Another DIY voltage reference...
 

Hello,

The recent discussions 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. Rather than clutter up Al's thread, it made most sense to start a new one.

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 idea for their own projects :thumbsup:

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

Hope this is of interest,

Mark

Re: Another DIY voltage reference...
 

Now that IS a well documented project.
Furthermore, it took you more than just a few minutes.
Well done that man!
Les.

Re: Another DIY voltage reference...
 

Excellent in all respects! A very neat idea, Mark.

I've been toying with the idea of using a single 2.5 volt precision reference in conjunction with a divider chain of matched resistors. But it's still just an idea, whereas you've already built yours! Well done.

Re: Another DIY voltage reference...
 

Remember that a 2.5v reference, however good, won't tell you that the higher ranges are any good if scaling resistors have gonebad or been damaged.

Properly calibrating a DMM means checking all the ranges.

David

Re: Another DIY voltage reference...
 

Absolutely true, David. However I was assuming the long-term stability of modern resistors. It would indeed be useful to have a set of precision voltages up to say 250 volts which, together with other tests, would be sufficient to check the calibration of most multimeters.

Re: Another DIY voltage reference...
 

Thank you for your kind comments :thumbsup:

One of my goals was to make something that was slightly more useful than the Chinese modules, so this unit has 10 output voltages rather than just 4. That's enough to let you check the functionality (not calibration) of the lower ranges of a typical DVM.

This unit is pretty good for that, and I must say that it's quite satisfying when you get away from the Keithley with its merciless resolution! A typical 3.5 or 4 digit meter generally gives an "on the nail" reading at all output voltages, which gives me a nice, warm feeling ;)

Certainly, if you wish to calibrate a meter, this unit is not really good enough on its own - and nor was it intended to be. However, in conjunction with a decent meter like the Keithley, things get a lot more interesting...

But it does depend on what the meter is. Some are calibrated via a single pre-set. Others have multiple presets for different functions. Most decent meters today are entirely done in software, with no preset resistors anywhere - DIY calibration is not an option.

If you want to run an entire calibration, you need serious hardware. I dread to think what this must cost: http://us.flukecal.com/products/elec...ction-calibrat

Having used the word "calibration" several times, it's probably worth pointing out that calibration does not involve adjustment. That's one of the big electronics myths - up there with "transistors are current-operated devices". Calibration is just a process to quantify the errors - how far does this instrument deviate from the standard, and is it in-spec? That's all. Adjustment is a whole other ball-game, and in many environments, adjusting a meter every 12 months would be disastrous. If you want adjustment, you have to request it separately.

But yes, I did think about higher voltages, which would need a boost converter and a high voltage op-amp of some sort, and I might build something in the future, just for the fun of it. I have got a couple of Maxim precision references (3ppm/C) that are crying out for a project. And there is something addictive about the quest for accuracy...

But naturally, DC voltage is just one part of the game - you'd also need AC volts, DC and AC current, and resistance standards as well. Where do you draw the line?

With a decent multimeter, the precision resistor networks determine the accuracy and precision of the instrument. These are well-protected by the input circuitry. I reckon that if a good meter is OK on its 100V, 10V, and 1V ranges, it'll also be OK on the 1000V and 100mV ranges. You can manually range to the 1000V range to quickly check, of course.

As you move away from KISS solutions like this, it's worth saying that anything involving amplifiers - even just buffers - is a lot more complex. Op-amps have offset voltages and currents, and while you might trim them out, they vary with time and temperature, so they become a systematic error. Chopper-stabilised amps can be used, but these are noisy. And of course, they will impact on battery life. I agonised for ages about this, because adding a unit-gain buffer to the output of mine would have made it even better, providing the cure wasn't worse than the basic problem ;)

Thanks again,

Mark

Re: Another DIY voltage reference...
 

It is also well known that if you set up a "guess the number of sweets in a jar" or "guess the weight of a pig" competition and have a sufficiently large number of people guessing, when you take the average of the guesses the answer is surprisingly close to the correct one. You can't quantify the tolerances of the answers in advance and so the maths of the explanation will be different. A simple explanation says that the high guesses tend to cancel the low guesses but I'm sure its more complex then that.

The passive component case was not something of which I was particularly aware. I thought if you took some 1% components to make up another value the result would be another 1% component, but as you have explained the real result is better than that.

Nice project BTW Mark and a very interesting read.

Re: Another DIY voltage reference...
 

IF the errors are randomly distributed (for example Gaussian distribution) then sigma does indeed go down with the square root of the number of elements involved. For the 5 element example a bit more that 2X better !

dc

Re: Another DIY voltage reference...
 

Thanks, Colin :thumbsup:

I first ran into this when Douglas Self published a pre-amp design back in 1996 - he was aiming for the best RIAA curve he could manage using non-exotic components, and of course, it's the capacitors that are a problem. The design called for a 50nF capacitor, so he used five 10nF 1% polystyrene types that gave 50nF at 0.48%. I was impressed!

At the same time, the Maplin catalogue had a "box-out" that explained that combining 1% resistors gave you an overall tolerance of 1%. Hmm...

Cheers,

Mark

Re: Another DIY voltage reference...
 

Absolutely ;D

For what it's worth, I did put my samples into an on-line histogram generator, and got fairly convincing results - certainly both datasets looked pretty Gaussian to me. I didn't include the results on the webpage, but perhaps I should? I was unsure because the sample sizes are relatively small...

All the best,

Mark

Re: Another DIY voltage reference...
 

Nice work Mark, that is a very neat veroboard, you can see a lot of attention to detail and work's gone into it. That and a clever, simple solution. I'm very impressed and will have to up my game when I do my next project.

Andy.

Re: Another DIY voltage reference...
 

Thanks, Andy ;)

The secret of Veroboard is simple: 0.1" graph paper :thumbsup:

I'm not sure how easy it is to find these days, but there are websites that can create it for you. I'd guess that a laser print would be better than inkjet, as you'll do a lot of rubbing out as part of the process - you'll see what I mean when you see this example: https://www.vintage-radio.net/forum/...1&postcount=52

If you haven't see it before, that thread is full of useful tips about Veroboard...

Re: Another DIY voltage reference...
 

Thanks Mark, handy tip - hadn't seen that.

Andy.

Re: Another DIY voltage reference...
 

I place the Veroboard, track side up, beneath a sheet of paper and rub a soft pencil over the sheet to obtain a full scale 'brass rubbing' on which I then develop the layout, using the components to hand. I guess that alternatively you could photocopy or scan the board.

Re: Another DIY voltage reference...
 

I would really like to have a go at building one of these sometime.

One question, with the inclusion of a preset resistor, is an existing separate reference voltage required to adjust the unit?

I haven't got a reference at the moment but I need to measure a 12 volt rail in a vintage oscilliscope accurately.

Is there anyone local(ish) to Walsall who has a calibrated DVM that I could compare my thirty year old BT Digital to?

Re: Another DIY voltage reference...
 

Hi Gents, on the topic of range extension, how about using a bridge circuit with a low current 250v supply and a 20:1 divider with "statistical" resistor chains as set out above for good accuracy.
This divider could be nulled against the 12v5 ref voltage with a centre zero galvo or electronic meter.
With the meter to be checked connected, the 250v supply could be adjusted to give null and the meter reading noted.

It would reduce the accuracy of the system at this voltage but should still form a useful check on the meter's higher voltage ranges.

Thoughts please gents.

Ed

Re: Another DIY voltage reference...
 

Brilliant :thumbsup:

Re: Another DIY voltage reference...
 

The preset is for the divide-by-10 range, so you can use anything to calibrate it, providing it is linear. The more digits the better, and preferably use the same range in case there are range-to-range errors. Ideally, when you switch ranges, all that happens is the decimal point moves ;)

If no-one is nearer, I'm not a million miles from you.

Re: Another DIY voltage reference...
 

It's a good idea.

I'd form the upper resistance from resistors in series, as many types of resistor have a voltage dependency at high voltages.

Obviously one has to be careful with those high voltages. And the simple Zener/FET protection that I built in becomes essential - should something go wrong, you might end up trying to dump excess current into the references. Stating the obvious, perhaps, but most DIY/cheap references omit protection.

For that reason, it makes most sense to make the lower resistance from parallel resistors.

Re: Another DIY voltage reference...
 

For 1% SMT resistors at least the errors in the values are systematic and a given reel tends to have resistors very well matched to each other even if up to 1% from their nominal value. At work I have just measured a 901:1 voltage divider made from one hundred 100k 1% 0603 resistors as having an error of better than 0.03%. That's considerably better than the error you would get with resistor values randomly distributed in the 1% range.

Only works if you're careful to ensure they're all populated from the same reel.