CC/CV digitally controlled power supply

[KeepItSimpleStupid]

This pic **broken link removed** illustrates straps and switches.

There is a movable bar between the green and black posts which can connect ground to earth. In general, what I used to see on Sorenson supplies is a terminal strip on the back with metal jumpers that configures various options such as series/parallel operation, remote sense and it had the inputs for remote sense.

An example: https://www.newark.com/molex/38002-...DPbNgWkl|pcrid|41566056141|plid|&CMP=KNC-GPLA

See the definition under electronics usage: https://en.wikipedia.org/wiki/Strap_(disambiguation)

 

[throbscottle]

Ohhhhh... So it is the kind of straps I thought - I just couldn't work out the context...

 

[Gobbledok]

Nice KISS thanks for the info. Those power supplies are very interesting. For the sense leads for my purpose I'm thinking I might run both the positive and negative sense leads into a differential amplifier, and the output of the differential amplifier will be used as negative feedback into IC3. I'm not sure if I will run into problems using so many differential amplifiers though. From the reading I have done, the resistors need to be extremely closely match in order to remove all of the offset voltage. I'm thinking I may be better using differential amplifiers with internal laser trimmed resistors such as the AD8271 or the LT1168 but they are expensive.

I have been slowly getting the hang of LTSpice. Great program for simple sims but I find it is extremely slow for a complex circuit. I'm talking a few hours before I get impatient and stop the sim. My computer isn't a slouch either, an i7 2600k with 8GB RAM. I'm not sure if it is just because of the millions of calculations it is doing, or if it is a problem with my circuit.

I have attached a screenshot of the circuit and the asc file if anybody wants to take a look. The op-amps and mosfets are substitutes and can be replaced if it will speed the sim up.

 

[throbscottle]

If it is taking such a long time to run the sim, it is most likely to be oscillating at high frequency, or calculating extreme conditions due to something you missed - I occasionally get Kv's and 100's of A, because I missed something. Try running the sim with a very short simulation time, you might be able to pick up the first few cycles that way if it is oscillating. If it's running for more than maybe 30s, it's usually a sign something is wrong, unless the circuit is complex (which yours isn't).

Also, simulators aren't perfect. Just occasionally (and it is very rare) you do get completely bonkers results. But again, the suggestion if you do is to figure out why - it more than likely due to poor design.

Go over everything extra carefully, I can see Q1 and Q2 forming a short circuit across the 25v supply if the input is not either fully high or low. U1 is going to have a serious problem unless it's one of LT's "C load" devices due to the 1uF capacitor across the output. Those are just the immediately obvious things. You should probably give your voltage sources some internal resistance, also.

I would try breaking the circuit down into sections and simulate each of those, before running the whole thing at once. Also, since it is a DC circuit, try decoupling all the amp inputs - if it is oscillating, this should reduce it enough to actually get some result you can see.

Other thought is, try a different simulator. LT Spice is generally the most accurate, but I have occasionally found Tina to be more useful (certainly easier to use).

Hope this helps

 

[Gobbledok]

Thanks mate very helpful. It's good to know that a few hours isn't normal

I'll try your suggestions and see how I go.

 

[KeepItSimpleStupid]

I forget how I implemented "sense" leads, but if you kinda make it SMU like, you could possibly make the device (measure V) or (source V, measure I) probably without much effort if your willing to live with poor resolution.

I did use an AD part which was a fairly expensive differential amp where the inputs were protected to 30 V and I also ran into an almost show stopper. I did not use a low value resistor between the output of that amp and the input of a regular OP amp, so the bias current wasn't dropping across nearly 0 ohms.

I supplied the bias externally because I didn't want the Ibias-Irf transfer function, so I think the power supply output was applied to one of the inputs and the measured output to the other in one case and in the other, I MAY have made it unity gain, so yea, it worked. I had auto-ranging system meters for the two outputs: Voltage and Current, but they were in effect A/D's because the final result and control was with LabView running on a PC.

minor edit: Parenthesis and commas

 

[Gobbledok]

Thanks again KISS for your thoughts.

Funny you mention the AD differential op amps, yesterday I ordered some AD8421 from farnell. I also ordered some AD822's because apparently the ones I had here were fakes from ebay (who would've guessed?!). Last but not least I also have some 1% 0R1 sense resistors coming.

I think I will have much better success with proper instrumentation amplifiers rather than trying to calibrate about 8 pots for the current sense differential amps

 

[KeepItSimpleStupid]

For the current sense amps use just that, a high side current sense amp.

 

[KeepItSimpleStupid]

I used the AD624C for my application, I think. https://www.analog.com/en/specialty...tation-amplifiers/ad624/products/product.html

See, https://www.linear.com/product/LTC6101 A high side current sense amplifier.

Depending on your resolution, you can look at the allegro hall effect current sensors too. e.g. https://www.pololu.com/category/118/current-sensors

Here https://www.google.com/url?sa=t&rct...fRmtokNXiZxMuzQ&bvm=bv.74115972,d.eXY&cad=rja

is an interesting chip you should look at.

 

[Mosaic]

Slighty OT, but the TFT board. Do you know of PIC based versions with an on board uC for the same kinda cost?

 

[Gobbledok]

... A high side current sense amplifier....

Thanks KISS when you mentioned that earlier, I though you were talking about the method of doing it, I didn't realise such an animal existed.

As I have already ordered my AD8421's I will give them a go first and see how they perform. AFAICT I should be able to get the MOSFETs sharing current to within a couple of percent (or less). I do like the idea of only having one pot for gain per sense resistor instead of 2 (my original idea). I had read somewhere that if there is a 1% error between the inverting and non-inverting resistor ratio's, then that will let 1% of the common mode signal through, which is way too much.

I do like those ADM1293's though. Nice and handy and, once again, I didn't know such a beast existed.

Hi Mosaic I replied in your other thread.

 

[throbscottle]

Article in LT journal of relevance - design around one of their devices that will make a cc/cv psu - will scan it for you later and post as pdf.

 

[Gobbledok]

Cheers throbscottle that'd be great. I take it that it is linear?

 

[throbscottle]

Will PM you - file to big to post.