Inside the new Korad KA3305P linear PSU

[torch]

Not quite what you asked for, more like "a punishing pulse test", but instead of using the on/off button, here I left the PSU on, made contact with then dragged a wire across a coarse file to repeatedly break the circuit. There's an interesting initial brief voltage spike before it ramps up as the bulbs start glowing. Because the load is inductive? Because the resistance increases dramatically as the bulb starts to heat up?

[Salas]

It can't recover so fast to hold a resemblance of a DC line on that. Using the file on a PSU rings a bell but I can't recall exactly. Was it some guru like Bob Pease or Jim Williams who did it first? Good it did not break something. At least now you know its tough. Money well spent.
Hmm... How a classic linear chip like 78XX, 317 etc. would react when dragged along that file for a load it could bear? That could give some kind of scaled down screenshot comparison.

[dom0]

How a classic linear chip like 78XX, 317 etc. would react when dragged along that file for a load it could bear? That could give some kind of scaled down screenshot comparison.

Not much, as stability is mostly an issue with "discrete" (i.e. op amp based) PSUs, while the integrated emitter-output regulators are very stable (they don't have a lot of loop gain to begin with and their fT is in the region of 500-1000 kHz). So output waveform will mainly depend on the output capacitor.

[george graves]

Can't say I'm a fan of that user interface.  Too bad they didn't make more room for it.

[torch]

Using the file on a PSU rings a bell but I can't recall exactly. Was it some guru like Bob Pease or Jim Williams who did it first?

I can't say I know of the file test being applied to a PSU. However, it's certainly not original to me: it is an old trick for down-and-dirty testing of automotive ignition coils. Scrape the primary wire along a coarse file and look for a series of sparks jumping a gap tester attached to the secondary.

[Salas]

How a classic linear chip like 78XX, 317 etc. would react when dragged along that file for a load it could bear? That could give some kind of scaled down screenshot comparison.

Not much, as stability is mostly an issue with "discrete" (i.e. op amp based) PSUs, while the integrated emitter-output regulators are very stable (they don't have a lot of loop gain to begin with and their fT is in the region of 500-1000 kHz). So output waveform will mainly depend on the output capacitor.

I would expect them classic chips steadier too. Maybe torch has one around to drag on the rough file. Is such a bench supply relying just on op-amps or an MCU controls the loop more likely?

[Salas]

Using the file on a PSU rings a bell but I can't recall exactly. Was it some guru like Bob Pease or Jim Williams who did it first?

I can't say I know of the file test being applied to a PSU. However, it's certainly not original to me: it is an old trick for down-and-dirty testing of automotive ignition coils. Scrape the primary wire along a coarse file and look for a series of sparks jumping a gap tester attached to the secondary.

So that testing has roots in automotive practice. Interesting. I must have read something in the past about it for an electronics PSU  torture test also. It could have migrated by some car tampering electronics engineer.

[torch]

It could have migrated by some car tampering electronics engineer.

Would have to be an old engineer -- someone who was around in the days of points and condensers...

[dom0]

How a classic linear chip like 78XX, 317 etc. would react when dragged along that file for a load it could bear? That could give some kind of scaled down screenshot comparison.

Not much, as stability is mostly an issue with "discrete" (i.e. op amp based) PSUs, while the integrated emitter-output regulators are very stable (they don't have a lot of loop gain to begin with and their fT is in the region of 500-1000 kHz). So output waveform will mainly depend on the output capacitor.

I would expect them classic chips steadier too. Maybe torch has one around to drag on the rough file. Is such a bench supply relying just on op-amps or an MCU controls the loop more likely?

MCU control loop would be insanely slow and unreliable. There are some very high power / special purpose SMPS which use DSPs for the control loop, though.

[Kleinstein]

The linear regultor is done analog, the µC just gives the setpoint and might give a controled ramp after power on. A µC might be used to controll the relays and the fan.

The reaction to this very rough test looks OK, but not very good - there is one spike going noticable higher than normal voltage. The funny spike at the beginning seems to be part of the normal way of switching from CC to CV.  So having short pulse of high current might be a problem for this regulator. This might even leed to overheating of the transistors, as the relays are not very effective with such dynamik loads.

For stability testing there are usually three critical cases:
1) having a low loss capacitor in the 10 µF range at the output and essentially no load. Internal Load might vanish at vry low voltage, so something like 50 mV might be the most critical volatge. This might give istability at high frequency (e.g. 100 kHz), especially if the output cap is small.
2) same as befor, but at a high current from an more or less ideal current sink. This might be rather difficult to build.

3) having a large low (as low as possible) ESR cap of some 1-10 mF and large load steps, especially down from a high current (but not limiting) to essentially 0 current. Here a manualy opening contact migh be enough, as reaction is likely rather slow in the ms range.

[torch]

The reaction to this very rough test looks OK, but not very good - there is one spike going noticable higher than normal voltage.

Yes, the worst spike appears to be about 5v (8%), bearing in mind the low resolution of the screen on my hobbyist-level Rigol scope. There's also two lesser spikes, apparently about 2-2.5v.

Frankly, I suspect there were more that just weren't captured given the time base.

The funny spike at the beginning seems to be part of the normal way of switching from CC to CV. 

That makes sense. I didn't think of that.

Maybe over the winter I can try your three stability tests. At the moment I don't have an ideal load nor do I have any really large capacitors sitting around. Out of curiosity, how much more stable would one expect a professional grade PSU to be, and how important is it to a hobbyist or student on a limited budget?

[Salas]

The linear regultor is done analog, the µC just gives the setpoint and might give a controled ramp after power on. A µC might be used to controll the relays and the fan.

The reaction to this very rough test looks OK, but not very good - there is one spike going noticable higher than normal voltage. The funny spike at the beginning seems to be part of the normal way of switching from CC to CV.  So having short pulse of high current might be a problem for this regulator. This might even leed to overheating of the transistors, as the relays are not very effective with such dynamik loads.

Could the MCU be controlling the CC/CV transition also and its speed having something to do with the magnitude of such a spike in your opinion?

[torch]

The funny spike at the beginning seems to be part of the normal way of switching from CC to CV. 

If that was it, then the sudden in-rush plays a role too. Turning down the current 10mA to force the switch from CV to CC or vise-versa results in a smooth transition. Here's a sample that switched it from CV to CC and triggered a relay to click at the same time:




The transitions where the relays kick in are quite smooth going up too. This is increasing the voltage by 0.1V, triggering a relay:



(both under load -- my 5 parallel lightbulbs)

[Kleinstein]

Stability (that is protection from oscillation and well behaved reaction to load changes) is something every PSU shoud offer. This a mainly a question of following basic rules. This about finding the rigth size for the nF range caps nothing that costs a lot of money per unit. It gets more difficult if the PSU is trying to regulate very fast. So there a cheap PSU can be very good in this respect, and a high end PSU that tries to push the limits speed wise is likely to be worse in this respect.

Switching from CC to CV mode is a slightly different thing. This includes nonlinear effects. In control theory the problem is called wind up - easy to solve in a digital regulator, but quite complicated analog. So here well testet high qualitiy PSUs can be expected to be better. At least there should be not complete desaster.  As comlete tests are difficult, however there still is a chance to find a bad one from HP and good chinese one.
Somehow the shape of the peak when changing CC to CV looks very much as I expect from a standard circuit. So it might even be a hint that stability is likely good.

The real drawbacks of these cheap chinese PSUs is the thermal design: they are made to deliver much power at low cost. Here less safty marging saves costs and thus things get hot right to the limit. Similar main filter caps are rather small as they are expensive parts. So some units that get the smaller ones show rippel at the upper limit.  For me this PSU looks good up to about 3 A. A more reputable brand might sell such a supply with just the upper current limit reduced and some fuses added.  So if you know and respect the limitations such a cheap PSU is not hat much inferior to brand named one.

One other point the chinese save some money is to pots / DACs. This might be the reason for the strange jigged shape on startup. Instead of a good DAC there may be a ADC / DAC loop to set the voltage.

Besides problems with constant high power, the other type of overload, might appear with dynamic loads. This is because the relays are slow, and some ealier stages might get hot as well. Here more reserve helps, but using relays to switch taps can not cope well with highly dynamic loads. So even a HP unit might fail or at least stop with overtemperature error. So the file test is allready pushing the limits of such a supply.

[Salas]

Solid state relays are out of the question?

[Kleinstein]

Solid state relais have a relatively large voltage drop. So they are OK at 110 V or 230V, but not that good at 12 V. Switching at AC also has a second problem: the filter caps have to charge, and this gives quite some current spikes.

If mechanical relays are not wanted SCRs are a real alternative. Old school, but less voltage drop than solid state relays, as the can also replace a diode.

The other alternative is doing switching an the DC side  in a class G like output stage. Slightly more diode drop, but as changeover can be very fast there is no need for a worst case reserve for ripple etc.

[torch]

I have found a serious bug:

If the M4 button is held down for about 2 seconds, the PSU turns on Channel 1 and Channel 2 and starts randomly changing the voltage of one or the other channel (usually Channel 2). While this is going on, the other channel displays 0.000 volts and 0.000 amps, but may be outputting power despite the display. The only way to get out of this is to power down the PSU.

The on/off button will turn off the outputs, and the voltages will drop to 0.000v but the random voltage changes will resume as soon as the outputs are turned back on. Also, the outputs may or may not be "on" when the PSU is powered back up the next time (outputs should be "off" when the PSU is powered up), however all controls will function normally.

This behaviour will occur regardless of the keypad lock. No other keys will function with the keypad locked, but leaning on M4 for 2 seconds will still result in this bizarre behaviour.

[torch]

Speaking of bugs, the programming software is full of them.

First, the manual gives specific directions about what cable should be used for serial communications and how to use a terminal program to send commands. As I mentioned earlier, it also comes with a virtual panel and scripting software to control it via PC. It can execute a script and log the results. What came with it installed without any hassle and at first glance,, appeared to be working correctly with the exception of a screen resolution issue when running on an old XP machine. (the work around for that on my machine is to set the resolution to 800x600, apply, then restore the resolution. That resizes the window properly thereafter)

Well, things sort of work. The pretty graphical interface works, except it does not change the memory storage location on the PSU, it simply lets you have a second set of independent memory buttons, resident in the computer. Voltage changes are truncated at the decimal point. For example, 7.77v sets the PSU to 7.00v. Also, despite the real-time graphs of voltage and current for each channel, it seems to be logging only one of the channels. I suspect #1 but haven't verified that.

So I decided to try the terminal route. I still haven't figured out the serial port connection but either it's dead or maybe requires a null modem cable (manual shows straight pass-through). I did finally successfully communicate through the USB cable to the virtual Serial-USB adapter installed in the PSU itself.

The manual suggests MTTTY is suitable. It is not. Neither is  Terminal, Hyperterminal  and a host of others that do not send data until "enter" is pushed. The only way I could communicate manually is with RealTerm. Fortunately its open source, available at Source Forge. After selecting the correct port, and opening it, I could send commands via RealTerm's Send tab. 

The keyboard locks automagically when the USB cable is connected. The keyboard lock can be overidden manually at the keypad, but not in software that I could see.

Most of the commands will work despite the keyboard lock. Except the memory recall and memory store commands. That explains why the memory buttons are independently duplicated in software.

What is needed here is some decent software to replace the crap that came in the box. The capabilities are there in the machine, but accessing them is not.

[torch]

Here is a short video showing the M4 bug:

http://picturehosting.verhey.org/korad/Korad_M4_bug_conv.flv

(Sorry for the separate link. I'm not sure how to embed a video here.)

[torch]

Voltage changes are truncated at the decimal point. For example, 7.77v sets the PSU to 7.00v.

I confirmed this behaviour on a  Win7 laptop this morning, so it's not because I have an old XP machine on the bench.

[torch]

Here is a short video showing the M4 bug:

http://picturehosting.verhey.org/korad/Korad_M4_bug_conv.flv

Update: I contacted the US distributor with a description of the problem and the above video link, but so far I've heard nothing but crickets. I have now submitted it to Korad via a feedback form on their website. We'll see what happens, but I imagine they've all gone home for the weekend by now.

[torch]

Update:

The US distributor advises that Korad has fixed the problem, they have the fix, and they have tested the fix.

The bad news is I have to send my unit back for the repair. For some reason it can't be just e-mailed and uploaded by me.

[Kleinstein]

The bug ist likely a software error. But the powersupply may not have the option to do software update vial USB. In addition they may not want to give away there software without the small amount od protection of the µC - so they may preferre to pay postage for the few early users that realy care about the bug.

[torch]

In addition they may not want to give away there software without the small amount od protection of the µC - so they may preferre to pay postage for the few early users that realy care about the bug.

That makes the most sense -- they are afraid the firmware will eventually get into the hands of potential clones.

$5 says it's roots are open source though 

Anyway, it's good to hear that they have produced a fix, even though it's a PITA to ship this thing back.

[Hydrawerk]

Here is a short video showing the M4 bug:

http://picturehosting.verhey.org/korad/Korad_M4_bug_conv.flv

(Sorry for the separate link. I'm not sure how to embed a video here.)

Just paste the youtube link.

Code: [Select]

https://www.youtube.com/watch?v=g94mpom2Ahs