Array 3645A Power Supply Review
I recently went on a hunt for a new power supply, because I needed about 2 amps for my new project, while my home made ones
could only deliver 1A. Since I'm on a budget, the choices were limited to either old equipment or cheap new equipment,
so I ended up with a list of old HP/Agilent power supplies and a bunch of Chinese supplies like the Korad 3005P and the Rigol 832 in the other camp.
I have to say that the most appealing one was the Rigol 832, but being such a big beast I quickly ruled against it - my bench
is not wide enough for it to fit in such a way that I could comfortably use it.
From my new power supply I wanted current limiting with mA resolution, digital display, over current and over voltage protection.
On top of that, I wanted it to be programmable so I can automate various tests. This left me with only a handful of choices
and in the end, the only one standing was the Korad 5005P, however due to expensive shipping and high bank fees (could
only pay by bank transfer), I had to reconsider.
A bit more digging around led me to another Chinese manufacturer, Array, which were also making a really nicely looking power
supply, that also featured a keypad (I was already drooling at that point). I had little information available on them,
but they seemed to make decent gear so I jumped over to ebay and found one used for 100 pounds. Yey!
After I already bought the Array PSU, I found out about a Tenma one that seems to be a re-branded Korad, which I found on
Farnell for a good price and I'm planning to eventually buy as a second "big" PSU for my bench.
That being said Array may be found under two other names I know of: Tekpower and Circuit Specialists.
As with most products in this industry, you're buying them for their specifications, so let's see what I paid for:
|Line Regulation||Voltage||0-3.999V||0.01% + 3mV|
|4-36V||0.02% + 10mV|
|Current||0.02% + 8mA|
|Load Regulation||Voltage||0.02% + 10mV|
|Current||0.02% + 10mA|
|Ripple and Noise||Voltage||< 1mVrms|
|Programming Accuracy at 25°C+/-5°C||Voltage||0.1% + 20mV|
|Current||0.2% + 20mA|
|Readback Accuracy at 25°C+/-5°C||Voltage||0-19.999V||0.2% + 20mV|
|Current||20-36V||0.2% + 100mV|
|Protection|| ||Over voltage, Over current, Over Power|
* actually missing
|AC Input|| ||110/220V AC +/- 15%, 47 to 63Hz|
|Operating Environment|| ||0-50°C, 80% RH|
|Dimensions|| ||W 212.6mm x H 88.1mm x D 250mm|
It also has a communication port on the back, but I'm not sure what it is yet, since it comes with an adapter both for
RS232 and for USB.
Seems to meet my needs just fine!
Let me sum it up for you:
You can set the voltage in 1mV steps between 0 and 3.999V and in 10mV steps between 4 and 36V.
The current can be set in 1mA steps across the entire range (0-3A).
The readback resolution is 10mV from 0 to 20V, so if you set 3.999V, the PSU might read back 3.99V.
The readback resolution is 100mV from 20V upward, so if you set 20.55V, the PSU might read back 20.5V.
Esthetically, it's a very good looking and compact unit. Sturdy, metallic enclosure with nice lines that go well on test
equipment. The carry handles on the sides are useless for me, but otherwise they don't seem to take anything away from
the design so I don't mind them.
The front panel looks like something you would enjoy using, with a power button that, at the first sight, seems to actually
toggle the mains line, a multifunction keypad, output on/off button and a big clicky rotary encoder (it actually clicks
when you turn it).
binding posts are of good quality too and the earth plug is on the right (the green one), unlike what you find on other
brands, where, for reasons that escape me, the ground plug is either in the middle or next to the positive plug.
At the back of the unit we find the power plug with the built in fuse holder, as well as the line switch and the communication
Before the real review
First thing I did after I took it out of the box, aside from admiring it for a bit and checking for scratches, was to actually
test the unit in every way I could think of. I think it's critical that you get familiar with your tools because it's
important to know how they really perform, what are their limitations, how they behave when those limitations are met and
what problems they may have.
Before I go on with my findings, I'd like to mention one more thing that I noticed even before I bought it, it doesn't
have LEDs to indicate output or constant current mode. That status is displayed on the LCD but I think two LEDs would have
made for much better indicators.
Let there be voltage!
The thing I was most curious about was how accurate it is. I don't have any fancy equipment to get really precise measurements,
but I do have a very good Brymen 867 meter (0.03% + 2 for DCV and 0.1 to 0.5% + 20 for current) which should be sufficient,
considering that the accuracy you're usually expecting from these cheap power supplies is way worse.
Voltage was measured at the binding posts with and with out load. I did not calibrate the unit, but the previous owner might
have. I tried to run the calibration software, but had no luck on my system (Windows 7).
Voltage accuracy figures, with no load:
* Yes, it can do 37V for some reason.
The voltage accuracy figures, when driving various loads can be found here - it's on a different page because there's too much data in there and it was causing some browsers to struggle
when displaying this page.
The output voltage accuracy is beyond my expectations and all I can say at this point is that I'm very happy about it,
so let's test further!
The constant current test
While some people might argue that it's not as important as voltage regulation, I find myself using constant current
mode on my home made PSUs quite often, because sometimes you just need to quickly test a component that needs current limitting
and can't work properly by itself.
When doing this test, I wanted to set a maximum settable current limit so I wouldn't blow the fuse in my multimeter by
mistake. It was nowhere to be found, but there was a Max Voltage setting and a Max Power setting so I suppose I could have
done the math there. This looks like something someone forgot about, since it's trivial to implement.
Well, it looks like there's a 6-7 mA offset across the board. Because I can't calibrate it, I can't tell if the
offset can be removed or not and as sad as it may sound, it's still in spec (0.2% + 20mA).
I can't declare it a failure in this regard, because it does regulate current properly, even tho it's off by a bit,
but I can't declare it a winner either, since I was expecting it to be off by at most 1mA on the lower ranges and ~10mA
in the amps range.
A closer look at the output
When I tried to connect a banana plug to BNC adapter, surprise! It doesn't fit, the plugs are a few mm too far apart.
That didn't stop me, so the first thing we're going to look at is the ripple on the output.
For the following screenshots, I had the oscilloscope on it's lowest setting (1mV/div) the entire time, because everything
I tried looked like a straight line at higher settings. This is why, so you don't get the wrong idea, I also took this
screenshot showing the noise floor of the scope, which is what I get when the power supply is completely powered off but
all the leads and the scope probes are connected.
(noise floor of my scope, with everything connected - PSU leads, load, etc)
There was no ripple in constant voltage mode and the oscilloscope screen looked just as if the PSU was stopped, so very clean
output, including at high voltages.
I did find some ripple in constant current mode:
(note: the load is a 1 Ohm resistor so V = A in this case, with some
error because of the leads)
Constant Current ripple @ 10 mA
Constant Current ripple @ 500 mA
| ||Looks worse, but it's still only 4mV/mA of ripple.|
While it's no current source, the output ripple is low enough to say that it has a clean output, both in CC and in CV
Ramp up and ramp down
Since it looks like it's using several taps on the transformer and it has two voltage programming ranges (0-3.999V and
4-36V), they all need to be tested since you never know if they all behave the same.
For ramping up and down, I didn't want to use my electronic load so I wouldn't introduce oscillations that otherwise
wouldn't be there, which is why I resorted to only using a 1 Ohm resistor whenever possible. Please note that because
of the value of the resistor, wherever it's being used, the read voltage should coincide with the current (I = V / R,
so I = V / 1 = V), but because of the resistance of the leads, that's not the case.
As you'll see, it has a clean and slow start with no overshoot while the ramp down looks like a capacitor discharging.
The cap discharging is visible because there's no relay cutting off the output - personally I would prefer it if it
It appears that the ramp is almost identical from 0 to 35 or so volts which is why I didn't include all the screenshots,
with a small surprise on the 37 volts ramp up screenshot.
2V into 1 Ohm
| || || |
3.999V with no load
| || || |
5V with no load
| || || |
37V with no load
| || || |
zoom in on the 37V ramp up @ 10ms/div
Alright, so it turns the output on and off just fine. I also tried to shut it down from the power button and turning it back
on to see if it backfires or has any turn off / turn on behavior but nothing worth mentioning surfaced.
Constant Current Overshooting
Another thing I was interested in, was to see how it's dealing with loads that will make it start in CC mode, so I set
the voltage to 5V, the current limit to 1A, connected the 1 Ohm resistor and turned on the output. What I found was a bit
of overshoot (about 700mV which would translate to about 3 or 400mA - considering the leads resistance).
Disconnecting a load that kept the PSU into constant current mode also generates an overshoot (quite big I might add):
I didn't expect it to behave so well on this test, but it did. I connected the 1 Ohm resistor while the PSU was unloaded
and outputting 2V. To my surprise I was able to make the voltage drop at most 500mV for less than 200us which is really
The reverse test (disconnecting the 1 Ohm resistor) went even better, only generating a 250mV spike for less than 500ns.
| || || |
connecting the 1 Ohm resistor
After seeing this, I didn't even bother thinking about smaller loads.
As I was playing with the rotary encoder and changing the voltage, I noticed that when you go from under 4V to 4V+ or from
over 4V to under 4V, the output completely stops for 2 seconds before it jumps back to the set value. I believe this has
to do with the fact that you can set the voltage with 1mV resolution under 4V and it's probably switching references
or voltage dividers. Whatever the reason is, it's bad and this kind of behavior shouldn't be present in a lab power
At the very least, there should be a setting that allows you to set the voltage in 10mV steps down to 0V with out switching
to the high resolution circuitry and vice-verso, it should prevent you from going over 4V if you are in high resolution
| || || |
from 3V to 4V
from 4V to 3V
Even Bigger Problem
While playing with the rotary encoder to get a glimpse at the above glitch, I stumbled upon something worse. Turns out that
if you have the current limit set to something low (I had it at 10mA) and you change the voltage with the output on, because
of the output capacitor it will get itself into CC mode. That's fine and dandy, however when that happens, it appears
that if you cross the bridge between 3V and 4V, the output doesn't get turned off (as it happened before) and the new
issue is revealed, a huge spike of about 22V that lingers on for 200ms+. I suspect the previous glitch was actually a bad
attempt at fixing this very issue via firmware, but it's not a complete fix and it just adds salt to the injury.
Although it's a good supply overall and I can look away from the various overshooting issues, the last two problems are
pretty much show stoppers for me because the glitch appears right in the middle of the range I need the most, 5V and 3.3V.
This prevents me from using it to power sensitive circuitry.
On top of that, you may have noticed that I said nothing about over current and over voltage protections. That's because
it's either not there, regardless of what the catalog says, or I can't find it.
That being said, it makes me sad to say, but I can't recommend this power supply to anyone until this issues get fixed
either via firmware upgrade or by releasing instructions for a DIY hardware patch. I contacted Array about these issue a
couple of weeks ago and also mentioned the missing (but advertised) over current and over voltage protections. If I ever
get a reply, I will update this post with their thoughts on the issues.