PAW9 - MrRef

[matseng]

As mentioned in an earlier post I'll try to do one small PCB project every week and then send for PCB's of it every Monday.

This week I'm in Bangkok for some business meetings, but it won't stop me from doing this weeks PAW.

Since the last PAW was about as simple as it can be, just a small pcb with a bunch of holes drilled into it, I'll plan to do something a little bit more fancy for the PAW.

Unless you have a properly calibrated "serious" DMM like the Agilent 34410A or the 8.5 digit Agilent 3458A or even a Fluke 87V you really can't be sure that your $20-$50 DMM is displaying values that are not too far from the real value.

So a high precision voltage (and possibly also a semi-high precision current) generator might come in handy to check the accuracy of your DMMs.

The MAX6350-series are high precision voltage references (max 0.02% error) and a typical tempco 0.5ppm/°C (=0.00005%) so it basically won't be affected by temperature variations at all.

The 0.02% error might sound very good, but it's actually 1 mV for a 5 volt reference, so the output could theoretically be 5.001 volts. This is good enough to calibrate a standard 3.5 digit DMM, but it might be nice to have something more. Luckliy the output can be adjusted for higher precision if needed, something that of course woule require a properly calibrated 6.5 or 8.5 digit meter to do...

If I would to sell a few of these boards (pre-built) I'm thinking of having my 6.5 digit Agilent re-calibrated and certified so those MrRefs could be calibrated for an order of magnitude higher precision than the initial 1 mV.

The MAX6350 is available in 2.500, 4.096 and 5.000 volt versions and I'm leaning towards the 5 volt version and then possibly either include a selected high precision resistor to add a 1mA output on the board as well or including a voltage divider to add a 1 volt output.

The chip needs at least 8 volts to do its job properly so I plan to run it from four lithium button cells in series.

I think that the 1mA current output would come in more handy than a 1 volt output - what do you think?


The output terminals would of course be gold plated 4mm banana jacks, but I'm debating with myself if I have to put them at the "standard" 3/4 inch distance or not. It would be much easier to make a small pcb without this constraint. Any thoughts?

[dolabra]

here is mine --- viewtopic.php?f=56&t=4815.

Much lower precision, but also very cheap and good enough for my crappy meters. I went the Makey-Makey route and just put plated holes for aligator clips --- adding to the inaccuracy I am sure.

[matseng]

Yep, the ADR510 is very cheap at about $1.75 in singles compared to the MAX6350 that is more like $12 so there's quite a price difference there. And the max 3.5 mV error you get on the 1 volt ADR510 is good enough for most low cost meters.

[matseng]

The first revision of the schematic isn't too complex.

Two thin batteries will be put in each battery holder for a total of 12 volts. The trimmer can be left out and it will still be 5 volts at 0.02% accuracy (I.E +/- 1mv)

I wonder if the undervoltage detection circuit is good enough to have the led go out at vcc voltages below 8 volts? At low currents a zener have a quite large slope from conductive to con-conductive. And it needs to be a quite low current since I don't want to spend too much of the battery mAh budget on having a led lit. 0.5 or 1mA should be quite enough for a small high brightness led as a battery ok-indicator.

Since I'm away from my lab this week I would be very grateful if anyone could hook up a zener, a resistor and a led in series and connect that to a variable power supply and see if it has a decently defined cutoff-point at about a 1 mA current. If 0.5 volts change would make it go from visibly lit to dark it should be ok.

MrRef with 5.000 and 1.00 mA outputs

[Alex555]

I would use 1 volt and 1 mA. I really need to buy a decent meter though.

[matseng]

Layout on a SoB DP5050 board...

[nickjohnson]

Nice project!

I'm not sure it's worth battery power, though - most people will use it infrequently enough it'd be simpler to add a DC jack, wouldn't it?

[Erl]

I'm not sure it's worth battery power, though - most people will use it infrequently enough it'd be simpler to add a DC jack, wouldn't it?

I suspect the batteries are to avoid the noise from a switch mode power supply.

[matseng]

I suspect the batteries are to avoid the noise from a switch mode power supply.

That is absolutely correct.. I want to avoid having a lot of switchmode hash floating around here in order to keep the voltage as stable and nice as possible.

I'm sure it can be filtered out with a lot of caps and inductors, infact I've alredy got an input filtering cap, an output filtering cap and a noise reducing cap in the schematics for running on battery. Better safe than sorry.

But it is also more convenient to use batteries as the power source. Just turn it on and it works - unless the batteries are dead :-( The batteries should last many hours though since the power consumption is very low.

[Sleepwalker3]

Mats, you may want to checkout this project that Silicon Chip mag did a few years back

http://archive.siliconchip.com.au/cms/A_111365/article.html

Myself, for something like this, I'd prefer to pay an extra little bit and get better accuracy, but others may not care much. You can only see part of the article without paying unfortunately (it's a commercial magazine). I have the article around somewhere, might try to dig it up if I can get a moment. Unfortunately none of the local companies took it up as a kit.

[matseng]

The AD588 is about double the price compared to the MAX6350, got 0.03% vs 0.02% initial accuracy and a tempco of 3ppm vs 0.5ppm...

The design might be interesting anyways since it has remote sensing for compensating for voltage drop in the cables. But if your meter is drawing enough current to cause more than a 1 mV drop you might what to change your meter anyways. :-)

[nlhans]

You may want to consider use a low-voltage reference instead of 5V. My reasoning is that 4000-count multimeters will autorange up to a 40.00V range at 5V. This way you only get 10mV resolution.
If you use a 2.5V reference, its measured in the 4.000V range and you get 1 mV resolution.
It may be a bit 'overkill' for a 4000 count multimeter, because they are usually 0.25%-0.5% accurate, but 40,000 count multimeter experiences a similar situation, and it may be interesting in that situation.

A resistor divider on the output can also work, but you will need to get some really really expensive resistors to not affect the accuracy that much. Something like a LT5400.

[fcobcn]

Made a very simple voltage reference using the max6350 2.5v and a 9v battery a couple of weeks ago.
My cheap multimeter is spot on, something I didn't expected.

In any case, I would rather have one of Matseng's boards : )

[rickrcomm]

Great idea! I just got the parts from Digikey for this. I went with the 4V version for the scaling reason - my Fluke up ranges at 4.1 volts. I also decided on a 400 ohm current source resistor to make the current reference 10ma vs 1 ma. Seemed like that would give more digits of accuracy (precision, whatever.)

Question: is the Zener in the LED ckt to give a low battery voltage indication? What should the Zener value be?

Thanks,

Rick