Simple DIY DVM Voltage Reference

[Mark Hennessy]

Hello,

Over on UKVRRR, a recent discussion about DVM voltage references coincided with some investigation work that I'd been doing, and I've finally found a bit of time to finish the project and document it.

That thread contained plenty of discussion that I don't need to repeat here, so I'll launch straight into an overview of my design:

The core idea was to combine a number of cheaper voltage references to get a better overall accuracy. It's well-known that combining passive components in series or parallel results in better accuracy, but the same argument should also apply to voltage references. To see if that was the case, I spent a few days measuring 60(!) of them, and went ever so slightly mad in the process. However, it works

The ICs in question are the "good but cheap" LM4040. The D-grade parts are rated to 1%, but in practice most samples proved to be better than 0.5% (which matched my expectations, based on my previous experience of them). The more expensive A-grade parts are specified to 0.1%, but I did find that not all of them met this spec, which was a surprise. More sampling and investigation required.

Anyway, four 2.5V 0.1% references in series reliably gives 10V at 0.05%. No selection or adjustment is needed, beyond checking that your samples meet the 0.1% spec.

Having done that, you also have 2.5V, 5V and 7.5V available. And while you're at it, why not add a 5th option for 12.5V? And a divide-by-10 option doubles the output options to 0.25V, 0.5V, 0.75V, 1V and 1.25V.

Finally, add a low-battery indicator and output protection (the latter always seems to be missing from the DIY and Chinese offerings). An afternoon with a drill and a basic die-cast enclosure completes the package.

So, a little bit different to the norm, but good results for not much cash. Not that it can compete on price with the cheapest Chinese offerings, but at least I've had the fun of building it myself, and I hope it gives people some ideas for their own projects

Lots more detail is available here: http://www.markhennessy.co.uk/dvm_reference/

Hope this is of interest,

Mark

[Nick]

Really interesting project and a great write-up   

It just so happens I did a "statistical regulator" project on low-noise HV screen regulators for audio pentodes & tetrodes last year - the project write-up is here (not nearly as detailed/impressive as yours!): http://www.audio-talk.co.uk/phpBB3/viewt...=11&t=6204 (I'm called "Jack" on Audio-Talk ) - it's based on Morgan Jones' write-up in the latest (4th) edition of "Valve Amplifiers".

In my case, the primary objective was very low noise, not absolute precision - also, I was using long strings of low-voltage Zeners so temperature correction was applied using the ZPWL model - see http://www.audio-talk.co.uk/phpBB3/viewt...45#p128632

I know noise is not a design metric in your project, but the LM317 is a noisy old thing so I used a pair of depletion-mode FETs (the ubiquitous DN2450)  as the CCS, delivering about 10mA to the Zener string. Also, with the large number of Zeners I was using (up to 80), voltage accuracy for each was not important - it averaged out to a nominal 5.6V, though the temperature & current-related compensation was significant.

As you note, in a statistical regulator the voltage accuracy improves as a function of the square root of the number in the series string - obviously, for the first one in your string (2.5V), there is no improvement - the improvements will be for the higher voltages where more devices come into play.

As a DMM reference, into MΩ input impedances, this would be fine, but being the forum we are here, maybe worth emphasising "definitely not for AVOs"

Possibly consider a precision voltage-follower to lower the output impedance so the loading becomes insignificant?

[Yorkie]

Excellent write-up Mark, with a lot of time, effort and technical expertise involved in developing the project.

It's a long time since I've seen transistor socket!

[Mark Hennessy]

Really interesting project and a great write-up   

It just so happens I did a "statistical regulator" project on low-noise HV screen regulators for audio pentodes & tetrodes last year - the project write-up is here (not nearly as detailed/impressive as yours!): http://www.audio-talk.co.uk/phpBB3/viewt...=11&t=6204 (I'm called "Jack" on Audio-Talk ) - it's based on Morgan Jones' write-up in the latest (4th) edition of "Valve Amplifiers".

In my case, the primary objective was very low noise, not absolute precision - also, I was using long strings of low-voltage Zeners so temperature correction was applied using the ZPWL model - see http://www.audio-talk.co.uk/phpBB3/viewt...45#p128632

Hi Jack,

Thank you!

That's an interesting project - I'll read the thread properly in a bit, but I've already lost half an hour there

The funny thing is, I'd never heard of the term "Statistical Reference" before I decided to call mine that! But it's proof that there's nothing new - every time I have a good idea, someone else has got there first

BTW, I have one of those Tek DVMs. The 914, so not as nice as yours. And yes, I did haul it out yesterday, and it compared well to the Flukes

I know noise is not a design metric in your project, but the LM317 is a noisy old thing so I used a pair of depletion-mode FETs (the ubiquitous DN2450)  as the CCS, delivering about 10mA to the Zener string. Also, with the large number of Zeners I was using (up to 80), voltage accuracy for each was not important - it averaged out to a nominal 5.6V, though the temperature & current-related compensation was significant.

Good point - I haven't measured the output noise of mine yet. Easy enough to do, and will do so at the weekend.

I did think about a simple JFET current source, but dropout voltage was a concern, as was stability. You should see the pages of scribbles that I dreamt up! In the end, I picked the LM317 because it's totally "jellybean" - and I was a bit surprised at how well it performs as a current source (as a voltage regulator, they aren't bad, but they're not the best).

As you note, in a statistical regulator the voltage accuracy improves as a function of the square root of the number in the series string - obviously, for the first one in your string (2.5V), there is no improvement - the improvements will be for the higher voltages where more devices come into play.

Absolutely. Selecting the "best" sample for the first in the chain is critical - which isn't good for mass-production

It'll be interesting to see what happens as these devices are used. In fact, I'm setting up the remaining 25 samples on a breadboard so I can run them in over the next few months... I'm expecting these to drift. I'll also experiment with thermal hysteresis (probably using the cheaper D-grade parts).

In short, I'm really regretting buying the Keithley now!

As a DMM reference, into MΩ input impedances, this would be fine, but being the forum we are here, maybe worth emphasising "definitely not for AVOs"

Possibly consider a precision voltage-follower to lower the output impedance so the loading becomes insignificant?

I agree - the "x1" readings will be OK, but the "÷10" mode will exhibit significant loading effects - a couple of % at worst. But then, that's not far off what an AVO can do anyway.

Even into a 10MΩ DVM there are loading effects - the 135Ω in the output protection JFET is there, and that's before you've flicked the ÷10 switch. However, as soon as you move away from the Keithley and use "normal" meters, everything is suddenly very satisfyingly nice...

Trust me - I did agonise over this. In the end, I decided against adding a buffer - this was very much a KISS project, and the references themselves are not the last word in precision. Instead, I tried to minimise the systematic errors and point out the errors so people can decide for themselves.

By coincidence, I'm awaiting delivery of a second-hand Time Electronics voltage source - it'll be interesting to see how it compares. Sadly, they put the interesting stuff in a potted box, but even so, I'll get a few ideas from there. I'm thinking about building a current source next, in the same size box. Strictly, it'll be a current regulator - the current will need to be provided externally - but that gives maximum flexibility and utility while saving battery costs...

All the best,

Mark

[Mark Hennessy]

Excellent write-up Mark, with a lot of time, effort and technical expertise involved in developing the project.

It's a long time since I've seen transistor socket!

Thanks, Yorkie

It was quite satisfying to see what could be done for relatively little outlay - as I think I said somewhere, those expensive reference ICs are a bit "brute force", whereas I was hoping to try a more "elegant" bit of engineering. Overall, I think it's a set of compromises that have been juggled fairly successfully - like all the best engineering projects.

I've had those sockets for years - they've been crying out for an application like this! There's plenty more where they came from - I'd happily donate a set of 5 to anyone who wants to build this exact project

Thanks again,

Mark

[Radio Fixer]

With all those meters you certainly need a reference Mark 

[Yorkie]

With all those meters you certainly need a reference Mark 

As my daughter-in-law says "A woman can't have too many pairs of shoes or handbags"!

Substitute handbags for meters, and there you go.

What the reference seems to indicate is that multi-meters are generally pretty much spot on.

I've got several multi-meters - the most often used one (and most expensive!) being an excellent £10.00 Toolzone one.

I checked them all  using the 10V reference I made from the EPE magazine. All more or less spot on, and quite adequate for my humble needs - I'm towards the lower end of the food chain!.

[Mark Hennessy]

That's a neat job, David

One day I'll set myself up with PCB facilities. I've certainly made them in the past, and I really enjoy the design/layout part of the process. But I don't mind Veroboard for simple one-offs - I'm used to its limitations, and spend a lot of time at the planning stage to get the layouts neat. But a PCB usually results in more compact constructions, and usually looks a lot nicer. Swings and roundabouts!

The multimeter collection is a bit out of hand. It peaked shortly after I made myself stop buying radios!

For years and years, I used an old Advance Alpha, which was only 3 digits. I still have it and one day I might photograph it and post a few details on-line, as there seems to be nothing about them available currently.

After that, I bought a unit from CPC - brand unknown - it seemed amazing at the time, being 3.75 digits and autoranging. It cost £30, which seemed like a lot to me at the time - money was extremely tight back then. It got a lot of use - indeed, I bought a couple for the workshop at work, and they were reasonably well received.

But a few years later, with a significant birthday looming, I decided to finally buy myself a Fluke. I'd used them at work for about 10 years at this point, but never owned one. What sealed it was an email from Rapid Electronics, offering "final column pricing" on anything I wanted to buy within a certain time-frame. Finally, a Fluke would be mine!

I looked at the 175 - which was at the bottom of the range, but would have been absolutely fine. But then I noticed that the 179 had a thermometer, plus a display backlight. And it's only an extra £whatever, so I might as well go for that...

Anyway, this process repeated and repeated, and before I knew it, I'd ordered the top of the range 189! I forget the price, but it would have been £300 or more. But, it's had a lot of use, and it still works perfectly, as seen in one of those photos. The only downside is the battery life.

The real shocker was just how inaccurate the no-name CPC meter was when compared to the Fluke. It was wildly wrong on several ranges, but no idea why - it had been looked after and never abused. Scary...

For years, that was fine. I picked up a vintage Philips PM2518X that was a bit kinder on batteries, and between these two, I was able to do most things. The Fluke was used whenever accuracy was required - also it was excellent for audio work as it has a 100kHz AC bandwidth and does dBm measurements. I used to take it to work several times a year because we didn't have anything good enough there. Overall, it has been a really great buy.

However, I started watching the EEVBlog videos. Not everyone is a fan of Dave Jones, but he talks a lot of sense, and he likes his test gear! And I found myself watching eBay for used Flukes at good prices - they are few and far between, but occasionally you get a bargain. For example, my 87V was £150 BIN - normally half-decent ones go for £200+, and the mint examples fetch £300 or more. The one I found was brand new and was the E2 combo kit that came with the nice case and some accessories. I sat watching it for about an hour - hoping someone else would buy it - but in the end I couldn't help myself! Check out the retail price: http://uk.rs-online.com/web/p/multimeter-kits/0237160/

Another example - a nice Fluke 45 bench meter for £70. These go for 2 to 3 times that amount, sometimes more. My Keithley cost a bit more, but not as much as you might think. Basically, a seller had obviously cleared out a big lab, and flooded the market with them - they probably sold 50 or more - so the prices were relatively weak. The early-birds got the best bargains - brand new in boxes and free postage. Mine was the same price, but it has been used, and postage was extra. Even so, I'm not complaining!

When I buy a new meter (or any other test gear), it has to be a bargain. I look for auctions that finish mid-afternoon on weekdays, or listings that are poorly done. I avoid the dog-fights and jump in at the last 10 seconds with what it's actually worth to me. And I try to buy things that I will actually use - which isn't as obvious a thing to say as you might think!

2 or 3 meters is easy to justify, and I frequently find myself using that many at once. Then you're in a zone where it's sort-of acceptable, because different instruments for different tasks, etc. But beyond that, it's an obsession - check this image: http://s185.photobucket.com/user/excavat...2.jpg.html

Luckily, I'm nowhere near that extreme, but I definitely need to slim down. Some of mine are regarded as "part of the collection", so aren't used in anger but are nice to own. But most of them have to earn their living.

That's just multimeters. I really need to slim down all aspects of my test gear collection, and am planning to make an effort this year

[Terry]

The problem with all these highly accurate DVMs, as far as fora like this are concerned, is the constant need to explain the error of their ways to novices (and some not so new to repair/restoration) when they say something like "such and such voltages read high" when the Trader sheet says 'All voltages measured with an Avo Model 7' ...

Then there's "I've changed all the [20% tolerance] resistors because they all measured [within 5% of their nominal value] high/low, and it hasn't made any difference!" One wonders how they would have got on in the more cramped areas of the AVO Model 8 resistance scale!

[ppppenguin]

Terry, as you have suggested, explaining the difference between accuracy and precision isn't easy. Nor are the concepts of adequate precision and accuracy. Then add the effects of loading by the meter. Experienced folk like you, me and Mark will normally get it right without thinking but it can be a struggle for a novice.

At a more advanced level, using a scope probe can give various problems. My favourite trick for checking if the probe is affecting the circuit is to touch a 2nd probe on to the same point. If the waveform changes significantly then probe loading is affecting the measurement.