Part 3: Static/Dynamic Load Testing [Updated: 10 Jan 2021]

Gough Lui it's not a diversion at all. Your input was very helpful and spot on (as always   ) to get to the bottom of it, and make the most out of the roadtest. As you mentioned I did use different wires, when I first used the front panel I got slightly worse results than what is displayed. I then used thicker core wires (specs in the main review page) and used the rear panel, which helped get a more stable output.

I am still looking into the calibration issue, but factory calibration is good (the difference between DMM and readback is a few uV at no load, and around 40uV max at full load).

Thanks @andrewj - glad to hear that you appreciated my feedback. You can probably tell that this is a RoadTest I was interested in and it is an area I have some knowledge about, so I couldn't hold back - of course I don't mean to cause unnecessary diversions for Waleed, but I just wanted to try and explain as much of the observations, share some of the knowledge I've gained in prior RoadTests/hobby experiments and understand just what is causing the reported issues. Sometimes you do need a little bit of extra thought to make the most of your gear, especially if it is high-end and you're expecting very fine accuracy.

- Gough

The two posts in this thread you made Gough are really useful.  I would never have known this without reading.

Thanks for the update waleedelmughrabi.

I'm not surprised that a variation of 1mV didn't do much - looking at the datasheet, Ch1/2 is specified for a programming accuracy of 0.05% + 5mV and Ch3/4 is specified for a programming accuracy of 0.05% + 10mV. Of course, in reality, a good unit (especially new) should do better than that - but this means that the actual output should be +/- 20mV of the set value. Likewise, the on-screen readback is only guaranteed to the same level of error. As a result, changing the output by 1mV probably wouldn't do much and that is borne out by your experiments.

When you adjusted the output somewhat more, you were able to get some current balance happening which is good, but the spread of voltages became somewhat large. I wonder whether your cable is a bit thin for trying to carry all that current, especially if you're jumpering one-to-the-next so perhaps you were losing close to 100mV from channel to channel. I suspect you might have had some differences in contact resistance in connectors or at cable ends which also added to some differences. You were also probably not using four-wire remote voltage sensing which could be very helpful to compensate for voltage drops up to 1V per lead. In that case, you'd probably jumper each of the sense pairs to a common point - i.e. where your load is. The sense wires can be quite thin as they're only used to sense voltage and won't carry current. I suspect four wire mode + a few tens of mV each way could be enough to balance. Perhaps if you have an accurate DMM, you could see how accurate the voltage output is when it is set to 20.000V to infer whether the imbalance is supply calibration related or cable/connector related. Perhaps if the self-calibration setup worked, it may have helped this situation as well.

While internal parallel operation is definitely a nice feature to have as it would save a lot of mess with wiring and configuration, I can understand why it might not be an option. All R&S supplies I've met do not have this feature and part of it might be the fact that the connections on the front panel just can't handle it - especially when used with questionable third-party banana plugs which may not have adequate contact area. This could result in risk to the user and equipment (fire, melting plugs), so by not offering it internally, people must explicitly devise their own thick external bus-bar (best) or stacked banana arrangement of an adequate rating (yikes!) to achieve it. I know some other supplies I've met have a warning that the banana plugs are good for a certain amperage with the full output available on the rear connectors - but sometimes people ignore this and end up melting something ...

- Gough

I don't believe R&S intended for the channels to be paralleled where the outputs are not of the same rating, although it should be possible provided the limits of the lowest capability channel are respected (or some care is taken to set approripate limits).

However, in general, whenever you parallel the outputs of the supplies, you should not set the same voltage on each channel, as this often results in regulation oscillation. This arises because each channel's voltage/current measurement isn't going to be perfectly corresponding - so here you've set 19V on each channel, but those channels where the calibration is slightly different and maybe it is issuing 19.00001V will take more load than a channel which is only issuing 18.99999V. This results in an uneven current split and depending on the way you have cabled up the channels, dissimilar lengths of wire can also cause voltage drops which cause one channel to carry more load than another. This is why commonly for paralleled loads, the positive may be taken from the left side of the busbar, while the negative taken from the right side to equalise cable lengths for all channels. As the load changes, however, the channel that is higher may suddenly collapse faster than the others, resulting in the load jumping back and forth between channels rapidly and maybe even pushing one into the current limiter. This is all expected behaviour of any multi-rail power supply operated with outputs in parallel.

While ideally you may expect setting the same voltage on each power supply channel spreads the load evenly across channels, I've been taught never to do this as the reality of calibration differences means that it is very unlikely even in precision equipment. Instead, I've been taught to set a slight voltage offset between channels - say if you need 19V then maybe I have a channel at 19.005V, 19.004V, 19.002V, 19.000V which would force the load onto the channels in that order. Then if I know I'd need a total of about 18A, then I might set a current limit of 5A on each channel to provide a bit of headroom, but this means that the first channel will take load up to 5A and then its output would start to collapse, pulling the second channel into play and thus also "forcing" the current share by setting a particular order (assuming your voltage differences are bigger than your cable voltage drops). The downside may be that your output voltage is not going to be "exact" - but I dare say that in the oscillating mode of operation, the instability may cause more voltage noise and heat build-up within the power supply. Perhaps you could use the "internal resistance" feature of the power supply (if it has one) to provide this kind of negative feedback without changing the voltages on the front. If you don't want potential voltage regulation differences of setting offset voltages, then setting current limits may be enough and the output would settle between the voltage output by the channel with the highest output under no load and the lowest output under full load (below the sum of all current limits).

While I didn't see it mentioned in the NGP800-series manual, the R&S HMP4040 manual (a power supply which I reviewed), mentions this:

- Gough