CircuitSpecialists aka Array 3645A PSU Teardown

[PedroDaGr8]

So I was looking around on ebay for a power supply because I desperately needed one. Being as I am short of money and can't afford much, I was looking for some deals. I came across a Circuit Specialists rebranded Array 3645A linear power supply. It does 0-36V/0-3A with both CC/CV modes. It also has an RS232 port for PC control, a nice jog wheel for adjusting voltages, a variety of buttons including the ability to do presets. (Un)fortunately for me it was listed as having a blown pass stage and a broken ground banana jack. Now I wasn't quite sure what a blown pass stage meant but when reading up on it the only reference i could find to pass stages were dealing with transistors. So I decided to give the guy a low ball offer (he rejected the first one but accepted the second at $40+shipping). So I figure great, lets see what I end up with. At worst, even if I can't repair it, maybe the parts inside will be worth the money. This PSU is sold under a few brand names Array, CircuitSpecialist, TekPower and a few others.

So I get the power supply from the box and plug it in, set the voltage to 12.00V, it defaults back to 10.00V. Try a higher voltage, it defaults to 10V. Anything lower than 10V, set just fine. This confused me a bit, and I initially though the logic board may be damaged. I later found out that there is a max voltage function to lock out higher voltages, once that was removed I had the full 0-36V. Anyways, I hook up my trusty UT61E to the outputs and turn the outputs on (it has an on-off button for the outputs on the front, which I like). Uh oh, 60.8V. Change the voltage to 3.000V, 60.8V. Clearly, something isn't right.

Time to open it up!


I see a rather large toroidal transformer, a large filter cap (Nippon Chemicon - 4700uF 100V), the circuit boards and some transistors and a rectifier mounted to a heat sink. The heat sink looks suspiciously like the ones I had on my old P4 processors. The eagle eyed amongst you will notice that I got ahead of myself and already removed the screws for the transistors and bridge. The transistors were held down by a screw with a split washer, a lock washer and an insulator. The bridge was held down by the screw, split washer and lock washer. Almost all of the screws in the device follow this pattern (screw, split washer, lock washer). The thick green and yellow wires are for the safety ground. Basically every single metal piece that is its own part (whether attached to another metal piece or not) is connected AT LEAST once to the safety ground. This was a bit reassuring to see.

The first thing I do is test the transformer. I see isolation between all cables (good) but one cable shows continuity between each wire. At first I think maybe the secondary has shorted, but then I check the resistance and realize that not only does the transformer have multiple secondaries, it also has multiple taps on one of the secondaries (it supplies 4 different voltages, as shown in the second image below). I then proceed to test all of the major resistors and the large filter capacitor. All of these parts appear to be within spec.



Now the sil pad insulators, the tapped holes for the screws and the black connector (i guess a ground to prevent charge build up?) are visible. All of the wires are sleeved in what appears to be some sort of heat/impact protection tubing. It's much thicker and stiffer than heatshrink. The green circuit in the back behind the heatsink is for the RS232 connection to the PC.


Close up of the bridge and transistor. The transistor is a ST TIP35C, the bridge is a Hy Electronic KBU 8D (200V, 8A). Soldering is pretty clean and simple. Not sure why certain traces have solder mask and others don't (heat dissipation maybe).


Now that I have the board isolated, time to check the transistors. One of the transistors shows a dead short between Gate and Source, the other shows dead short between all legs.  Since source and gate are paralleled between the devices I remove the clearly dead one. So I can retest the other it checks out as fine (I ended up replacing it anyways later on just to be safe as the source-drain resistance was only 1Mohm whereas the new stock was around 20Mohm). This is the offending transistor. I broke the leg off post-transistorectomy, its basically a dead short between all three legs (about 2.4 ohms between each).

So I go on a spending spree at Tayda and buy a boat load of stuff: TIP35c, TIP42C, 2sc1815, 2n5551, the various op-amps on the board (TL082, LM329N and JRE4855), the four voltage regulators, plus a bunch of other stuff that I wanted for my component collection.

Since it would take a week for delivery, lets tear this baby apart. 


This is the main board removed from the device (the leads are soldered to the front panel unfortunately). The layout is pretty clean, power devices on the left, op amps and related stuff on the right. Soldering is acceptable with no noticeable flaws (leads could be trimmed a bit shorter but thats just me being nit-picky). Note there are 4 leads going to two jacks. Two carry the power, the other two I guess are to sense the voltage at the jack. This means it might be pretty easy to add a couple more jacks and carry the sense connections outside of the case. I'm quite curious why there are so many op-amps (there are 5 single op-amps and 1 double op-amp). I am guessing they are all parts of the control circuitry that I am to much of a novice to understand. Everything else on the board are the required resistors, loads of box caps, voltage regulators, etc. Everything appears to be quality name-brand parts. The relays are from Tyco, the main cap is from NCC,  the other parts are from companies like TI, ST Micro etc. The only let down is the various other electrolytic caps besides the main filter, they are all damn Jamicon.



Moving on to the heatsink, we see what appears to certainly be an old P4 heatsink but it does have a Sunon fan (which IIRC from my case modding days is a good company).

So I get the parts in from Tayda and everything looks perfect. I install the new transistors and mount everything back as it should be. Cross my fingers and power it up.



VICTORY IS MINE! Clearly it needs to be recalibrated a bit, especially on the lower ranges. Anybody notice anything odd about the second image? Yep, this PS is rated 0-36V but is actually able to do 0-37V. Lastly, I decided to test it on mV. Testing it out with my multimeter (and no load) it was able to go down to around 1mV but it bounced around a decent amount. I am going to grab a few loads from work (and try those out) see if it stabilizes. Clearly,at 10mV it needs calibration here as well. That being said, they include the calibration software on their website, so that if you have the right equip you can do it yourself. Keep in mind, all of these measurements are basically 5 min after turning it on. When I get the load resistors from work, I will let it have a full hour to warm-up and then test it out.

I'll let others much more knowledgeable than me evaluate the quality of this PSU. I'm just too new to this to really evaluate it properly.

[Deckert]

Hi Pedro,

That's a really cool supply. Oh and very nice job on the repair!

--deckert

[PedroDaGr8]

Hi Pedro,

That's a really cool supply. Oh and very nice job on the repair!

--deckert

Thanks, so far, since I haven't really loaded it, it's been whisper quiet. Just checked my stash at work and I have a 13R00 100W resistor and a 5R00 100W resistor to mess with, both which served as braking resistors in a dead centrifuge. As an odd aside, the centrifuge died by the oddest way possible. It wasn't anything dealing with the main function of the centrifuge. Instead, one of the motorized latches for the lid shorted out and took the whole control board with it.

[robrenz]

Nice job and thread

[PedroDaGr8]

Hi Pedro,

That's a really cool supply. Oh and very nice job on the repair!

--deckert

Thanks, so far, since I haven't really loaded it, it's been whisper quiet. Just checked my stash at work and I have a 13R00 100W resistor and a 5R00 100W resistor to mess with, both which served as braking resistors in a dead centrifuge. As an odd aside, the centrifuge died by the oddest way possible. It wasn't anything dealing with the main function of the centrifuge. Instead, one of the motorized latches for the lid shorted out and took the whole control board with it.

OK I decided to load it up tonight. Mounted the large 100W resistor to the side (both the 5ohm and 13 ohm, one on each side). Sure enough, once loaded all of the bouncing at the low mV levels stopped. THis meant that I could now reliably dial in 1mV and it would hold a steady value (albeit a wrong one because it really needs calibration). I set all of the voltages between 1 and 37V and graphed their offset from the typed in value as well as their percent error and a few things become clear: one, the actual voltage offset remains pretty constant and two, i need a better DMM. The UT61E is great but you can clearly see an error appear from switching ranges at 22V. Before the range switch, the average offset is 0.0405, above 22 the offset is 0.119V. Otherwise, the STDEV of each range is around 0.012V. Too bad I'm tight on funds but maybe in the future.


Also, as for sound/noise level, its not bad at all even after 60min at 100+W and 60min at 52W. I have had desktop PC's with louder and more annoying fans. It's definitely there but far from a nuisance.

[saturation]

thanks Pedro, this is a good review.  The Array PSU does look like a good unit; don't see many Chinese branded supplies that would opt for a toroidal transformer if they weren't thinking of lab quality from the outset.

As for a better DMM, given your skills you'd appreciate an HP 3456a, it provides the functions you need and is typically under $100, delivered.

[PedroDaGr8]

thanks Pedro, this is a good review.  The Array PSU does look like a good unit; don't see many Chinese branded supplies that would opt for a toroidal transformer if they weren't thinking of lab quality from the outset.

As for a better DMM, given your skills you'd appreciate an HP 3456a, it provides the functions you need and is typically under $100, delivered.

Yeah, I've been pleasantly surprised with this PSU, especially considering I paid so little since it needed repair. Between the  safety features (like grounding all panels to a central point), the component selection  and the top notch customer support, it seems like they desire to be more than just another budget player.

Also I did a couple of tests tonight using my oscilloscope. The first image is loading the inputs and turning the device on. You see a nice (albeit a bit slow) turn on with no overshoot. In the second, I have the scope single shot trigger on me plugging in ~50W of load. The idea is to see how the power supply goes from no load to 50W. Yeah I  know this isn't the best method, but it's all I  had the ability to do right now. The noise at the beginning is from me plugging in the banana plug. To be honest the slight spike only appears around half the time so I can't guarantee it's not noise from me plugging in.



In general this supply seems very well behaved. At most minimal over shoot in worse case scenarios and no overshoot at  turn on.

As for the DMM, I picked up a couple Keighley 199s of eBay (with the scanner cards) supposedly not working. One works great the other will soon.

Sent from my EVO using Tapatalk 4

[madscientisthut]

I have one of the ARRAY 3645A power supplies, and I just finished calibrating mine since it just started reading a little off, it has always been spot on until this week and I have had it over a year.

Calibration was really simple, the hard part was actually connecting the coms to the supply. I used a FTDI cable (TTL-232R-5V) and cut the connector from end of the cable off, then I soldered on a pin to each of the following wires: GND (Black), TX (orange), and RX (yellow). I then connected the wires to the back of the supply, GND to pin 5, TX to pin2 and RX to pin 3 on the 9 pin d-sub connector. I plugged in the FTDI to the computer and found out that it set up for COM6 on my system. I downloaded the calibration program from array (software.rar), unzipped the rar, installed and ran it. Under setting, I setup for com6 at 9600 baud, ID 0. Double check the baud and ID setting on the power supply itself by using the menu button (8 key) on the supply then scroll to baudrate set, set it to 9600 use OK key to go back, then scroll to address set, set it to 0, select OK, then press ESC. You need a good meter to calibrate, I used a fluke 87 (it is the best I have). The calibration runs through 5 voltages( 100mv, 3.999, 4.01, 37.0, and 20) and asks you to input each value displayed on the meter. It then does 100mA and 3.1A when it is finished it says successful.  I checked it against the fluke after cal and it was spot on. I hope mine does not have something going wrong and loses calibration again.

If you need the program and cannot find it, I have a copy of it.

Hope this helps you out.

BTW I liked your tear down article it was nice to see.

[PedroDaGr8]

I now have a meter I trust a bit more than my ut61e. Anyways, I checked out the meter using my HP 6114a and across the board every voltage I tried either matched perfectly or was within 1 count. So I checked the 3645A.  Basically, it reads 1-2 counts low on all settings. I'm pretty certain i could calibrate this out if i wanted as its pretty uniform but honestly at this close it could be the meter or the power supply. Not too shabby considering it was used and abused to the point of failure before I got it and was not calibrated after repair. 

After owning it a year i must say its a great budget power supply: Acceptable ripple(I measure under 1mV), surprisingly accurate for the price range, capable of going down to a few mV, the company cares about after sales support, etc. 




Note: on this one if I set the meter to mV it reads 99.9mV.





Sent from my EVO using Tapatalk

[BravoV]

Anyways, I checked out the meter using my HP 6114a and across the board every voltage I tried either matched perfectly or was within 1 count.

So the 6114A now also being abused serves as an adjustable voltage reference huh ?   

Congratulation on the new Fluke DMM btw.

[robrenz]

That's because the 6114A if in cal. is 4 times more accurate than the 83V at 10V 

[idpromnut]

Anybody notice anything odd about the second image? Yep, this PS is rated 0-36V but is actually able to do 0-37V.

I seem to remember that this may be related to the transformer taps + the actual line voltage being a bit higher than the nominal rated input voltage (which results in a higher maximum output voltage). You should see if your line voltage is running a bit higher than normal for your area, and that might explain the higher output voltage.  I know on my power designs supplies, the output voltage is spec'ed at 60V, but is actually closer to 62V.

[PedroDaGr8]

Anyways, I checked out the meter using my HP 6114a and across the board every voltage I tried either matched perfectly or was within 1 count.

So the 6114A now also being abused serves as an adjustable voltage reference huh ?   

Congratulation on the new Fluke DMM btw.

The 6114A says its a low cost calibrator. Honestly, my 6114A was dirty but based on the sheer number of calibration stickers I doubt it was abused. Well-used yes, abused doesn't look like it. That being said, I mainly used it as a sanity check. As I have ZERO calibrated equipment right now.

The fluke was one of a lot of 5 "busted" fluke meters I purchased. Two (83V and 87 series I) just needed to have the contacts cleaned for the display.  The others need some work but I think I can revive at least one more.

Anybody notice anything odd about the second image? Yep, this PS is rated 0-36V but is actually able to do 0-37V.

I seem to remember that this may be related to the transformer taps + the actual line voltage being a bit higher than the nominal rated input voltage (which results in a higher maximum output voltage). You should see if your line voltage is running a bit higher than normal for your area, and that might explain the higher output voltage.  I know on my power designs supplies, the output voltage is spec'ed at 60V, but is actually closer to 62V.

This isn't the case here because all of the settings are WELL within spec. I can actually set the voltage on the settings to 37V (despite the max listed at 36V).

[chickenHeadKnob]

Pedro writes: Not sure why certain traces have solder mask and others don't (heat dissipation maybe).

Hello Pedro Enjoyed this teardown, was usefull. The only routine intention/purpose I have seen for bare traces in power electronics is so that extra metal can be added there for increased current carrying capacity. Say the board is two ounce copper the circuit may still require greater current capacity in certain connections so solder is added, or solder plus copper wire or bus bar.