Programmable Electronic Load - Analyse the Summing Node Zero Point

I'm going to redo this exercise now that the OpAmps are replaced with lower bias ones and I have more precise measurement tools.

From the blog Programmable Electronic Load - LabVIEW Test Automation: Characterise the Instrument we can see that the behaviour is very linear.

But we have a non-0 start point. When th DAC is driven to 0, there is an output current flowing of a few mA.

I assume that it's our control circuit that's doing that, because if I disable the output (pull the gate to ground), the current that flows is neglectable. And it's the same FET that's fully blocking the current flow then.

I've re-stacked the load so that the control board is on top and I have easier access to the measure point.

I can't use the LCD display in that case. I'll use the SCPI interface to set values.

I'll measure DAC output and the voltages at relevant control circuit points to see what drives the FET to conduct a little.

I think I'll document the results in new blog post. We can then think on measures to get the lowest current lower than 2 mA.

I've been checking for a while and get 0.6 µA when measuring with a multimeter. I usd a fairly high input voltage of 14 V to compensate for any cable resistance.

I have this working. I'm logging pin P7.A via the scope ...

jc2048  wrote:

 

Can you measure the voltage at P7.1 so we can see what the integrator thinks it is telling the MOSFET to do as you go past the cut-off in the current.

 

...

No offset remediation, 5 V input:

Current:

Gate voltage:

I've done the same measurement a second time. Even though the current graph is stable, the gate graph shows significant difference:

I'm doing the gate measurements again. I've just learned how to program the Bandwidth filter of my scope.

Well, it didn't smoothen that out

The input voltage this time was 6 volt.

Here is the same exercise with Scope bandwidth on and input voltage = 5 V.

The MOSFET behaviour is very interesting and I'd like to investigate that. I don't know much about how these power MOSFETs behave if you try and use them in the linear region like this - the manufacturer intends them as power switches, so the datasheet doesn't help very much - but if I had to guess, I'd say that maybe you're seeing the effect of having an array of MOSFETs in parallel, all with slightly different characteristics, and characteristics that change with temperature and so are dependent on how the currents arrange themselves (a power MOSFET like this will be an array of smaller MOSFETs). It's possible that down around the threshold voltage there's a certain amount of somewhat chaotic behaviour to the gate voltage (ie there might be several arrangements that allow for the small total current and which you get may depend on very subtle variations as you approach - if that were the case, it ought to be possible to repeat the measurement many times and map the various 'paths' through that area).

Anyway, that's just wild speculation [quite out of character for me!] and might be totally wrong - where are all the experts when you need them? It doesn't matter to you because the loop seems to be dealing with it quite happily.

Back to the load. I've just sat thinking about all this and it seems to me that you're asking a lot of this circuit. The sense resistor is 50mOhms, so for 1mA current you get 50uV on the sense. You then give it to an op amp with an offset voltage of 180uV (typ)/ 600uV(max) to deal with. I suspect we're getting to the point where you would need to think about trimming the offset on that first amplifier.

But do you actually need the load to be able to do this? Do you need the load to operate continuously over more than 4 decades of current? If you don't, perhaps you might do better to build two instruments covering different ranges.

jc2048  wrote:

 

...

It doesn't matter to you because the loop seems to be dealing with it quite happily.

 

...

Yes, it's impressive that the control mechanism gets such a good grip on the current with the MOSFET being a tough device to control in this range of operation.

...

 

But do you actually need the load to be able to do this? Do you need the load to operate continuously over more than 4 decades of current?

....

No. My main goal was to see if I could turn Peter's design into a programmable one. I started from that point, with no other goals.

It's certainly not a perfect exercise.  I learned and re-learned a lot while doing this (I'm expected to be a diploma'd control mechanism guy but I needed loads of studying to get hold of the theory and practice again - this exercise bumped me back into the stratosphere of control theory ). And I'm pleased that it yielded those blog posts you made that explain the behaviour of the control part.

I think this instrument as it stands now is useful in its operating range. Mine is fairly stable and linear between 2 mA and 6 A for the control part. The measurement part I have only roughly tamed. It needs more work before we can see how well that behaves.

Maybe, because everything is published, there is a chance that someone improves on it. Maybe by increasing range and precision. Or by reducing the costs by factoring out some of the ridiculously expensive components I used ... (and improving the PCB design)

For a comparatively simple circuit, it's been a very good vehicle for blogging. I'm learning a lot from it, though I think it will be a while before you give me my diploma in servo systems (I never was any good at maths, and I'm certainly not now).

One worry I have about those blogs you refer to is that I'm not an expert in this. All the simple project stuff [I'm thinking about Arduino stuff here] doesn't matter too much (it's just fluff, though it's fun and serves a useful purpose in getting people interested and acts as a way of introducing concepts), but those blogs were moving into the area of theory, but theory that I'm not qualified to teach. I think, on balance, it's worth doing this stuff (as much because it gives an idea of how electronic design is done and how you think about things and solve problems as the actual detail of what you're blogging about), but I often think that there are lots of people out there who could do it so much better than I could.

To extend the range down you just need a higher value sense resistor (though you obviously lose the top end). A 5R sense resistor should get you down a couple more decades. You might need to swap the MOSFET (though it would be an interesting experiment to try the 5R in the existing circuit and see what happens with the MOSFET if you take the drain current down further). That then leads to the idea you could switch the load resistors into circuit, but that falls on the difficulty that anything you use as a switch has a significant resistance compared to the 50mOhm. Another possibility would be the two resistors in series, with a relay to short out the 5R for higher currents, but you'd need two sense circuits and some way to switch between them.

A further idea, if the current when forced off is very low, is two parallel two circuits with either being allowed to operate - a high current one and a low current one. You might even be able to arrange for both to be operating together, though I suspect the loop stuff gets a bit hairy if you do that. [One thing to watch, though, is that the clever scheme of shifting the ground point to the source of the MOSFET rather than the negative load connection would compromise that.]

I've got a much, much lower cost design using transistors (a completely op-amp free zone!) - I did that a while back as a challenge to myself to see if I could design something more complex than a textbook circuit. I haven't prototyped it yet but it works ok in the simulator. (I'm a bit slow in my old age - it was only after I'd started it that I realised I was actually designing a discrete op amp, though it does differ a bit from the traditional do-anything op-amp design). It's nothing like the precision of this one and it's not a true servo system, but it's much faster. Component cost would be a few pounds. I'm saving it for the next DIY Test Equipment contest.

Here's a long running sample. 5400+ samples. Ignore the blip at 2400 because that's when I paused the test to go out with my youngest daughter. I don't want to burn my house.

It's still running. I'll stop it when I close for the night (bloody early flight, need to be up at 4:00 ).

Jan Cumps  wrote:

 

Pushing the offset higher doesn't solve the issue. It just pushes it uphill.

In this graph, I set DAC B, that creates the offset correction, to 300.

 

Behaviour is the same, a steep jump from nothing to > 2 mA, but at a higher starting point of DAC A.

 

 

1840,00051850,00251860,00251870,00261880,00281890,00291900,003

 

(are we fighting against the diodes at the backside of the integrator?)

The conclusion may be premature, but the cause for the bump seems to be the offset in the current sense opamp.

When working that offset in the U3C away to under 200 µV, I get nicer behaviour when the regulation circuit starts.

Jan Cumps  wrote:

 

Pushing the offset higher doesn't solve the issue. It just pushes it uphill.

In this graph, I set DAC B, that creates the offset correction, to 300.

 

Behaviour is the same, a steep jump from nothing to > 2 mA, but at a higher starting point of DAC A.

 

 

1840,00051850,00251860,00251870,00261880,00281890,00291900,003

 

(are we fighting against the diodes at the backside of the integrator?)

The conclusion may be premature, but the cause for the bump seems to be the offset in the current sense opamp.

When working that offset in the U3C away to under 200 µV, I get nicer behaviour when the regulation circuit starts.