EEVblog #314 - Korad KA3005P PSU Teardown

[PA0PBZ]

Interesting to see that it has 5 memory leds but only 4 memory buttons, guess they make more than 1 model

[firewalker]

WTF... They scraped out the infos of a chip that has a unique id? Idiots, just an "lsusb -vvv" will reveal everything.

Alexander.

[Bored@Work]

I must say I dislike several of the things Dave managed to like about that thing. The earth wiring all over the place, instead of all connected properly to one central point The cooling directly at the fan, so the near-field fan turbulences can generate the maximum amount of noise when sough at the heat sink borders The opto isolators, well, where are the cutouts below them? The tinning to increase the current capacity. Well, that never works well, cf. the resistance of copper vs. solder and you get an idea where the current flows. The hot glue right in the area where the PSU gets the most warm, and the hot glue gets the most soft again.

The big cutout above the DB-9 and USB is likely not for a DB-25, but a GPIB interface. Ok, as if anyone would waste money on enhancing that PoS with a GPIB interface.

The solder bridges and missing resistors in the DAC area could be the manufacturer's idea of a coarse calibration, the rest then done in software.

The metal plate could be to direct a little bit of air flow over the transformer.

[Rufus]

The plate in the middle is to guide airflow from the vents in the side past the transformer.

The choke and capacitor bodged onto the USB board supply along with the earth strap look like fixes for EMC problems

[veryevil]

Hey, would love a DaveCAD diagram of that '595 resistor network DAC. Seems interesting? What sort of accuracy / resolution could you get with that?

[markus_b]

At least it is not expensive ($99 on Amazon).

Dave, can you perform some basic measurements ?

Are the USB and serial ports insulated from the power output ?

[KTP]

@Bored

Come now, a little harsh weren't we?  I am sure the power supply is perfectly acceptable holding a door open or some such suitable application.  The heavy transformer almost guarantees it will work for that.

[rsjsouza]

The cooling directly at the fan, so the near-field fan turbulences can generate the maximum amount of noise when sough at the heat sink borders

I thought about a similar thing... Although I can't see the heat sink in detail, it seems to me it is just a thick plate with those thick but very short fins on the side, thus the fan seems to be blowing directly to a slab of aluminum instead of creating the proper airflow through longer fins..
But oh well... Maybe Korad's procurement department found out that a "real" heatsink would be more expensive than the fan/aluminum combination they chose...
I also think that scraping the device markings in the 21st century is really childish...

[Ketturi]

What is that white residue on PCBs, nearly every low-cost electronics have it. It's not the best designing, but if it works then no complaining about(my own diy linear "lab" PSU's are much worse, but heck they still work well).
Why so often there is design flaws etc. in cheap electronics. It's possible to engineer things cheaply with inexpensive components but bad design and poor parts together is dead last thing to do. And still it happens so often because consumers insists cheap and good, both rule out another, and cheap usually wins.

Resistor network, the old way to do D/A conversion, even some soundcards used that practice. Those good old times. I wonder how many bits that DAC correspond . One option could have been to use some micro with integrated DAC. But in this application there is no need to high precision or pcb space saving so probably they figured out it is most cost efficient way.

Microcontroller in PSU might be some generic cheap one, maybe even AVR. It could be reverse-engineerable comparing usual micros pinouts in that package to PSU's one. If manufacturer is stupid they might be even left flash/spi/whatever programing interface open but blah who cares. Pretty pointless to copy cheap design to even cheaper device :-D

[mikeselectricstuff]

I didn't like how close the current sense resistor was to the wiring from the mains switch.
And that DAC is ridiculous - PWM would be simpler and more accurate.
 

[gxti]

14:10 - Those look like JST VH connectors (or a clone thereof), rated for 10A.

How did I miss this when I was searching for my first bench supply? I ended up getting a $90 20V/3A unit from Mastech with a much poorer (although still acceptable) build quality, passive cooling only and it definitely can't sustain 3 amps at any voltage for more than a minute. I did a test run with a thermocouple and had to cut the power at 125C on the transistor case. Would have readily ponied up for something with memory and decent cooling! Still a decent unit, but now I'm looking for a switching supply for my second.

[DarkPrince]

I assume this is just another clone of the well known units. Question is, are there artifacts on the outputs (during power up/power down)? Granted nothing should be hooked up during the start-up anyways... I know i'd forget one of these times. POP.

[free_electron]

what is this obsession with underrated heatsinks with humongous fans on them .... its a piece of stupid stamped aluminum with a crappy fan.
if you buy an equivalent range supply ( max volt and max amp ) of a 'name;' brand the transformer is double the size and the heatsinkg 5 times larger ...

shave a nickle and dime... for 99$ no thanks.

i got an old analog tenmayears ago (2001) for 99$. that has a big honking transformer ,digital readouts and everything.
they spent more money making a cusomt rubber keyboard and controller pcb than the entire transformer rectifier cap pass transistor combined.

FAIL

[AntiProtonBoy]

Dave, I think you were a bit too forgiving in this review.

[nitro2k01]

I laughed a little when I saw the resistor network because I knew what it had to be. That appears to be a total of 24 bits, although they may be distributed between different things, say 16 bit voltage control and 8 bit current control. (There are three TL082's right to the right of the resistor network, so it's plausible.)
I can imagine that some people these days would give away 74hc logic for free in large quantities, or even pay you to dispose of a ton of it. I can see how a bunch of 595's and resistors would be cheaper than a "real" DAC or a microcontroller with enough PWM precision and channels, but good enough. And since it's going to generate a DC voltage most of the time, you can get away without the filtering needed for a PWM ( = less precious engineering time).

Also agree with Mike that the shunt is too close to the mains wire. One of the things I spontaneously reacted to.

[free_electron]

Just watched the whole video. Looks like that metal plate above transformer is to force the airflow down over the transformer before it goes out throught the side vents. That transformer is probably 'critically dimensioned' ( read: it runs hot .... )

The resistors look like they are marked 1002 and 2002 10k and 20k that would fit a classical R-2R style dac. They probably read the voltage and current using a sar algorithm...

This is most likely cheaper than something like a dual i2c adc and dac .. As they can use the dac as adc .. And they can get away with a larger output voltage.

[Bored@Work]

Maybe the MCU has an ADC and they use that for measurement, and the DIY DAC for setting. Or they skip the measurement (analog control loop), and the front panel just has a conventional ammeter and voltmeter.

Some more speculation. Maybe they need the external memory to hold "calibration" data for the "high precision" DAC.

[Scopeman1]

Dave, I reckon the piece of aluminum in the center that you did not know its function is to divert the air flow when the case is on. It will send it down past the txfmr to the power board.

[Scopeman1]

I think the fixed transformer voltage stops us from bringing in the USA versions on Amazon.  Mind you I don't mind paying a bit extra for a local supplier with local warranty. I see Trio Smartcal has it at $199 AUD.  Fair I suppose but would not want to pay more.   The return freight for a warranty repair to the USA would be more than it's worth!!

[tramjoe]

What a coincidence. I was exploring a (very) similarly spec'ed PS just last night, and my findings left me wondering about distributors margins and, as for many chinese lab hardware, the inevitable rebadging / copying dance that goes on there.

Look at that (in french, but you'll get the idea, and google translate is your friend). €300 ! Just for fun, I contacted the maker's Alibaba's distributor and his first price before even any bargaining (though in volume for 30 pcs) was lower than $100, that's about €100 after taxes and shipping to France :-) 200% markup aka 66% margin, life is good for electronics distributors in France :-) No wonder why they offered me a 30% rebate on this unit to compensate for a slight hotline problem I had with them. But even at that price (€210), it does seem a lot too similar to that Korad to be a good deal ;-) I guess the Korad is the ripoff/rebadge and the MCP the original one, because the MCP does have the 5 function keys, and looks a bit better constructed or at least european-market-conscious with  IEC1010 protected plugs and a bit nicer front panel.

What do you think guys ? Looks really _closely_ spec'ed, however the MCP unit is totally non-existent on the internet, apart from that french distributor and alibaba wholesale distributors, at least under that name and ref # :-)

Interestingly, MCP also makes 1A and 3A versions of the same unit, and a surprising 300V/400mA version of it.

Oh, and BTW everyone is quoting Dave's $99 amazon deal on the Korad, but there is actually a "D" version there which is non-programmable but has 110/220v input for $10 less, and I could swear I saw a couple of hours ago the P version for $75 with another amazon vendor with only 2 units left.... even showed it to my wife :-D

[amyk]

Maybe the MCU has an ADC and they use that for measurement, and the DIY DAC for setting. Or they skip the measurement (analog control loop), and the front panel just has a conventional ammeter and voltmeter.

Some more speculation. Maybe they need the external memory to hold "calibration" data for the "high precision" DAC.

After watching the video I believe the MCU is this, from the location of the crystal and arrangement of the pins:

http://www.stc-51.com/datasheet/STC12C5A-english.pdf

8051-based, not surprising. I've noticed '51 is very popular in Chinese electronics.

[T4P]

Maybe the MCU has an ADC and they use that for measurement, and the DIY DAC for setting. Or they skip the measurement (analog control loop), and the front panel just has a conventional ammeter and voltmeter.

Some more speculation. Maybe they need the external memory to hold "calibration" data for the "high precision" DAC.

After watching the video I believe the MCU is this, from the location of the crystal and arrangement of the pins:

http://www.stc-51.com/datasheet/STC12C5A-english.pdf

8051-based, not surprising. I've noticed '51 is very popular in Chinese electronics.

Well yeah, it's tried and tested for them. Not that they will venture much outside of their "safe" zone
The "weird" and rare ones they use are ST's STM32F103

[poorchava]

Ya know, I'm seriously thinking about buying one (or two) of those (half-year bonus sallary coming |^^|). Turns out we have an authorized reseller in Poland offering most of the single channel solutions from Korad (on their website even 3-channel versions can be found).

KA3005P is priced at 335PLN (=99 USDB at current stock exchange rates) while KA3003P is at 325PLN (=96USD). I wonder what's the actual difference in hardware between those two. Obviously can't be too big if they are almost at the same price.

Maybe that's some kind of Rigol-bandwidth-method? Like different firmware or some jumpers on the pcb limiting the current?

[EEVblog]

KA3005P is priced at 335PLN (=99 USDB at current stock exchange rates) while KA3003P is at 325PLN (=96USD). I wonder what's the actual difference in hardware between those two. Obviously can't be too big if they are almost at the same price.

Maybe that's some kind of Rigol-bandwidth-method? Like different firmware or some jumpers on the pcb limiting the current?

I have been told that the 5A one has a bigger transformer, but that's the only difference.

Dave.

[poorchava]

Huh, that kind of explains such small difference in price. Of course software has to be different to go up to 5A with current setting. I'm wondering what kind of pass transistors did they use. I mean a pair of old-school 2N3055 bjt's could easily do it (plus they are very cheap and being manufactured by a dozen companies or so)