EEVblog #315 - Korad KA3005P Review/FAIL

[pickle9000]

I think the biggest problem here is the level of "cheaping out". 3 or 4 shortfalls and the product hits the floor. The manufacturer could turn this into a win by fixing the problem and then sending out a few copies to be reviewed. I would certainly buy a product that has been redone and brought up to snuff.

The fact that they cut corners, case, finish, design, lack of cleaning and so on are an issue but if it fails out of the box your name is mud. The unit has a good front panel layout, simple to operate form factor, so a win there.

Looking at the web site you can see that most of their product line is based around a variant of this unit. That being the case I hope they have the incentive to fix it.

Is this another example of Chinese manufacturing cheaping out. Yes, and extra 10 bucks per unit and this may have been an acceptable product. I'm not saying great but OK for the money.

Can you blame the engineer, the build team, management. Yes all of the above.

...mike

[T4P]

I usually side with HK companies
but it all changed with this event  nah i'm kidding, just not this one. For now.

[gxti]

I agree that some of the early weirdness during testing is probably due to the control loops of the load and the supply interacting in non-obvious ways. But then the thing burnt itself out, no excuses there!

[LaurenceW]

Well, by the time I got here, others had already made the same points! But I concur:

• It looks like the design uses a software loop for voltage stabilisation, hence the ramps (up and over) while the software catches up. NEARLY ALWAYS BAD.
  I wonder if your earliest glitches when the output died were software related?
  The soft on/off switch will probably work by simply turning off the base drive to the (two) main series pass transistors. Since current continues to flow, that suggests fried output devices (or possibly fried common driver).
  The internal voltmeter probably maxes out at around 36V (was it?) because in normal use it would never even need to go that high. Under fault conditions, the output rises above this to the main supply cap figure, however. And at 57V, that seems a bit OTT anyway for a 3OV power supply.
  I doubt that encoder is "optical" - just a simple two-switch affair. CHEAP!
  I didn't like the quality of the PCB design, at all. Not laid out by a seasoned designer. Inconsistent use of vias, track widths and routing styles.
  The discrete D>A converter seems bonkers. Surely there are plenty of fine chip solutions which would use less components, hence less PCB real estate, assembly time, etc.?

SMUG MODE :[ON] I've just bought me a new  end-of-line TTI 3 channel power supply PL330TP  (30v x 3A). It is built like a brick dunny (as I think the antipodeans would say). The only problem with it is the HERNIA it gives me when I try and pick it up. It will still be going strong when I am long dead. Pays yer money. I might do a teardown one day.

[hans]

All power supplies fail.. At college we had an Agilent 3630A triple output. We used it to power a transistor signal amp of +/-10V.
There was an unit that was faulty on -10V, where it would ramp up to -30V and sometimes stay there (the panel meters indicated , but the scope confirmed). So, yeah, that kinda sucks if you make fireworks TWICE with tantal caps of 16V as decoupling. Oh well. At least we made fun of each other when caps blew up on the negative rail (ah you screwed that up mate!) and then it happens to myself the day after.  Then we figured out the PSU was bad.. 

The overall quality of this supply is really crap. When I saw I thought, yeah, it will do 30V/5A for a few moments and then explode in flames. Not built to last. The PCB's are dirty, soldering is bad, bodges everywhere, weird case construction, weird DAC, the current shunt looks more like an inductor (dave, can you measure it? - maybe it explains a part of the weird load transients, unfortunately it's now fried but the turn-off response would be interesting too), weird holes/isolation idea.

Furthermore, an R2R dac is a horrible concept. If you read the WIKI, you need to have matched resistors for that. Well, you can take them from the same batch, but as Dave showed with his batch measurements some time ago: 1% resistors measured over a range of 1000s will have a variation of 0.5% - 1%. I see they stacked up 16-bits for current and voltage. That's just a bullshit thing. I bet the actual accuracy isn't that great. Furthermore, if you control  3000 to 5000 counts, why do you need it?
Apparently PCB space and assembly is free there!

Furthermore, the relay switching seems too slow. Too little capacitance? 6800uF at 5A & 100Hz isn't a great deal.
I can understand them switching the relays.. They of course look at the live output voltage (after current limiting) and make sure the voltage difference across the transistor isn't too large. Those other cheap PSU's recently linked here do the same, and basically have the same problem. I've encountered this a few times on a rather large buck-boost LED driver.

I basically tried the same abuse. Put the output at 30V/5A, and turn on a 60W load with 15A inrush currents. The output drops to 3V/5A, and that's it. The PSU didn't die, though.

[ivan747]

Also rebadged as a "Quakko" supply (quack  )
http://www.mastechpowersupply.com/dc-power-supply/linear-power-supply/programmable-dc-power-supply-32v-5a-lab-grade-hy3005dp/prod_49.html

[Bored@Work]

UPDATE: Triosmartcal have withdrawn it from sale,

The thing I am asking myself is what was the procedure Triosmartcal used to decide to add it to their offerings?

Did they get a sample from Korad? Did they at least open it? Was the sample a special or from the normal production? Did they at least checked one when they took delivery of the batch they ordered? Did they check the manual for completeness?

and Korad are investigating...

Sure. Wana buy a bridge?

[ivan747]

It seems M5 memory led is the LOCK led:
http://www.ebay.ca/itm/KORAD-PROGRAMMABLE-PRECISION-VARIABLE-30V-5A-DC-POWER-SUPPLY-DIGITAL-LAB-GRADE-/110888868115?pt=BI_Electrical_Equipment_Tools&hash=item19d17dd913

by the pictures

I believe Dave received an early version.

[ivan747]

I almost wonder if they are doing something stupid, like implementing the control
loop in the processor, given the strange ramps at turn on.

Of course you would get terrible transient response doing it that way, but
I've seen designs on the net doing just that.

Scott

+1

[Electr0nicus]

The white residue comes from cleaning the PCB with an acetone like solvent. Acetone breaks up the molecular bonds of the flux leaving back this white substance if not washed off thoroughly . This indicates that they actually don't wash the PCBs in the solvent, but rather only dipping it into. When the solvent dries off, the flux is still there, with the difference, that it is now a white powdery substance, due to the broken up molecular bonds. You can easily brush it off using an old toothbrush f.e.  I know that, because I myself use acetone the most time to clean my homemade PCBs. Of course, you have to know, what plastics are on the board, because f.e. PVC gets attacked by the acetone. But apart from that, almost all plastic components on a PCB are acetone tolerant.

[wiedehopf]

Regarding:
Thread Original Video at 18:30

Once the Power Supply is in CC Mode, the constant power load will further reduce it's resistance and hence the Voltage across the load drops to near nothing.
How it gets there is beyond me.
Perhaps the initially low load resistance sucks dry the power supplies capacitors and when recharching them the supply goes into CC mode. Just a guess.
But that is not sth bad i'd say. Where in real life do you have a constant power load?

The breakdown on the other hand is really bad news. And escpecially the Over Volt Protection really works like a CHARM

Edit: grrrr ... auto youtube embedding destroyed the timecode so i took it out

[Electr0nicus]

The breakdown on the other hand is really bad news. And especially the Over Volt Protection really works like a CHARM

There are power supplies out there, which generate a overvoltage condition, that is not so obvious like on the Korad PSU. I f.e. have a ELV PS9530 30V 10A PSU and one day I discovered mostly by accident, that it puts out a huge overvoltage pulse when switched off (proper mains power switch).
http://www.mikrocontroller.net/attachment/70639/PS9530_KM_G_071030.pdf
This phenomenon only occurred when the relay was switched on and thus the two transformer taps were connected in series to deliver a output voltage >15V. I found out, that it was actually a design glitch with the freewheeling diode of the relay, thus the control circuitry was influenced and switched on the serial pass transistors completely, delivering 40V to the output for at least 20ms. I found a solution to prevent that and emailed it to the company, which has designed the PSU. But they weren't even interrested in solving the problem and said that "because nobody had ever complained yet/found it out, they wouldn't bother correcting the mistake they made". Absolute dickheads.
But end of rant! What I want to say, is that even if you buy a higher priced PSU, (mine costed 359€/425AUD) you are not immune to such problems. 

[free_electron]

The white residue comes from cleaning the PCB with an acetone like solvent.

cleaning pcb's in acetone is not done. acetone eats too many plastics. the displays for example would melt ... same for at exle of the rotary encoder ..

this residue is most likely water soluble flux that has not fully saponofied and washed away.
water soluble flux is hygroscopic and will yield the classic white soap-scum appearance if not washed properly.

[Electr0nicus]

The white residue comes from cleaning the PCB with an acetone like solvent.

cleaning pcb's in acetone is not done. acetone eats too many plastics. the displays for example would melt ... same for at exle of the rotary encoder ..

this residue is most likely water soluble flux that has not fully saponofied and washed away.
water soluble flux is hygroscopic and will yield the classic white soap-scum appearance if not washed properly.

Thanks free_electron!
I didn't know that up to now.
As I wrote, I use acetone to clean my PCBs, when there are no sensitive plastics on the board. Then I get the same type of white residue, but it's not soapy but powdery.
If this residue is hygroscopic, then I assume it must be a acidic or basic? Or I'm completly wrong. So if it's a base or acid, then this will be conductive soapy layer on the whole board, which can't be good 

[Denno]

Almost certainly a blown pass transistor in the series regulator. Be interesting to see the part number on the MOSFETs to see if they are under rated. The is why some form of crowbar output protection is employed in better engineered supplies. If the output goes higher than some threshold then an SCR or transistor shorts it to ground and blows the supply fuse. Supply still needs to be fixed but does not take your circuit with it.

The test with the constant load of 50W where output was dropping to 0.2 volts was understandable at the electronic load would present an almost dead short when the supply output was switched on. A better test is to use the load switch on the dummy load to turn the constant power load on and off. This way the electronic load sees the set supply voltage and only switches in the appropriate current load for the power setting.

[Bloch]

The is why some form of crowbar output protection is employed in better engineered supplies. If the output goes higher than some threshold then an SCR or transistor shorts it to ground and blows the supply fuse. Supply still needs to be fixed but does not take your circuit with it.

Why not a relay ? Is it not fast enough ?

[JoannaK]

UPDATE: Triosmartcal have withdrawn it from sale, and Korad are investigating...

Dave.

Next time.. Try to get some smoke out of it 

[free_electron]

The is why some form of crowbar output protection is employed in better engineered supplies. If the output goes higher than some threshold then an SCR or transistor shorts it to ground and blows the supply fuse. Supply still needs to be fixed but does not take your circuit with it.

Why not a relay ? Is it not fast enough ?

A relay is slower than an SCR to kick in but the main reason lies elsewhere
a relay needs power to energize and close itself... shorting the power rail .... removes the power from the relay ... so that relay would become  buzzer .... until the fuse pops
under a certain voltage the relay will also not have enough megnetic pull to close the switch...

an SCR is triggered on its gate and stay s in conduction until the current through it has fallen under the hold current. this hold current can be very low.
So: fire the SCR and it stays in conduction until the fuse is blown and the capacitors in the circuit are discharged far enough that the current has dropped to a few milliampere. Then it comes out of conduction . the voltage across the scr can be 0.1 volt . it doesn't care as long as there is current flowing it stays on.

[TRIO_Smartcal]

Greetings and thanks for all the feedback.

Firstly yes we have withdrawn this from sale until we get to the bottom of the failure. To answer the previous question about what process do we use to decide what to sell, well we have a process and this passed it.

We got two samples a 30V 5A and a 60V 2A.  As an ISO-9001:2008 accredited company, a structured product introduction process is mandatory. We look much further than just the build quality. We look at support scenarios and how we and the supplier can deal with them so :
- what is the warranty?
- Can we repair locally?
- Is there a calibration procedure?

The  samples do get pulled apart. For example the white residue on the boards was known to us, we've seen it on more than one suppliers' boards and it has never been a reliability issue.

For the price as Dave says the build quality was aceptable and we liked the user interface too.
We also have already fed back the incompletness of the manual plus we'd like to see the programming commands in there too.

Our performance testing entails putting a product through our NATA accredited calibration lab to make sure it meets its specifications which these supplies did. (NATA is similar to UKAS , NVLAP etc) Thanks to Dave's dynamic (in more than one sense of the word)  testing of the supply a potential weakness has been exposed that the placid conditions of a Cal lab would not reveal. The product was immediately pulled from sale as we will not sell a product we don't have 100% confidence in, and after Dave's review we have our doubts on this one until we get a manufacturer recommendation and/or rectification proces.  We are certainly not scared of putting a product through an EEV Blog tear down and product review as the result can only be beneficial for customers, us and of course the manufacturer that will need to act on the findings so that the market gets better products and we get happier customers.

We try to offer two classes of product and if you visit our web site   www.triosmartcal.com.au  you will see this in action. We offer world class brands like Agilent, but sometimes that's not what people need or can afford. Wants and needs are different. So for folks with budget constraints, or simply needing something that does not have a world class brand label on it we try to offer lower cost, value-for-money reliable products that do what they are supposed to do. Plus we need to be able to support  our customers locally after they have bought something from us. Our customers are everything to us, it does not matter how good our products are without happy customers and good staff to look after them we won't be here.

There is a formula used in marketing-speak that is very simple but very true:

Value = (Performance/Price) + Service .........OK I know the brackets are redundant but most marketers are not engineers.  Many people do not consider service when buying a product but it is an integral part of acquiring and owning a product.

The other true marketing statement is that "you get what you pay" for. Well it's mostly true unless you are being ripped off. So our challenge is to choose products that meet these criteria. We thought this power supply met our criteria and until we are sure it does, it is withdrawn from sale. By the way we know it is cheaper in the USA. That is to do with volumes when purchasing, freight costs, customs clearance etc. We set our prices to be competitive, offer value for money and provide an adequate return for the company so we are still here for you if something does go wrong. It is all a big balancing act. Proof of this can be seen in another post I saw on here comparing our price on a 25MHz digital scope to a much larger company's price here in Australia. We are almost 50% less expensive and at that price we meet all our criteria necessary to take that scope to market. (thanks to whoever did that for pointing that out).

Ultimately you pay your money and make your choice. At $199.00 +gst here in Australia it is competitively priced for a locally sold and locally supported product.  If we find out afterwards that something does not perform as initially expected, we won't sell it.  We have sold a few of these and have not had any returns. Nevertheless we did not know it could not handle the Dave treatment and now that we do know, it is off-sale.  For folks that want a main brand, the big-guns are coming down in price.  You can pay $360+gst and get an Agilent U8001A with 30V 3A. It is an excellent product for the price. Maybe Dave should do a tear down on one of those for comparison.

Anyway I hope that answers your question and also why we have withdrawn this product from sale. Until we know the reason for the FAIL and the preventative fix it stays off-sale. Dave has full permission to rip it to bits to find out what went wrong as it seems there is quite a bit of interest in this.  Just don't drive over it please Dave ;-)

[grenert]

Unless you really need the programmability or the digital display, I think many people would do much better (at the same price) with an old tank of a power supply like this one:
http://www.ebay.com/itm/HP-6227B-Dual-DC-Power-Supply-W-Option-40-0-25V-0-2A-30-Day-WARRANTY-/300743973561?pt=LH_DefaultDomain_0&hash=item4605bcdab9

There will certainly be no funny business with it!

[amspire]

Unless you really need the programmability or the digital display, I think many people would do much better (at the same price) with an old tank of a power supply like this one:
http://www.ebay.com/itm/HP-6227B-Dual-DC-Power-Supply-W-Option-40-0-25V-0-2A-30-Day-WARRANTY-/300743973561?pt=LH_DefaultDomain_0&hash=item4605bcdab9

There will certainly be no funny business with it!

Hey, that is my power supply!

It is a good supply, reasonably compact (compared with other supplies of the same vintage) and no fan to annoy you or to fail.  Proper independent thyristor based overvoltage circuit for when you are testing that board you cannot afford to accidentally damage.

For average use, I prefer the moving coil meters to the digital readout. I like having a coarse voltage and current control that can go from 0V to maximum in a single twist.

Richard.

[ivan747]

The is why some form of crowbar output protection is employed in better engineered supplies. If the output goes higher than some threshold then an SCR or transistor shorts it to ground and blows the supply fuse. Supply still needs to be fixed but does not take your circuit with it.

Why not a relay ? Is it not fast enough ?

It's too slow. About 3 orders of magnitude slower than a good transistor. Relays take milliseconds to switch. You don't want your $2000 dollar 1.8V prototype being exposed to 30V, do you? It is not uncommon to see transistors switching in tens of microseconds. Although, I reckon the slowest component in a overvoltage protection circuit would be the fuse.

[Alexei.Polkhanov]

I have BK PRECISION 9110 on my table right now. I got it for $299 from Newark. I connected it to 10 ohm resistor, set current limit to 20mA, voltage - 10V and pressed ON button - result is what seems to me rather large overshoot in voltage (see attached photo). This is is the only lab supply I have right now. It is a new device, good brand,  so I am not sure if it is so unusual to see large overshoot on these supplies in current limiting mode, unless you want to spend over $2000.
After all what matters, in my opinion, is power delivered to connected device over set limit in watts not how high voltage is or in other words 10v overshoot for 100ns is better than 500mV overshoot for 100ms, right?

[Scopeman1]

That looks worse than the KORAD.  Also BK Precision re-badge other products, like the Siglent Scopes, the Atten spec-an, Protek HH-spec-an and some of the power supplies look similar to those on Extech's website.  It would be great of someone started a cross-reference for all this re-branding nonsense so we can buy from the OEM and cut-out the big-name companies adding the brand name and price tag.  As for Alexei's power supply, does it say "made in China on the back panel?"

[EEVblog]

Greetings and thanks for all the feedback.

Firstly yes we have withdrawn this from sale until we get to the bottom of the failure. To answer the previous question about what process do we use to decide what to sell, well we have a process and this passed it.

*snip*

Thanks for the detailed reply Charles.
I will try and find some time to investigate what failed on it...
Would of course be nice if Korad released the schematic! 

Dave.