reference device might be the only way to gain a little confidence in what their Low-end device is telling them. So, the gap.

We would all probably agree that a DMM Check+ and the like would be overkill for most of that low-end market so I would make no attempt to duplicate or impinge upon the typical DMM Check+ buyer/user demographic. I'm strictly at the very low end with specific focus on the most common basic functional tests.

Why all this preamble?  Because I appreciate all your inputs.  So, please don't think I'm ignoring your recommendations if I don't address them individually or implement them right away. You guys are helping me LEARN and grow. My dilemma is keeping the sale price LOW while providing VALUE to the buyer. It's a tough balance. Your input helps me move towards achieving those goals, while on a strict budget.

Thanks again.
Major

[Majorassburn]

Sorry, I missed your question earlier....
1) Where did you see "relative pure sine wave" ?  I don't recall using that term in the AC Reference description.

Your top post, third paragraph.

AC Voltage Reference:  Here's a device that generates a relatively pure, NON-FLUCTUATING, adjustable sine wave from 0 to 6VACrms @ 100Hz (fixed frequency) for those who do not have a quality signal generator or alternative.

Thanks for proving that my age is creeping up on me! 
My reply should have said that I never used that type of term in a device description or ad.
Of course, in this thread, my statement that my AC reference generates a "relatively pure sine wave" was meant for forum discussion. What does "relatively pure" mean to me? Almost, but not quite perfect, in shape. 

[shapirus]

What does "relatively pure" mean to me? Almost, but not quite perfect, in shape. 

You should be able to characterize that using a sound card and some software spectrum analyzer.

[Majorassburn]

Asking again; does it have distortion specs?

A “relative pure sinewave” is not a spec.

Sorry, I missed your question earlier....
1) Where did you see "relative pure sine wave" ?  I don't recall using that term in the AC Reference description.
2) Distortion is less than 10%. Depending on the designs I've used (phase shift, dual integrator, etc. and the type of components and amplitude designed for, it can be less than 1%. The newest designs have about 1-2%.

When designing these references for DMM checking, the frequency, stability and amplitude are much more important than a few percent distortion. These are fixed, low frequency sine wave generators and they aren't aimed for audio use. I have a line of  low distortion 1KHz-20Khz sine gens for that.

How about you guys with much more sophisticated measuring setups buy a few of these AC & DC references and actually put them to the test? I would make a special price of 50% of my cost to you and after you evaluate and report, I'll give you a free sample from the first production run. Partners anyone? 

I want to re-address, clarify and correct my reply:
1) All AC references that I have advertised and sold on eBay in the past year have output pure sine waves  at fixed frequencies at controlled amplitudes, as each sold device indicated.
2) Distortion specifications for all such outputs are: Less Than 1%. (My previous reply had typos and brain-farts in it  ) Below photo.
3) The most important benefit (in my mind) of owning one of my AC references is that it outputs a STEADY, NON-FLUCTUATING, NON-NOISY, NON-DISTORTED, SAFE sine wave that you can actually measure with a high degree of accuracy, AS OPPOSED TO the more common method of trying to guess at any derivative of measuring dirty, unstable mains voltages (like direct probes or step-down transformers, variacs, etc.). That's if you don't have a nice signal generator.

[Majorassburn]

When designing these references for DMM checking, the frequency, stability and amplitude are much more important than a few percent distortion.

I'm afraid you're wrong there; you do need low distortion. If you look at mains voltage, it has flat tops (caused by capacitor input supplies) and a true-RMS meter gives a very different answer to a mean reading meter calibrated RMS of sine wave. Total harmonic distortion on mains with those flat tops is typically 3%.

I agree with you but disagree with you??? 
Yes, the mains are very dirty and distorted. About the only thing you can depend on is that the frequency is usually very close while voltage fluctuates and so-called sine waves are flat-topped and almost unrecognizable due to the influences that you pointed out.

BUT: When using one of my AC references at 6VACrms at 100Hz, with less than 1% distortion of the sine wave output, the frequency, stability and amplitude are much more important than sub-1% distortion concerns. And, the typical audio and RFI distortion concerns aren't really relevant when checking an under-$100 DMM with a 100Hz, 6V sine wave as long as the sine wave is a sine wave.

You won't even be thinking about distortion if the amplitude and frequency of the sine waves are jumping all over the place, right?

Please check the attached photo and let me know what you think of that compared to sticking your probes into a mains socket.

[bdunham7]

2) Distortion specifications for all such outputs are: Less Than 1%.

How are you measuring that?  You probably won't see 1% harmonic distortion on a scope trace.  1% might matter if you were intercomparing TRMS and average-responding meters.  For TRMS to TRMS comparisons, it might not matter very much especially at the handheld level of precision.

[Majorassburn]

Just an aside....

With a source of a precise, 50% duty cycle square wave, 12.000VDC @ 100Hz, crest factor of 1, in theory, that should equate to a precise 6VACrms display on a properly calibrated true-RMS meter, right.

Interestingly, the display on our new Siglent SDM3055 shows "6.0004" but on all of our less-than-1-year-old Fluke 87's the display is "6.011 to 6.013" on their 6V range (They have about 10% overrange capability).

It is my understanding that they both use similar TRMS conversion protocols. Comments?

[bdunham7]

With a source of a precise, 50% duty cycle square wave, 12.000VDC @ 100Hz, crest factor of 1, in theory, that should equate to a precise 6VACrms display on a properly calibrated true-RMS meter, right.

Interestingly, the display on our new Siglent SDM3055 shows "6.0004" but on all of our less-than-1-year-old Fluke 87's the display is "6.011 to 6.013" on their 6V range (They have about 10% overrange capability).

It is my understanding that they both use similar TRMS conversion protocols. Comments?

I believe they both use AD chips, but different versions.  What is the "precise" source of your sqware wave and how precise is it specfied to be?

A TRMS meter should do a reasonable--but not perfect--job in this instance.  Again, I concentrate on the source of errors and there are some additional sources of error when using a non-sinusoid as your test stimulus.  The specifications for the Fluke 87V allow for up to 64 counts of error (1% + 4 counts) for signals within 1kHz BW and then 2% for signals above 1kHz BW.  To begin any analysis of the errors involved with non-sinusoid signals, calculate the contribution of the harmonics beyond the 1kHz point, in this case the 11th harmonic and up.  Or just Google it. 

[Majorassburn]

With a source of a precise, 50% duty cycle square wave, 12.000VDC @ 100Hz, crest factor of 1, in theory, that should equate to a precise 6VACrms display on a properly calibrated true-RMS meter, right.

Interestingly, the display on our new Siglent SDM3055 shows "6.0004" but on all of our less-than-1-year-old Fluke 87's the display is "6.011 to 6.013" on their 6V range (They have about 10% overrange capability).

It is my understanding that they both use similar TRMS conversion protocols. Comments?

I believe they both use AD chips, but different versions.  What is the "precise" source of your sqware wave and how precise is it specfied to be?

A TRMS meter should do a reasonable--but not perfect--job in this instance.  Again, I concentrate on the source of errors and there are some additional sources of error when using a non-sinusoid as your test stimulus.  The specifications for the Fluke 87V allow for up to 64 counts of error (1% + 4 counts) for signals within 1kHz BW and then 2% for signals above 1kHz BW.  To begin any analysis of the errors involved with non-sinusoid signals, calculate the contribution of the harmonics beyond the 1kHz point, in this case the 11th harmonic and up.  Or just Google it.

That was also my understanding that they both use similar AD chips. Of course, the Fluke 87V's are all well within specs. In fact, they appear to be within 0.25% rather than the currently published 0.7% in the User Manual (as compared to the 1% in the Cal Manual) according to my calculations. What caught my curiosity is that all 8 of the 87V errors agree so closely. That tells me it is by design.

I tested a few of the Flukes on sine waves rather than square waves and their errors decreased by about 50%. No big deal, just a heads up that there is a possible benefit to sourcing sine vs. square waves when checking TRMS DMM's.

[ebastler]

I saw that this explosively growing market had a gap of available, affordable (very low cost) but still useful "ballpark" references to do some function testing of their low cost devices.

Frankly, I don't see that market need. I already have a "ballpark reference", namely the digital multimeter I bought. If I were to spend time and money to obtain another reference on top of that, I would want something a bit more definitive.

"I have always wondered whether I can trust my meter. Nevermore! Now I wonder whether I can trust my cheap homebrew voltage reference."