Open Source Multimeter

[codeboy2k]

A bit more to add  regarding the protection diodes... The BAS116 is about $0.25, and has 5-10nA of reverse leakage current. 

The diode connected 2N4117A has up to 200fA leakage at Vgd around -20 to -25VDC ..depending on where your signal ranges are set, then your max signal might actually be lower than this, and the leakage could be as low as 150fA @ Vgd -10VDC .  It's a great part, but the 2n4117 is extremely expensive and mostly non-stocked.   I might consider a purpose made low leakage diode before I chose this part, unless I absolutely needed it's fA leakage (like in a 7.5 or 8.5 digit meter)

Finally, the much cheaper jelly bean 2N3904 BJT has a typical reverse leakage of between 30-50 pA, and can go as high as a few hundred pA.  These are nice because they are available as MBT3904 in a TSSOP-6 or SC70-6 package for $0.25 so you get two on the same die with similar leakage for $0.12 each.

The MBT3904 is what I am using for my LCR.  I also happen to have many around here so they are readily available to me. Depending on your leakage budget, you can use them as is with an expected leakage max of 200pA (if that works for you) or you can bin them and choose parts with 10-20pA, an order of magnitude smaller.

I use a 1k input resistance, then the MBT3904s, then a 100 ohm, then the opamp input buffer. The 200pA max leakage across the 1K series input resistance is 200nV error max, and you can bin for less.  I use a 24bit ADC with a 1.00V reference, so 1 LSB is 60nV, and this error is > 3LSB but that's ok for me, I only want 22 bits, so I can accept 4LSBs error.  If it turns out that I have other error sources elsewhere that I can't control and blows my error budget, then I can bin the MBT3904s for better than 60pA leakage, or 1LSB and bring it back under control.

No matter what, the cheaper 2N3904, even if you use it in a TO-92, is much better than the BAS116 diode.

Cheers!

[mrflibble]

Finally, the much cheaper jelly bean 2N3904 BJT has a typical reverse leakage of between 30-50 pA, and can go as high as a few hundred pA.  These are nice because they are available as MBT3904 in a TSSOP-6 or SC70-6 package for $0.25 so you get two on the same die with similar leakage for $0.12 each.

The MBT3904 is what I am using for my LCR.  I also happen to have many around here so they are readily available to me. Depending on your leakage budget, you can use them as is with an expected leakage max of 200pA (if that works for you) or you can bin them and choose parts with 10-20pA, an order of magnitude smaller.

Thanks, that sounds like a good candidate for the protection diodes.

Stupid question though, how do you get a reasonable estimate of what to expect for leakage current (in diode connected configuration) from the datasheet?

[SeanB]

Icex is a good place to look, as it is a measure of die leakage. You might want to look at higher voltage devices as well if you are going to have a high input voltage, though they will be likely to have worse leakage but still they will be better than a regular signal diode.

[mrflibble]

Icex is a good place to look, as it is a measure of die leakage. You might want to look at higher voltage devices as well if you are going to have a high input voltage, though they will be likely to have worse leakage but still they will be better than a regular signal diode.

Thanks. I understand the concept of "is a measure of", but that sometimes is a bit limiting in use... 50 nA max for the 2N3904 ... you state "can go as high as a few hundred pA", so lets say 500 pA. So that's a factor of ~ 100. As a wild guess ... leakage current for the diode connected bjt is very roughly collector cut-off current divided by Hfe?

Regarding the collector cut-off current, for example the 2N5962 looks good with only 2 nA max. Compared to 50 nA max for the 2N3904.

[SeanB]

When you read the appnotes look at the footnotes, where often the measured value is limited by the test equipment in the ATE cell. You basically have to do it yourself and see the best parts to use, or just buy a lot of cheap ones and bin them in a test jig for pass/fail devices.

[mrflibble]

Good point. Don't know how feasible it is going to be to run those through a test jig. I don't mind some testing because this is a hobby thing. But there's a limit to what I am willing to have-to-bin as part of a design. And binning bjt to find the ones good enough as protection diode might be one bridge too far. Hell, just buying two 2N4117A and just be done with it is "cheaper" that way.

That said, I have to make a jig anyways to check for other leakage (integrator capacitor), so if that can be used for this as well I just might do so just for the fun of it. But barring that I am not going to go on an extensive bjt hunt, because before you know it you have a little side project.

[Marco]

Oops yeah, dunno what I was thinking there ... still don't see the point in using BJTs or JFETs when the Central Semiconductor diodes are readily available.

[sync]

Oops yeah, dunno what I was thinking there ... still don't see the point in using BJTs or JFETs when the Central Semiconductor diodes are readily available.

5nA leakage is a lot. An example, on a 100kOhm protection resistor 5nA causes a voltage drop of 500uV. Not a problem for a 3 digit meter but a bad for a 5 digit one.

[mrflibble]

Oops yeah, dunno what I was thinking there ... still don't see the point in using BJTs or JFETs when the Central Semiconductor diodes are readily available.

Well, one reason might be that Central Semi's marketing monkey needs to have his banana supply revoked. Write a datasheet for a diode marketed as low-leakage, and then you give max leakage only at 25 degrees C. Seriously? Because who needs to know leakage over temperature when they are looking specifically for low leakage, right? Or maybe it just sucks horribly at 40 degrees C. Not saying it's a bad part, but I am saying that whoever wrote that datasheet won't be getting any bonus bananas from me any time soon.

Guess we can only reasonably assume that the reverse current will quadruple every 10 degrees. Oh if only the datasheet had included an Ir versus junction temp graph. 

[mrflibble]

Oops yeah, dunno what I was thinking there ... still don't see the point in using BJTs or JFETs when the Central Semiconductor diodes are readily available.

5nA leakage is a lot. An example, on a 100kOhm protection resistor 5nA causes a voltage drop of 500uV. Not a problem for a 3 digit meter but a bad for a 5 digit one.

The central semi diode Marco refers to has 500 pA max leakage. At 25 degrees C @_@ So logically 234234000 pAmax  at 40 degrees.



Yeah yeah, roughly double every 10 degrees.

[Marco]

John Larkin is a pretty reputable source (moreso than datasheets)

[mrflibble]

John Larkin is a pretty reputable source (moreso than datasheets)

But John Larkin isn't downloadable in pdf format from centralsemi.com for easy review.

[Marco]

From the threads I linked earlier :

I got some samples of the Central CMPD6001S dual SOT-23 diodes.
Through a local rep, since their registration/password thing is so
absurd.

I measured about 50 fA leakage at -5 volts. I say "about" because this
is sorta hard to measure... 5 mV across a 100G resistor.

[sync]

The central semi diode Marco refers to has 500 pA max leakage. At 25 degrees C @_@ So logically 234234000 pAmax  at 40 degrees.

Oops! I mixed it up with the BAS116.

[mrflibble]

From the threads I linked earlier :

I got some samples of the Central CMPD6001S dual SOT-23 diodes.
Through a local rep, since their registration/password thing is so
absurd.

I measured about 50 fA leakage at -5 volts. I say "about" because this
is sorta hard to measure... 5 mV across a 100G resistor.

I read that. But hopefully you agree that there is a difference between "an individual on the internet that measured several samples" and "the manufacturer that characterized a large number from different batches". If nothing else, statistical relevance. Not only that, but care to make an estimate of the sampling bias that is encapsulated in the phrase "I got some samples of item XYZ through a local rep"? I know I don't.

At any rate, rest assured that I have the CMPD6001S in the schematics now as alternative to the BAS116.

[SeanB]

Looks like a good unit to use, and preferably at lower voltages the leakage will be a lot less. Running it at 15V should give a leakage that is close to unmeasurable with most measuring tools.

[codeboy2k]

I hadn't really looked around for specific low leakage diodes before, they've always been too expensive.  Ok, so today I looked at the CMPD6001 datasheet.  looks ok, MAX 500pA, got it... but.. where's the charts(!) in their whole 2 page datasheet???

where's the gaussian distribution (histogram) of leakage vs number of units?  I want to know statistically what to expect from their part.  This saves me having to do binning on another part, for example, if I can predict what the failure rate might be, that rate of failure might be more cost effective than time spent binning, given the part is twice as much money as qty (2) BJTs. If I can statistically trust their part to be better than 10pA 95% of the time, for example, then it's worth it to spend the extra money for these low-leakage diodes, and rework the 5% outliers that fail. 

Yes, Mr. Fibble.. that data sheet is a joke. 

[mrflibble]

I hadn't really looked around for specific low leakage diodes before, they've always been too expensive.  Ok, so today I looked at the CMPD6001 datasheet.  looks ok, MAX 500pA, got it... but.. where's the charts(!) in their whole 2 page datasheet???

You are being generous. It's one page packaging info, 1/2 page standard and 1/2 page with some actual information.

where's the gaussian distribution (histogram) of leakage vs number of units?  I want to know statistically what to expect from their part.  This saves me having to do binning on another part, for example, if I can predict what the failure rate might be, that rate of failure might be more cost effective than time spent binning, given the part is twice as much money as qty (2) BJTs. If I can statistically trust their part to be better than 10pA 95% of the time, for example, then it's worth it to spend the extra money for these low-leakage diodes, and rework the 5% outliers that fail. 

Such information would have been nice yes. Given the anecdotal evidence of low leakage (the thread Marco linked to) and the awesomely informative one-liner spec of 500 pA max @ 25 C, the part looks interesting enough that I will try it. But the datasheet doesn't really inspire all that much confidence.

Yes, Mr. Fibble.. that data sheet is a joke.

Heh, no kidding. That datasheet deserves an honorable mention here.

[Marco]

Well if you really want a low datasheet value I don't think you're going to do better than the LM662, it's cheap too.

Will just slightly obfuscate the schematic.

PS. this whole discussion is putting the cart before the horse though, you'll not find many muxes or opamps without input protection ... unless you build your own MOSFET input stage or something you'll get clamping diodes as a side effect of your part selection ...

[Felicitus]

I'm also on my way to give this a shot. Can't promise anything though…

A few considerations and ideas:
- 4½ digits is what I consider the minimum and yet maximum reasonable for an open source multimeter. Everything else requires a lot more effort in the analog frontend.
- System performance could be optimized  with software tricks by using the internal temperature sensor (this is just an idea, I have done absolutely no research on this so far)
- Use 4 full-fledged, fully configurable channels - switching between V and A could be done with bipolar relais (this is also just an idea, and no further research has gone into that)

Notes on the design draft, which I have done too early due to a cold and boredom in bed:

- The draft below uses components which you can buy, including the case
- The case is absolutely huge with 209x98x34mm (bigger than the Metrahit 25S), and a bumper is also available
- The 4mm plugs fit in the case with a headroom of ~3mm
- Uses tact switches and the MEC NaviMec solution. The navigation serves dual purpose - when a menu is selected, it serves as navigation buttons. Outside of menus it selects the multimeter functions. Pressing 2 buttons activates the function in that segment (no idea how to describe that)
- It's using the 2.7" Sharp Memory LCD, which is slightly too small for the case, but probably I can fit everything in the Bopla BS 703 case, which is 20mm smaller in height than the BS 803.
- It's notable that the cost for buttons, display and casing is already 138.79€ including VAT at quantities of 25 copies. The price includes the switches, LCD, case and input jacks (64€ for the input switches alone). Let's be realistic about the price - even if this project works out, it won't come cheap, unless it's possible to save costs somewhere. However, I'm not intending to trade of awesomeness-factor against cheaper parts; personally, even if I didn't do the project, I would be absolutely willing to pay 300-400€ for such a meter, as it replaces up to 4 meters.

That draft was put together in probably 3 hours. Things might be missing, plain wrong, not doable etc - yes, I will figure that out during the further design process.

[mrflibble]

PS. this whole discussion is putting the cart before the horse though, you'll not find many muxes or opamps without input protection ... unless you build your own MOSFET input stage or something you'll get clamping diodes as a side effect of your part selection ...

What discussion? You mean the side track about diodes? Not as if that was the main track as far as I'm concerned. The horse & cart were already properly located even before the random discussion. But now thanks to the semi-random diode discussion I have a few alternatives for the BAS116 that was already in place. Around the opamp that was also already in place.

It's the way of threads like this. People come and go, while leaving some remarks that happen to interest them at the moment.

[Felicitus]

It's the way of threads like this. People come and go, while leaving some remarks that happen to interest them at the moment.

I agree. Different people have different approaches to design things, and none of those is more correct than the other.

[Marco]

But now thanks to the semi-random diode discussion I have a few alternatives for the BAS116 that was already in place. Around the opamp that was also already in place.

My point was that with the LTC6240 I think you're using and a 100 kOhm input resistor you're already good to 1 kV (10 kV transient) any way without diodes.

[mrflibble]

My point was that with the LTC6240 I think you're using and a 100 kOhm input resistor you're already good to 1 kV (10 kV transient) any way without diodes.

Speaking of input resistors...

In sync's post about the HP 3456A input stage you can see the 4 resistors, with 2 on each side of the relay. I get the one before + one after thing for EMF purposes. Having 2 per side is also for EMF compensation I presume?

[codeboy2k]

I would say two resistors per side (4 in total) is designed to increase the voltage standoff ability.  For any high voltage transient event, each of the 27k resistors would be dropping 1/4 of the voltage, and if you make those large body resistors, then they would be able to withstand 1kV each, better than 1 resistor that has to take 4kV (which would likely flash over a small resistor).

For example, 4 resistors in series would have 1kV on each resistor and would not flash over from lead to lead.