Adding remote sense (4 wire) function to a generic PSU with CC/CV output

[shapirus]

Sometimes the standard method of sensing voltage on the output terminals of the PSU is not good enough: one might want to keep a certain voltage across the load and have the PSU adjust its output voltage automatically so that the difference compensates for the voltage drop on the connecting wires.

A typical example would be charging a single battery cell (or multiple cells connected in parallel) to a certain (e.g. float) voltage at a large current, where the voltage drop on the wires becomes quite a nuisance, as the PSU reaches its set voltage much earlier than the desired voltage is reached on the load (cells), and current starts dropping earlier than it would if the voltage was sensed across the load, wasting time.

How do we solve it unless we've already got a fancy PSU with an additional pair of wires for remote sensing?

A quick and easy way is to build a differential amplifier to sense voltage across the load, convert it into single-ended output, cut the original feedback track where it was connected to the positive output terminal and connect the amplifier's output there. This allows to keep all the original circuitry intact.

Here's my implementation that works fine in simulation and seems to work as expected in a TL494-based SMPS like one of these: https://www.aliexpress.com/item/4000055526485.html
Mine has an adjustment range of 0V..15V, used, usually, up to ~14.4V.




Main potential problem to be aware of: if output goes too low uncontrollably, the negative feedback won't work properly, and the PSU will keep raising the output voltage up to potentially blowing things up. This has to be dealt with.

Points of interest:

- Resistors R3 and R8 provide signal to the sense lines to protect against them being unconnected. Values shouldn't be too low to avoid too much current going through them when the voltage on the load is too low compared to the power terminals or if the sense terminals are shorted. How high we can go with their values here, I don't really know. But I checked some existing remote sense schematics, and they all seem to use quite low value resistors.

- Resistors R5, R6, R9, R11 should preferably be of high precision. I used 120k 0.1% ones, because I have them. But if I understand correctly, nothing too bad will happen if they aren't perfectly matched: the voltage adjustment potentiometer's action will become less linear and that's it.

- Op amp selection: OPA2192 is a pretty cool inexpensive RRIO op amp capable of a wide supply voltage range, so I wanted to try it. It allows to set output voltage down to almost zero and up to almost its positive supply. The venerable LM358 can be used instead just as well, except that it won't allow to properly set output voltage in the lower range and will require a larger "headroom" between its max desired output voltage and the positive supply.

- Power supply: it's required to find a source of low-voltage supply in the PSU, which typically comes from an auxiliary transformer or an auxiliary winding. That's the one that's used to provide power for the PWM controller and other things. In my case its voltage was too low, about 14 volts (and not quite stable), so I had to add a boost converter (based on MT3608) followed by a linear regulator to get a stable +15V supply.

- Safeguards: Q1 protects against failures of the op amp: output stuck low, output floating. In these cases, Q1 conducts and passes the positive power terminal's voltage on to the output, thereby keeping the feedback network working reasonably. If the op amp's outputs fail high or Q1 fails open, then D3 will do the same job as the last resort, albeit passing one diode drop lower voltage than the source.

- Side effect of the protections used: the amount of voltage compensation will not exceed approximately one diode drop (and there will be no compensation at all when Q1 is fully conducting, which can happen if the voltage at the sense terminals is too low).

- R7, as pictured, might be too low for many op amps: make sure it's high enough so as not to overload the op amp when it has to sink the respective current when the op amp's output is low.

- This will work as shown without modifications (save for the power supply levels) for output voltages up to about 30 V or whatever the op amp is capable of. Higher than that, an additional voltage amplifier output stage will be required. I haven't tried to model this.

- D4 protects the output of the op amp when the device is unpowered and voltage is applied to the output (e.g. in a battery charging scenario).

- C5 was required to make the op amp happy in simulation, as it tended to oscillate on fast transients without it. Probably not required in the real world, but I still soldered it in.

- The sense wires are connected using a shielded multi-conductor cable. The shield is connected to the mains earth terminal. Not sure if shielding is really required here, but, considering the relatively high input impedance of the amplifier, it's probably good to have.

- An additional overvoltage protection may be desired; for TL494-based power supplies a possible approach is described in the app note on page 21: https://www.ti.com/lit/an/slva001e/slva001e.pdf; for others, we could use some comparators and operate the shutdown pin or inject bias into the feedback line.

- I decided to reconnect the original PSU's voltmeter's sense terminal from the power output to the feedback line so that it displays voltage across the load. But then I don't know the output voltage. A second voltmeter would be nice to have to see both voltages.


Actual board:



...of course I fixed the wires with hot glue and isolated the whole thing with several layers of heat shrink tubing before final assembly.


What did I miss in the schematics? Any obvious fails? Any room for improvement? It works as it is, but maybe I should fix/improve it before modifying another PSU.