This is interesting. Several times when I was doing the transistor blogs I wondered about about doing some kind of modular test setup [though it was fun designing the ad-hoc circuits and wiring the prototype boards each time]. I couldn't make up my mind whether separate boxes for each function would be best or some kind of mainframe with plug-in boards.

It shows how ignorant of test and measurement I am that I didn't even know that SMUs were a thing, though I had come across curve tracers and was vaguely conscious of automated testers with 'pin drivers' and all that. I obviously don't move in the right circles.

I was thinking of separate modules for driving current and voltage, though I can see how they might be combined [though, if you do, it kind of then starts to move towards what Michael is doing with his impedance analyser - I'd worry then that the spec is going to run away faster than you can chase after it]. I also toyed with the idea of using an isolated serial communication bus between modules [maybe something like a fast MIDI, or differential like DMX], to save the cost of working with ethernet [I'd have had some kind of a hub, with a small SBC, to relay stuff back to a laptop].

One complication for me, with the MOSFET tests, was the necessity for doing quick tests because of the effect of temperature from the dissipation. That then moves you away from DMM type techniques for the measurement to something that works sample and hold, with the timing and synchronisation issues that introduces [as far as I can see, SCPI won't be very good at synchronising and sequencing things at speed]. Although you might think that temperature effects would only be important for power devices, I was seeing such effects at much more modest levels with very small packages like SOT-23 and smaller.

Anyway, it will be interesting to see what you come up with.

One request: please don't put it all in one blog. That programmable load blog with 16 pages of comments is getting impossible to trawl through.

I was thinking of separate modules for driving current and voltage, though I can see how they might be combined.

I had some inspiration from the DAC8775 used on the Quad-Channel, Analog Output Module  RoadTest kit (I got DAB 's board, traded it with him for other electronics).

They use a separate current and voltage driver, then combine the outputs:

from https://www.ti.com/lit/ug/tiducv5/tiducv5.pdf?ts=1591530663544&ref_url=https://www.ti.com/tool/TIPD215 , page 2, 3 and 5

I don't think that TI are telling the whole story with their diagram of the current output of the DAC8775.

As they draw it (page 35 of the data sheet)

considering just the lower or sinking half:

Iout = -(DACvolts - VNEG_INx)/Rsense, so the sink current is a function of the difference between the negative supply and the DAC voltage.

And worse, at zero DAC volts both amps will be trying to turn their MOSFETS on enough  to drop the full rail voltage across their sense resistors.

You could fix this by using a diff amp to look at the voltage across Rsense.

If jc2048  is reading this is right up his street for simulation.

The problem is that with two diff amps this circuit needs 10 precision resistors to work.

Here's a very quick and messy LTSpice of it, with diff amps, can't make it work without.

The classic Howland current source (TI app note snoa474a) only needs 4 precision resistors.

MK

I don't think that TI are telling the whole story with their diagram of the current output of the DAC8775.

As they draw it (page 35 of the data sheet)

considering just the lower or sinking half:

Iout = -(DACvolts - VNEG_INx)/Rsense, so the sink current is a function of the difference between the negative supply and the DAC voltage.

And worse, at zero DAC volts both amps will be trying to turn their MOSFETS on enough  to drop the full rail voltage across their sense resistors.

You could fix this by using a diff amp to look at the voltage across Rsense.

If jc2048  is reading this is right up his street for simulation.

The problem is that with two diff amps this circuit needs 10 precision resistors to work.

Here's a very quick and messy LTSpice of it, with diff amps, can't make it work without.

The classic Howland current source (TI app note snoa474a) only needs 4 precision resistors.

MK

michaelkellett  wrote:

 

I don't think that TI are telling the whole story with their diagram of the current output of the DAC8775.

As they draw it (page 35 of the data sheet)

considering just the lower or sinking half:

Iout = -(DACvolts - VNEG_INx)/Rsense, so the sink current is a function of the difference between the negative supply and the DAC voltage.

 

 

 

...

MK

Yes, just shopping for ideas at the moment, browsing, reading and collecting. I'm fully aware that the drawings and text in the TI application note are a vast oversimplification.

I'm also looking at your circuit and the way you approach building for performance and precision.

And worse, at zero DAC volts both amps will be trying to turn their MOSFETS on enough  to drop the full rail voltage across their sense resistors.

They don't show it in the drawing, but in the operating narrative they explain that they never drive both outputs: "Only one of the current stages are active at any given time".