Keithley 195A DMM Amps/TRMS option board

[qu1ck]

Hello everyone,

I just published designs for a project I worked on and off during last year. It's an add-on board for Keithley 195A dmm that enabled AC/DC amps measurements and AC TRMS voltage measurement. I tried to use off the shelf components to replicate this board so that anyone who has this DMM without the option board can build one.

KiCad and gerber files are here: https://github.com/openscopeproject/Keithley1950

BOM that also showcases the board can be viewed here: https://openscopeproject.org/InteractiveHtmlBomDemo/html/Keithley1950.html

This was fun to work on and I learned quite a bit in the process.

Many thanks to forum user @Ordinaryman1971 who gifted me the multimeter to tinker with and use as a testbed for the project.

Schematics for the board are not too complicated, it's just AD8436 chip, couple of opamps, shunt network and digitally driven switching logic.

[graybeard]

Cool!  I have a 195A with the option which I love because it measures very low (nA) currents.  It is a great meter when this option is included.  This will ensemble the capability for the many units without the option.

Do you have a photo of the board installed in the meter?

[qu1ck]

It installs in the same place where the original goes in. Just needs a few holes drilled in lower shield.



Few more pictures are on github I linked in first post, also here:
https://imgur.com/a/Ajf4ZLl

[Kleinstein]

The board to a large part looks like a near copy of the Keithley circuit - it kind of needs to to work with the DMM.

There are 2 small points that could alow for a slight improvement:
One would be some measures to reduce the current of the relays when on for a longer time. The hold current can be much lower and this can reduce the power consumption quite a bit. This could help especially K6, as here heat can cause extra thermal EMF.  The other relays seem less critical. In the DCV case the relay is off an not a problem, the ranges in question are DC-amps. So one could see the effect as small offset to the current.

The second possible upgrade would be using the buffered voltage at U1A also to bootstrap the protection diodes. This would be 2 resistors in the 10-100 K range to both ends of D3. This could reduce leakage current for the diodes, that could be there (but hard to see) at upper end of the lowest current range.

[qu1ck]

Thanks for the notes, that is interesting.
I don't think 0.36w relays used in this circuit will affect anything, they don't even get warm to the touch during prolonged operation.

But the bootstrapping of diodes is a great idea, it should divert some portion of leakage current to opamp instead of leeching it from measured circuit. As a side note maybe something with lower than 10uA Ir should be used instead of the S3M I currently have.

[Kleinstein]

The diodes for the current protection need to be quite beefy, as they have to carry the current needed to blow the fuse. So they should stand something like 1.5 times the fuse rating with sufficient peak rating. So the obvious choice are slow rectifier diodes.
The forward drop on the S3M looks relatively high though. So the shunts may worst case see quite some power.
With bootstrapping the diodes one might consider using only 1 diode. However this might start to limit peaks in AC current.
Having the diodes close together with not much cooling is a good thing as this lowers the drop when the diodes get warm.


The diodes to protect the AC amplifier should be smaller to avoid the high capacitance. So more like a 1N4148 or BAV199. That is definitely a point to change.

One could check the effect of the relay getting warm by looking at the offset in the current ranges. Ideally there would be essentially no current measured with open terminal. Thermal EMF can introduce an offset of a few µV that could be visible as some zero error in the current.

p.S: the lower shunts also look rather small form factor. So they may show quite some self heating effect with higher current. 2 A at 200 mV burden gives 400 mW power loss. This normally calls for a resistor that is more like 5 W normal power rating. The power rating of SMD parts is valid on a massive copper plane board not the more practical layouts  .  Also a symmetric thermal layout around that resistor could help to avoid errors.

[qu1ck]

It's kind of hard to notice small offsets because this meter naturally has them in all modes and requires zeroing for any low value measurement. But I'll try to see if there is drift in 0 offset in current mode. Btw you mean K3, right? That's the one switching on for highest DC amps range. K6 is for ACV on my board and is off in that mode.

I will take a note on the AC amp diodes.

Also here is a video of routing the board in KiCad. This is rev A, I fixed some mistakes later.

p.S: the lower shunts also look rather small form factor. So they may show quite some self heating effect with higher current. 2 A at 200 mV burden gives 400 mW power loss. This normally calls for a resistor that is more like 5 W normal power rating. The power rating of SMD parts is valid on a massive copper plane board not the more practical layouts  .  Also a symmetric thermal layout around that resistor could help to avoid errors.

Lower shunt is a constantan wire, it will easily take 5W. With so much surface it will dissipate 400mW without heating more than few degrees.

[Kleinstein]

The relays are not that bad - my fault  . K3 and the other relays for current switching are not critical at all with thermal EMF. It would only effect the voltage seen as a burden.
K6 is only used for DC volts and thus is off when not used. So no problem there. It would only effect settling when switching from AC to DC volts.


The footprint for the shunts looks quite small. It kind of is a balance with the TC - with a very low TC resistor more heating is not a problem, but normally it is easier to get higher power than very low TC.

For the diodes it's probably still possible to fit a much smaller diode (e.g. MELF) to the large foot-print.

[qu1ck]

That's why I chose constantan for low tempco. That said, quality of chinesium constantan is questionable but it does shine like nickel and solder like copper so probably is close enough.

[Kleinstein]

The better low TC shunts are usually made from Manganin (or similar alloys)- compared to Constantan the thermal EMF is smaller, resistance is lower (so longer wire and less localized heat) and it is easier to solder.
One has to see how much the shunt changes under load. One can still be lucky even with cheap resistor.
A 5.5 digit meter could make use of a relatively good quality shunt, better than the simple wire found in 3.5 digit handhelds. Still cheap shunt is better than no current range at all.

There is a thread on testing low value resistors / shunts, that has quite some information on how to test.

[tcX]

Hi, I am currently building the project and already have the PCB. Unfortunately the BOM is not very helpful as it only shows footprints but often no specific components and I have already spent hours searching for the correct components. The AD8436 is also hard to get in Germany. I need more precise specifications on the relays and reed relays, which ones did you use? I have never seen the footprint of R17. Can I just put a 0.1% 2512 resistor in there?

[Kleinstein]

The reed relays should not be that critical. They are used for current or AC. So the main point is getting the right footprint.

The Shunt R17 is difficult with the given layout. No need of 0.1 % , but ideally low TC and high power rating, which is not really working well for the given footprint.

The PCB is not made for a high voltage anyway: C21 and R25 would limit the safe voltage to some 100 V, depending on the resistor type. Ideally there would be some 4 resistors in series for R25, or at least a somewhat larger type. One may get 2 x MELF vertically mounted in place.
So the relay type is probably also not that critical - mainly the right footprint and ideally not too much heat.

The AD8436 is likely tricky nearly everywhere. Mouser may get some in the CP20 case in few weeks.

[snik]

Hi, I am currently building the project and already have the PCB. Unfortunately the BOM is not very helpful as it only shows footprints but often no specific components and I have already spent hours searching for the correct components. The AD8436 is also hard to get in Germany. I need more precise specifications on the relays and reed relays, which ones did you use? I have never seen the footprint of R17. Can I just put a 0.1% 2512 resistor in there?

Hello, i build one too at the moment. When you look at the 3 Pictures on Github you can see the relais Markings. The reed relais are marked with TRI-1A05 (DC5V) and the normal relais are cheap SRD-05VDC-SL-C. For R17 - It's marked in the boom as 2512 Resistor too.
The AD8436 is at the moment really hard to get. I buyed one in china via eBay. But as i also haven't all components yet, i don't know at the moment if it's a real chip or a Fake.

[qu1ck]

For bigger relays any 5v coil 250v rated relay will do. There are many manufacturers that make them in this form factor
SRD Series Form C, there are Sanyou, Songle and other branded. Search for SRD-05VDC-SL-C.

Smaller reed relays are 5v spst, search for SIL05-1A72-71L or SIP-1A05.

R17 is main current shunt I recommend getting a constantan wire coil 0.8mm diameter. Available on aliexpress for few bucks for a bunch.
Alternatively use 2x 200mOhm 2512 resistors in parallel to share the load. Not ideal for long term measurements of high currents.
And yeah as Kleinstein said, dont worry about precision for the shunt, it can be calibrated in software, low tempco is more important.

[Kleinstein]

For R17 a wire (so just a bare metal piece) shunt is definitely the prefered choice. The material is usually not constantan, but the more modern manganin, which is easier to solder and has lower thermal EMF.  Such shunts are also avialable from the normal distributors. If needed one can bend the wire in shape or combine 2 of them in series.

The plan has the relatively expensive TLE2072 for U1. In this place the cheaper and lower power TL032 and even a TL062 should work as well.

The diodes may not give a 100% protection for the 200 mA range, as R14/R15 may get quite hot. So one may consider to replace D3 with a short to lower the clamping voltage. This comes with a slight limit on the peak current / crest factor allowed for AC currents (especially smaller ones). One could than reduce the effect of diode leakage with a resistor from the output of U1A to the D3 position.

[qu1ck]

I don't see how 36 mW can get 2 2512 resistors hot.

Good point on U1, I just put the same value as U7 to minimize bom but if you have cheaper jelly bean fet input opamp it will do fine.

[tcX]

Sometimes life is a bit funny, there was no activity in this thread for 2 years and as soon as someone comes up with a problem, others show up as well - including the same problems.

I will be ordering most of the parts from Digikey and some of the more exotic stuff from Reichelt. AD8436 is really hard to get but 505-AD8436ARQZ-ND on Digikey (which is exactly what we need) is expected to be in stock again in about two months. I will order and wait.

[Kleinstein]

I don't see how 36 mW can get 2 2512 resistors hot.

The problem is not the normal 200 mA current. 36 mW is already the point where the self heating can become noticable (e.g. some 2-5 K temperature rise) and effect linearity, though still acceptable.

The problem is the overload case, e.g. when applying 2 or 3 A to the 200 mA range. With 3 diodes in series, the voltage may reach some 2 V, maybe even 2.5 V for a short time before the diode take significant current. 2 A for the resistors would than be 3.6 W and this can be too much for the 2 resistors.  With 2 resistors close by they would not get there full nominal power rating (~ 1 W) and would likely heat up enough to melt the solder.
The power rating of SMD resistors is valid for a near infinte copper plane, a bit like the power rating of transistors that is valid for 25 C case temperature. So a number that should be used with some correction factor for real life conditions.

[qu1ck]

The problem is the overload case, e.g. when applying 2 or 3 A to the 200 mA range. With 3 diodes in series, the voltage may reach some 2 V, maybe even 2.5 V for a short time before the diode take significant current. 2 A for the resistors would than be 3.6 W and this can be too much for the 2 resistors.  With 2 resistors close by they would not get there full nominal power rating (~ 1 W) and would likely heat up enough to melt the solder.

Fair enough. So I should use autoranging

[qu1ck]

Btw here is a close up of fancy custom high voltage trim caps on the original board

[Kleinstein]

The critical capacitor is C21, as this one has to withstand the peak input voltage. To reduce the change for damage, I would suggest a 20-100pF capacitor in series, that could withstand the voltage.

The capacitors at the range switching should should not be as critical.