DIY-SMU Project

[wergor]

I love and hate such threads... love because I learn a lot, hate because I see how far I still have to go
very interesting project. I'm designing an open source SMU myself (but I'm going in a different direction: +-24V, +-3A, no UI, instead it will be a module for a NI PXI / EEZ BB3 type system), your writeup, schematics and this thread have given me both insight and motivation to go ahead with it. thank you!

I was asking about improvements to my current sense circuit a while back, maybe this thread is interesting to you: https://www.eevblog.com/forum/projects/improving-a-current-sense-circuit/msg2986694/#msg2986694 - you may be able to eliminate U17 from your design (at the cost of 1 additional opamp).

[smgvbest]

Hi,

Love the project,   following you on youtube and this thread.   Great work.
sorry to be another one of those with suggestions though

I'm curious why not go for a Teensy 4.1 Board instead of the 3.2?
The 4.1 for almost same price gives more of everything but most importantly adds a SDCARD socket on the board which would really make you ready for datalogging storage and playback.  Also supports Ethernet so you could have USB/Ethernet SCPI support which would likely save you in the future adding those options.

I of course realize what is likely the reason,   you have 3.2 boards on hand

[djerickson]

You are right, I'll likely end up with a Teensy 4.x or maybe even an STM32. Currently a cohort is working on a version of the embedded code with Ethernet on a Teensy 4.x. I've designed HW for embedded systems with ethernet and RTOS, but never wrote the code myself. Also looking into changing the UI from Nextion  to LGVL.
Thanks,
Dave

[RikV]

Very tempting project! I would like to build one. Would it be possible to share the software and Nextion configuration of the current version? I do realize this is far from completed but nevertheless would be a nice starting point for som experimentation.
Thanks in advance.
PS: damn nice project!

[RikV]

Dave,
For what reasons would you leave Nextion diaplays?

[voltsandjolts]

Thanks to Dave for an interesting and well documented project.

For a lower voltage and simpler SMU, could a power op-amp be used as the output stage?
For example, OPA548 or OPA541, for say +/-30V and few amps SMU.
They have a programmable current limit which could be used as the 'back-stop' current limit but a more accurate current sense/limiter could be wrapped around it.

I'm thinking of an smu add-on for a standard lab bench supply that provides +/-32V @ 3A.

[jbb]

I was thinking about that. A lower voltage version could use reduced supply rails (ie <60V) which are safer to deal with.

But higher current outputs lead to high power dissipation, especially in ‘active load’ mode. Consider using the SMU to apple 2A load a 24V supply. On the face of it, that looks like 2*24=48W, which is substantial. However, that power opamp doesn’t have a current path to ground so the current must return to the -32V rail. So the dissipation is more like 2*(24+32)=112W. Yikes! I expect that a power opamp which could dissipate this energy would be expensive. That suggests a discrete output stage is the way to go.

[voltsandjolts]

Yikes indeed! Thanks for pointing that out, I hadn't thought of that.
A couple of power op-amps might be sufficient for a compact dual smu for small transistor characterisation.
But as you say, going discrete helps with heat dissipation, probably the only way to go.

[prasimix]

Ok, what's about adding pre-regulators?

[voltsandjolts]

While fine for a DC supply I suspect an smu pre-reg could be a bit trickier, if you need good dynamic response from your smu.

[Kleinstein]

The usual SMU circuits use a few supply rails to choose from and this way limit the loss. Switched mode preregulation gets tricky with EMI and also the need to sink and source - so one would need 2 rails. A preregular would still not help when used as a sink / load.  So one needs the relatively high power loss capability anyway.

The SMU power stages look that complicated because they often are made also for high voltage and use series connected transistors. With a more moderate voltage range (like +-45 V) the circuit can be quite a bit simpler, like 4  power transistors from levels like +-10 V and -+50 V.  As a sink it would add the opposite side 10 V and thus have quite some power loss. However just saving the 10 V by sinking towards 0 V would not help that much.

[jbb]

Yes, a pre-regulator gets a bit tricky. Conceptually, it could help a lot with the sourcing modes by limiting main amp voltage drop. You’d probably want one for positive side and one for negative side.

However, when we consider the sinking modes, we have two options.
1) if the pre-regulator goes down close to zero, we still need to dissipate all of the sink power in the linear amplifier
2) if the pre-regulator goes down below zero, the linear amplifier dissipation is reduced. BUT the pre-regulator then runs ‘backwards’ and will push energy back into the main DC link. We then need to get rid of this energy so the DC link doesn’t blow up.

So, as usual, Kleinstein is right: it looks like an adjustable preregulator isn’t particularly helpful here. A multi-tap supply (+-40V and +-10V) could deliver quite an improvement without the additional complexity of tracking regulation. All four power rails could be sourced from a single power stage (perhaps LLC for lower switching noise?).

Choosing the ‘outer’ voltage is based on the desired instrument voltage range (plus cable and shunt resistor drop, amplifier headroom and  supply tolerance).

Some spit-balling for a ‘touch safe’ SMU:
- say +-30V output
- amplifier headroom: 9V (note we need headroom for bias circuits)
- shunt R drop: 2V
- cable drop: 2x 1V
- Hence outer DC supply >= 43V
- Allow 10% tolerance: 43 / (100% - 10%) = 48V
- Funny how those numbers worked out :-)

Choosing the ‘inner’ voltage is... probably similar, really. If we just want to improve the power sinking situation, we could just use 0V as previously suggested in this thread.

If we want a ‘high current’ output from the inner voltage rail:
- say +-6V output (enough for ‘USB’ and LiIon testing)
- amplifier headroom: 3.5V (perhaps optimistic, can get bias from outer rails)
- shunt R drop: 2V
- cable drop: 2x 1V
- Hence outer DC supply >= 13.5V
- Allow 10% tolerance: 13.5 / (100% - 10%) = 15V (ish)

[jbb]

I realise that Keithleyhas already produced some nice diagrams which illustrate the point for their https://www.tek.com/datasheet/series-2400-sourcemeter-instruments]2400 series SMUs[/url] (attached).

If we look at the 2440 (bottom row), we see two distinct ranges in play: 10V 5A or 40V 0.5A. Also worth noting are the shaded notches in the sinking quadrants: looks like a power dissipation limit to me (or perhaps power transistor Safe Operating Area limits).

[dany-boy]

Hey Dave,
Thanks for sharing such a complex project! Specially thank you for explaining how it works and the analog design decisions. Reading your blogs along with other materials has been quite a learning experience. I'm finalizing my order for most of the important analog components, and I wanted to ask you directly if its ok for me to convert the boards to kicad and modify the design a little bit. Particularly in the digital isolation department and the ±15v isolated module. I want to change the recom module for the SN6505 controller that @prasimix suggested, and change the digital isolator for a couple of cheaper options, as well as changing the SOIC footprints for SSOP which are also cheaper. Of course these modified files would be public as well.
Cheers for such a great project!

[jbb]

... modify the design a little bit. Particularly in the digital isolation department and the ±15v isolated module...

I saw a new part recently that might be handy. The Texas Instruments https://www.ti.com/product/UCC25800-Q1]UCC25800[/url] is a little LLC power driver intended for isolated gate drive. It works with high leakage inductance transformers, which means they can be wound using split bobbins and have very low coupling.

Wurth electronics apparently makes the https://www.we-online.com/icref/en/texas-instruments/UCC25800-LLC-Resonant?sq=750319177&sp=https%3A%2F%2Fwww.we-online.com%2Fweb%2Fen%2Fwuerth_elektronik%2Fsearchpage.php%3Fsearch%3D750319177#750319177]750319177[/url]transformer which has <1pF coupling capacitance.

——

I had a look at the Keithley patent previously mentioned. I see the relationship to the crossover Dave came up with, and I can see it does indeed have lots of current limiting / constant current diodes.

Those parts are hard to get and/or expensive, so I’m now thinking about using diode bridge switches (more often used for RF switching) in their place. The current flow through a diode bridge switch is limited to the control current (I.e. current limiting is available) and the behaviour of the bridge can be changed by adjusting the drive currents (e.g. can open the switch by removing drive currents, or can only supply current to one side to get ‘diode’ behaviour). I can set a low current for the ‘force’ channel and a high current for the ‘clamp’ channels such that the clamps can override the force. Interesting.

It may also be practical to put in another pair of ‘oh no’ outer current clamps which are always active and ready to override voltage clamp limits in the Force Current Clamp Voltage mode (which Dave has had trouble with previously).

[voltsandjolts]

I saw a new part recently that might be handy. The Texas Instruments https://www.ti.com/product/UCC25800-Q1 is a little LLC power driver intended for isolated gate drive. It works with high leakage inductance transformers, which means they can be wound using split bobbins and have very low coupling.

Wurth electronics apparently makes the https://www.we-online.com/icref/en/texas-instruments/UCC25800-LLC-Resonant transformer which has <1pF coupling capacitance.

That's an interesting little chip and transformer, nice find.
Links were broken, I fixed them in the quote above.

[wergor]

DIY-SMU and the Keithley 236/237 use integrators as error amplifiers. why don't they use simple inverting amplifiers? how come the integrator's high gain at DC is not an issue?

[Kleinstein]

A control loop allways needs a limited bandwidth to keep it stable. The simplest case is a kind of integrator.
The usual OPs (dominant pole compensation) are actually working as an integrator with the capacitor inside, setting the GBW. The output of an OP is the integral of the difference between the 2 inputs. It only become a more normal amplifier with feedback, or at very low frequency (e.g. < 1 Hz), where the limited open loop gain gets visible.
Inside the control loop one usually wants a very high DC gain.

An explicite integrator allows to choose a lower speed independent from the OP. It may not be a pure integrator and could have additional tweaks to improve the stability.

[djerickson]

Hi All,
I've updated the DIY-SMU web page and added 2 new project videos on Youtube.
Chapter 3 is the new case, front and rear panels built by PCBway. I show how I used PCB layout tools and FreeCAD to design the enclosure.
Chapter 4 is the SCPI interface and how it was implemented.
Enjoy,
Dave Erickson

http://djerickson.com/diy_smu
https://www.youtube.com/channel/UC6agVsOGqt6E8cMNyzgdOAg

[ducreux]

Hi,
I don't have license for DipTrace : is it possible to have gerbers files ?
Thanks

[RikV]

Hi Dave,
Thanks for the update of the schematics and for including the software! This is an amazing project.
However, I believe you forgot to include the Nextion file.
I cannot get the soft compiled: DIY-SMU_SCPI\DIY-SMU_SCPI.ino:397:3: error: 'my_instrument' was not declared in this scope

Any suggestion?

There is however 1 remark I would like to make. There is no logic in the naming of your files: SMU_Main becomes SMU1 Cross...
It would be logical that the file name reflects the board  AND the version Number. SMU MainRev.2 10_2020 and SMURev2. which one is most recent at first sight?

Dave, just found out that U26 (74AHC1G14GV,125) is missing in the BOM

[RikV]

Hi, Ducreux,
to save Dave some time I include the Gerbers of the most recent version of the boards. I suppose Dave won't mind.
If you need a free "Gerber viewer", look for FlatCAM.

[ducreux92]

Hi,

I make a prototype with test plate, see photo.

I use only cheap op like TL074, without adc or dac.

in FV mode my two references are variables resistances.

I have a big oscillation problem in out: about 200 Khz with >5V pp

[ducreux92]

The photo.

[Kleinstein]

The TL074 may be a bit on the fast side on a bread board. Parts of the circuit may be to high in impedance for the added parasitic capacitance and there may be OPs with intentional different GBW in the design. The compensation may have to be adjusted for slower OPs.

So the simplified circuit / breadboard version may need adjustments / new calculation (simulation) for the compensation.