DIY voltage reference for DMM calibration

[exe]

Hi guys,

I wish I could buy the famous DMMcheck from www.voltagestandard.com . But they don't even ship to my country (besides hussle with customs). So I would like to build one by myself. I'm looking for a good (and not complicated) scheme, but so far I only found this: http://www.avdweb.nl/tech-tips/multimeter-calibration-diy.html . May be someone could recommend something else? Particularly, switching voltage by a button would be useful .

If not, then I would go with the link above, or just with lm317 + some precision  reference like lm329, lt1021 or something.

May be someone could also enlighten me how to solder precise components? Apart from avoiding overheating I heard there could be a nasty impact on precision from thermoelectric effect of solder joints, how to avoid it? Can I use a conventional protoboard instead of a custom PCB (no breadboard of course)?

[jpb]

I am surprised that they don't ship to Italy. Have you e-mailed them?

When I'm soldering through-hole parts that are heat sensitive I use solid copper crocodile clips as heat-sinks on the leads. I've not done any scientific studies as to the effectiveness but someone else on these forums suggested using them a while back.

Joe Geller closed his labs a couple of years ago, but I managed to buy a pcb off him about a year ago (I've yet to make it up though) so you may have some luck if you e-mail and ask him nicely - it all depends if he has any old stock. It is a very interesting site to go on still. I particularly like his article on making a 10:1 divider.

http://www.gellerlabs.com/

[exe]

I am surprised that they don't ship to Italy. Have you e-mailed them?

Not yet because I think our customs will be a big-big hustle for me. I'll furst check how much time I will spend filling-in declarations, sending mails (not e-mails, but, ughm, paper communication, though now I think I can send them a scan) etc. It can be easy and straightforward, or it can be such a big pain that it is easier to decline the parcel rather than fight for it.

Joe Geller closed his labs a couple of years ago, but I managed to buy a pcb off him about a year ago (I've yet to make it up though) so you may have some luck if you e-mail and ask him nicely - it all depends if he has any old stock. It is a very interesting site to go on still. I particularly like his article on making a 10:1 divider.

Yeah, I came across his site yesterday. I did my homework before posing this . He is in the US, right? Then this can be a problem. Last time I checked no matter what you order from the US you have to deal with customs. Thanks for the tip with crocodile clip .

So, building an own reference can be an option. Or may be not, I see some skills, precise equipment and deep understanding are required to do it properly.

[michaeliv]

This comes with readings so it will probably be more accurate than anything that can be built with off-the-shelf components:
http://www.ebay.com/itm/-/181219600361
Ships from China, review Here : youtube.com/watch?v=zm_ZBT4t4dA

[suicidaleggroll]

What kind of precision do you need?

An LT3080 plus LTC6655 is easy to design yourself and will go a long way.  I built a board with the 5V and 2.5V LTC6655 on it, they read a steady 4.9999 and 2.5000 respectively on my Rigol DM3058.

[exe]

What kind of precision do you need?

An LT3080 plus LTC6655 is easy to design yourself and will go a long way.  I built a board with the 5V and 2.5V LTC6655 on it, they read a steady 4.9999 and 2.5000 respectively on my Rigol DM3058.

I want to be sure my uni-t 61e does not drift too far . I think +-1mV precision is enough at 3.3V. Soo... as I understand, I can just repeat you arrangement (lt3080+ltc6655-XX) on a piece of a protoboard and.. it's done?

[suicidaleggroll]

Don't use a protoboard for high precision stuff.  You need a solid ground plane, good decoupling, etc.  The LTC6655 datasheet has a lot of information about the recommended layout, and they also have a demo board with full schematics, BOM, and layout that you can reference.

I laid out a 2" x 2" 2-layer PCB for mine and had it fabbed at OSHPark for about $20 for 3 copies.  Add in $4 for the regulator, $15 for the reference, a few bucks here and there for decoupling caps and headers, and you're looking at about $100 out the door for three of them, or $50 for one.

[altaic]

If it's just a one-off going into a box, go dead bug style for both highest performance and lowest cost. The LTC6655 is available in a MSOP-8 package, so you can solder caps and such directly to the leads.

[hli]

I build my own reference a while ago: https://blog.hendriklipka.de/archives/2014/05/volt_multi_ref.html. It has five references from 1.25V up to 10V, so you can test multiple ranges regardless of what the counts of your DMM are. Its accurate enough to verify my not-so-expensive multimeters. And since I got the references as samples it wasn't that expensive to build... As long as the reference comes in SO8 package, the footprints are typically the same. So you can use what you have already available.

[exe]

You need a solid ground plane

Could you please clarify why? My impression from reading the datasheet of LTC6655 is that the precautions are needed for sourcing that reference voltage to the load (to avoid voltage drop over long traces). But multimeters have quite high input impedance (and long test leads), so this shouldn't be a problem.

I'd add some shielding, though.. But I see most of them are not shielded and this makes me wonder why. May be because test leads are not shielded, so shielding a small box wouldn't add much.

I build my own reference a while ago: https://blog.hendriklipka.de/archives/2014/05/volt_multi_ref.html.

I'll go this way, thank you. So, a battery, pre-regulator and a few references.

[suicidaleggroll]

Could you please clarify why? My impression from reading the datasheet of LTC6655 is that the precautions are needed for sourcing that reference voltage to the load (to avoid voltage drop over long traces). But multimeters have quite high input impedance (and long test leads), so this shouldn't be a problem.

It's for noise rejection.  This is arguably not important for your specific application of testing the long-term accuracy of a DMM, but high precision references have uses that extend far beyond the occasional DMM test.  Compromising its noise performance in order to save a few bucks just because that aspect of the reference isn't that important for this exact task is a waste of an expensive reference IMO.

[exe]

It's for noise rejection.

Why they are not fully shielded? Like DDMCheck, etc. I mean, most of them are in plastic enclosures or even just a bare pcb like dmmcheck.

[hli]

I'll go this way, thank you. So, a battery, pre-regulator and a few references.

Important: check the data sheets for the references you want to use. They differ all in what the accept and/or require in capacitance on their outputs. I created a somewhat generic PCB, which could hold up to 3 caps per reference. Then I populated only the ones I needed after selecting the references. And to avoid self-heating in the chips, maybe you should use different pre-regulators.

[exe]

to avoid self-heating in the chips, maybe you should use different pre-regulators.

I thought about this and I find this shouldn't be an issue: each reference needs only about 1-3mA and I'm not going to use more than four of them (because they are expensive!). Anyway, a few tens of mA shouldn't heat-up too much the pre-reg, imho. I faced such problem with lm334z when drawing ~5mA, but this shouldn't be a problem for, e.g., lt3080 which supposed to be much more stable and ceramic-friendly regulator.

So, what I've done, I wen't to digikey, chose the right output voltage, sorted by accuracy (I don't have precise DVM to measure the actual output, so I want a factory-trimmed chip) and stability. So finally I chose ADR4520BRZ and thelike. I'll install it on a soic-dip adapter, hope this will provide mechanical relief.

What do think about this? Is ADR4520BRZ any good? Anyway, I think the biggest problem is not the chip itself, but the one that sits between chair and keyboard.

[hli]

I thought about this and I find this shouldn't be an issue: each reference needs only about 1-3mA and I'm not going to use more than four of them (because they are expensive!). Anyway, a few tens of mA shouldn't heat-up too much the pre-reg, imho.

The issue is not about the power lost in the regulator, but about the power dissipation of the references. When you have a 10V reference, you need about 12V input (or thereabout). This means a 1.25V  reference needs to handle 11V - and this causes heating in the reference so its voltage drifts a little bit. Maybe its not much, but when you can avoid it then you probably should...

What do think about this? Is ADR4520BRZ any good? Anyway, I think the biggest problem is not the chip itself, but the one that sits between chair and keyboard.

I thinks its a good choice. Quite good accuracy, low temperature drift, and not that much noise. You can look at Scullcom's where he discusses what to look for in a reference (he did chose the ADR4540 as 4.096V reference).

[exe]

The issue is not about the power lost in the regulator, but about the power dissipation of the references. When you have a 10V reference, you need about 12V input (or thereabout). This means a 1.25V  reference needs to handle 11V - and this causes heating in the reference so its voltage drifts a little bit. Maybe its not much, but when you can avoid it then you probably should...

Ah, I see, thanks. I decided to use only 2.048V, 3.3V and 5V references, so I'll set the pre-reg output to 5.5-6V. I think there won't be much self-heating to screw the results.

Concerning 10V references. Their precision is specified in % of voltage. So, they are not very precise in terms of millivolts and I'm not sure that having 10V ref is beneficial in addition to 5V (other than to verify full-scale error if there is a non-linearity). Also, my junky ut-61e has only one trimpot (or two) so I cannot correct any non-linearity. But I can trim it around "sweet spots"  (2.048  and 3.3V) as these I use the most.

One day I will go for real DMM, but right now this won't pay me back.