DIY-SMU Project

[djerickson]

Hi All,
I've written a few Python Curve Tracer programs for DIY-SMU, Single channel for 2-terminal devices, 2 SMU channels for transistors and tubes.
Here are the results for a 2N3904 and a 12AX7 tube. Check out the latest at http://www.djerickson.com/diy_smu/smu-bringup3.html
I'm pretty happy with the results.
Thanks,
Dave

[desert]

Very exciting project and thanks for sharing! 👍

One important feature of SMU and V/I instrument is glitch-free range-switching: no voltage drop or voltage glitch during up-range-switching, no voltage overshot or voltage glitch during down-range-switching. This is valid for both current range-switching and voltage range-switching. And it takes very short time to finalize rang-switching (less than 100us).
It needs state-of-art design to achieve that.

Have you tested that behavior? It's interesting to see the result.

[djerickson]

Hi. Thanks for your interest and excellent question. The low current ranges use fast CMOS switches with 100-200nS switch times. There is no significant glitching when switching between low ranges. The 100mA range uses a slower Photomos switch and does have a so has a 200uS big-old glitch when changing to that range. I need to change the code to turn on that switch >200uS before turning off the previous range switch: Make-before-break is bad. That should fix the glitch.

Range switch glitching is one reason I didn't use relays for range switching. Photomos delay is better and cleaner than a relay, but still needs care. Next time I'm in the firmware, I will fix it. It is added to my to-do list.

Voltage range switches also uses Photomos switches, but there is always a parallel resistor for the other range, so no break-before-make glitching there.

Again, thanks!
Dave

Very exciting project and thanks for sharing! 👍

One important feature of SMU and V/I instrument is glitch-free range-switching: no voltage drop or voltage glitch during up-range-switching, no voltage overshot or voltage glitch during down-range-switching. This is valid for both current range-switching and voltage range-switching. And it takes very short time to finalize rang-switching (less than 100us).
It needs state-of-art design to achieve that.

Have you tested that behavior? It's interesting to see the result.

t and excellent

[ddrl46]

Are the files which are available on your website (https://www.djerickson.com/diy_smu/files/DIY_SMU_Files.zip) still the latest revisions? I may be interested in building one or perhaps two units.

[MathWizard]

Cool project, someday I hope to try it or some version.

What are the higher voltage ranges used for in general ? Curve tracing ? I guess I mainly do lower voltage work. When I learn a few more things, I'd like to make a very low power, low voltage, 4-quadrant supply, even just on a breadboard, with BJT's and maybe op-amps.

[Hydron]

High source voltages (including well above those seen here) are very useful for looking at semiconductor leakage etc when you have good low current measurement capability. I regularly use my keithley 237 above 100 volts (usually on the microamp ranges unless I'm looking at insulation resistance).

[gmac34]

Hi, has anyone ordered PCBs, if so I would be interested in getting a set. thanks!

[chilternview]

Impressive design Dave, this has piqued my interest.

What would be involved in lower current ranges? I'd be interested in getting down to the 1fA or so of the HP4140. That uses reed switches for range switching, and also a zero offset capability for correcting repeatable leakage below 100fA.

[Kleinstein]

The very low current range may want a few more changes. Especially if getting in the low current constant current one may need a lower speed. Chances are one would build a seprate version for only small currents (e.g. up to 1 µA).  Having the instrument for a large range of currents comes with compromises and at the extremes there is not much room for compromises.

[SebastianH]

Hey guys,

it's been a while and now I have some need for advise/feedback. My schematics are very similar to the original ones, with some changes discussed a few pages earlier. The board currently looks like this (placement will certainly change a bit here and there; obviously not routed yet):



Is there anything placement-wise that is a big no-no? My goal is not to build the "perfect" SMU, but I try to avoid the bigger problems here .

It's going to be a 4 layer board, since I don't care about a few bucks for ease of routing alone. But now, I would also be thankful for advice regarding grounding (ground plane, star grounding etc.). What about the signals that are referenced to the floating GND? Where on the board should I connect the "reference GND"/Amplifier output? Yeah, that's when I notice that I'm still a beginner

Thanks
Sebastian

[SebastianH]

I'll try to make my question more specific, here is the voltage ranging/measurement circuit, the buffer/ADC section and the PCB section in question. The difference amplifier has a GNDF reference, so do the ADC buffers and the ADC. I could just connect all GNDF's to the ground plane, risking some noise to be picked up by the ADC from the (low current) analog circuits. I could also connect the GNDF's via a trace in an attempt to use star grounding (example in the PCB picture; my guess would be a 5 to 10 times higher DC resistance than with a ground plane). Does any of the solutions make sense? Are both horribly bad? Would really appreciate any advise.

[Kleinstein]

For the GNDF a star ground may make more sense than a ground plane.  There should be not much higher frequency signal going via this ground.

[djerickson]

Hi SebastianH,

I like what you've done. I see that you reduced the VM gain at the diff amp to allow 10% headroom, and the ADC is now +/- 5V, allowing the use of convenient 10K resistors. Same with the DAC buffers.  Good idea. Did you do something similar on IM?

4 layers allows you to shrink the board and improve grounding. I get a small amount of SPI clock noise on the output. 4L should help that too. Makes the layout easier. Nice.

Did you guard the current sense and + output? If you send me your files I'm happy to review. And borrow your other good ideas!

Thanks,
Dave Erickson

[SebastianH]

Hi Dave,

great to hear from you! I took a lot of inspiration from Jaromir's schematics and if I remember correctly he had a lot of 10k (single) resistors in his schematic, so I don't want to take credit for this approach necessarily. I probably have changed a few things a bit and found very convenient to use Vishay's 10k ACAS precision resistor networks wherever possible. The IM circuit is full of 10k's as well, even though I'm looking for a gain of 2 => 20k at one point, but that is what networks are for, I guess. I attached a PDF of the current version of my schematics.

Right now I still try to figure out how to do the grounding. I don't fully understand how I would implement a star ground approach (Kleinstein hinted at, probably rightly so) properly in a two-converter design. This is also where AD' application notes state that for multi-converter designs star ground is not possible and in a strict sense it simply isn't. They also admit that a solid ground plane may not always a good idea, but fail to point out concrete solutions. A few examples of well working and poorly performing examples with explanations would go a long way. So if you like to share some advice on how I could proceed - I'd certainly appreciate it

Haven't thought about any guard rings etc. in great detail, although I briefly discussed it with Jaromir. Besides the grounding this is the other part that I still have questions about in this more complex circuit. As a power systems EE, these topics are a bit outside my area of expertise/experience, unfortunately 

- Sebastian

[SebastianH]

Regarding the guard rings. I drew some guard rings in the schematic, (symbol G shows the node the guard ring is connected to). This probably is very similar to Jaromir's implementation (I think). The problem is that the potential of the nodes of the right side of U302 and U303 will change when switchting into different ranges. With this solution the guard ring potential matches the guarded trace except for the range switch being activated. For the activated range there is a voltage drop across the sense resistor - obviously. Guarding both ends of the resistor with the same guard ring seems weird to me. Maybe it's better to only guard all the sense resistors in the same way as R9 (and C301) and not guard the switched nodes at all?`

Edit: I don't have diptrace, so I'm not sure how you did it (@djerickson). Do you have the gerber files of your project posted somewhere?
Edit 2: A slot between the pins of the switch seem to be more suitable. This might not be an option with many prototyping PCB manufacturers though, since the slot width would have to be 0.8mm (or so) with virtually no space between the pad and the edge of the PCB.

BR
Sebastian

[Moriambar]

This is awesome, I want one!

[RikV]

Build one (or two)!!

[RikV]

It is so quiet over here. No evolution?

[Martinn]

Definitely on my todo list.
But while Dave loves SMT soldering, I prefer to order soldered PCBAs from JLC.
Not sure if any PCB is already ported to KiCad - I don't have DipTrace either.

[luudee]

I really would like one as well !!!

Should we organize a Group Build ?


Cheers,
rudi

[Roehrenonkel]

Hi All,
 
since a second K237 came my way it's unlikely that i finish this project.
Even if i have all the parts (SMD and THT) except the PCB.
If someone needs them.........

Best regards

[Hawaka]

Hi All,
 
since a second K237 came my way it's unlikely that i finish this project.
Even if i have all the parts (SMD and THT) except the PCB.
If someone needs them.........

Best regards

PM sent regarding the parts

That project is also on my todo. If I ever order PCB's I'll post it here to see if there is interest for some.

Only thing I wish is that it would go to 450V… Guess I know what is my next project   

[saliherensagirli]

Hello, I am trying to build an amplifier for a project and stumbled upon this project. While trying to understand the circuit I am baffled at U14.2 opamp and its complementary passive components located in Mainboard schematics (I am guessing it is some sort of a voltage buffer). I would appreciate if someone could help me understand how they affect the rest of the circuit and how the passive components values calculated.

Sincerely,
Salih Eren Sagirli

[RikV]

Hawaka,
Since that K237 is the only thing keeping you from building this SMU, I am willing to help you out!
Ship that K237 my way an set yourself free! Of course, you pay for the shipment.

[RikV]

Dave has an earlier project where he details the functioning of these 3 amplifiers (13.2, 13.3, 14.2). They are part of the feedback loop of the amplifier. 13.2/13.3 correct the drive voltage for the positive / negative clamp of the amp, 14.2 is the "error" amplifier. It is easier to understand the circuit if you omit 13.2 and 13.3.

Here http://www.djerickson.com/ps-load/index.html Dave explains the basic operation of the circuit.