Playing with the Atten PPS3005S DC power supply

[bobski]

I'm a noob here, but what the hey.
I'm starting a thread since doing this in someone else's classifieds ad seemed a little inappropriate. I posted this yesterday and updated it a few mins ago:

One of these arrived at my doorstep this afternoon... It seems to have a few issues that need re-engineering.
- The buttons are slow. If you just quick punch a button, you'll get no response no matter how hard you press. You have to slowly and very deliberately press the button to get the desired action. It really gets annoying when you're trying to move the selected digit over.
- The switch to constant-current is slow, or maybe there's too much filter capacitance on the output. If you set the supply to 10V, 10mA and (expecting the current limiter to do it's job) touch the output to an LED, you'll end up with a quick flash and one less LED in your parts collection.
- The fan is constant speed, always on, and it's distractingly loud.
- The zip ties keep things tidy, but for some reason I feel like the AC lines straight off the transformer shouldn't be bundled with the wiring for the DC regulators.

On the up side, the unit is controlled by an ATMEGA88PU running at 8 MHz and 5V. There's a nearby 6-pin header which may or may not be ISP (I'll have to trace the pins)... This is all visible in the pic. I tried swapping in a 16MHz crystal to see if that would help with the control responsiveness, but got no noticeable change in behavior what so ever. I wonder if someone forgot to set the fuse bits and the controller is running on its 1MHz internal clock.
The displays appear to be driven by a bunch of 74HC595D shift registers and a mess of SMD current limiting resistors. Low tech, but effective.
Since the 'mega88 doesn't have on-board USB support, I expect the 4-pin connector consists of the TX and RX lines along with 5V and ground. That's all that would be needed to interface with a USB UART bridge.

[edit]
Yeah, it's ISP. From left to right (left most pin is numbered 1, so 1-6): Ground, VCC, MOSI, MISO, reset, SCK.
Let the hacking begin.
[/edit]

So next I'm going to try hooking it up to my AVR Dragon to check my theory on the fuse bits.

[bobski]

Bingo.


Unfortunately, the limiter operation still seems flakey.

[saturation]

Good work!  If we can get some of these Chinese test gear working properly, its does give you a bang for buck.  Can you characterize the PSU for us in terms of its ripple and output stability?

[bobski]

Can you characterize the PSU for us in terms of its ripple and output stability?

There's does seem to be some kind of harmonics going on.

The peaks hit about 8 mV at 15 us invervals, while the noise at the base is about 0.8 mV.

How would you characterize stability? Output voltage drift over long periods?

IMHO there is no need for such restrains.

Well, If nothing else, it was in the wrong forum section. ^_^

Seriously, that design decision might be the reason the fan needs to blow constantly.

Yeah, ok... I can see that. They would need temp sensors at board-level as well as the heat sink at the back. But still, variable speed would have been nice. Don't necessarily stop the fan, but merely spin it down so it isn't so noisy. So long as there's airflow, the heat sink temperature should suffice, as all the air flowing through the case has to pass around or through the sink on its way to the fan and out the back.

Is the ATMega part of the control loop (digital controller), or just providing the settings for analog control loop (digitally assisted controller)?

I'm not sure. Since the surge-y behavior of the current limiter doesn't seem to have changed with the clock speed, I suspect it's just digitally controlling an analog loop.

[bobski]

Can anyone recommend some good, free (or close to it) diagraming software? I worked with multisim in school, but it's not exactly affordable for the hobbyist.
I've got the basic AC-in to DC-out path of the analog board down on paper:

That covers about 1/3 of the board's area. Another quarter is simple 5V and 12V supplies based on 7805 and 7812 linear regulators. This is starting to seem manageable. Not included on that scribble are various little circuit offshoots from the main traces, as well as whatever controls the big NPN transistors... It looks like a couple of daisy-chained transistor amps, but amplifying what I don't know. Expanding the diagram so it's actually useful is going to get messy, so I would prefer to transfer what I have to software and continue from there. Which leads back to my original question: Does anybody know some good diagraming software for the hobbyist?

[bobski]

Slowly sprawling...

Turns out there's an unused 2-pin "Temp" connector (pins labeled B and W) on the analog board. I guess it goes with the offset threaded hole on the heat sink?

[saturation]

I haven't given this much exercise as I use paper too:

http://www.dz863.com/index.php

Can anyone recommend some good, free (or close to it) diagraming software? I worked with multisim in school, but it's not exactly affordable for the hobbyist.
I've got the basic AC-in to DC-out path of the analog board down on paper:

[bobski]

Alright... I think this covers all the bits on the analog board that play a role in output regulation. This doesn't include the 5V and 12V regulators (LM78xx linear), associated components, or anything else that might be on the board, like fan control. I just traced everything that seemed to have a role in controlling the big regulator section.

Any thoughts/analysis would be appreciated.

[bobski]

The wiring and polarities are correct. D13 is a glass case diode, but I can't read the part number as it is facing the circuit board (as is the case with a good number of the diodes). Perhaps it's a zener? I'll check it out tomorrow, along with D8, D9 and D10. All the rest look like regular rectifier diodes.

[Richard W.]

What happens when you set the output voltage close below the switching threshold of the transformer-coil-relay and then connect a high power load?

What happens with the output voltage when you increase or decrease the voltage and the relay switches?

I guess there are some nasty spikes...

[bobski]

Turns out D13 was the only zener diode.

[bobski]

I guess there are some nasty spikes...

Maybe. On the other hand, that's part of the A/C section, so it's swinging positive/negative at 60Hz. That line of capacitors after the rectifier might smooth out that hiccup just fine.

[Richard W.]

Richard W. on July 01, 2011, 12:52:13 PM

    I guess there are some nasty spikes...

Maybe. On the other hand, that's part of the A/C section, so it's swinging positive/negative at 60Hz. That line of capacitors after the rectifier might smooth out that hiccup just fine.

I don't know if 6600µF are enough in this case. I would have added more Cs for the relay thingy.
But 5 amps, capacitive load, that doesn't sound like a "longlife" construction.

I prefer a real, 100% analog lab-supply.

[bobski]

I decided I should probably be a bit more scientific about my tinkering. After replacing a suspect filter cap on the internal 12V supply (appears to have a puncture on its blowout seam, but no evidence of electrolyte leakage) I put everything back together and set up a simple test rig for the current limiter. It's a 120 ohm resistor as a load in series with a 1 ohm resistor as a current shunt. I hooked up my oscilloscope across the shunt (scope ground on the supply's negative post, test point on the shunt's remaining lead). Since it's 1 ohm, volts across the shunt should equal the current through it.
Anyway, I set the current limiter to 5mA and got the following when I close the power supply circuit:

The yellow curve is with the supply set to 30V, magenta is 15V and cyan is 5V.
Looks to me like in each case, it takes a full second for the limiter to get where it's going. The curve looks like capacitor discharge or maybe an averaging routine if the uC is playing an active roll. The flat top is interesting... I would like to know what's causing it.
[edit]
30V. Yellow @ 100 mA, magenta @ 50 mA, cyan @ 25 mA:

[/edit]

[Miti]

Can anyone recommend some good, free (or close to it) diagraming software? I worked with multisim in school, but it's not exactly affordable for the hobbyist.

Try Eagle from Cadsoft. They have a free version with some limitations but is more than enough for most of a hobbyist needs.

http://www.cadsoftusa.com/downloads/

miti

[Miti]

Richard W. on July 01, 2011, 12:52:13 PM
I don't know if 6600µF are enough in this case. I would have added more Cs for the relay thingy.
But 5 amps, capacitive load, that doesn't sound like a "longlife" construction.

I prefer a real, 100% analog lab-supply.

I'm not sure what you mean by capacitive load but this is what you'll see in most of the analog power supplies. If you like 100% analog lab power supplies this is one of them. The digital part is basically two digital potentiometers one for the voltage and one for the current.

What happens when you set the output voltage close below the switching threshold of the transformer-coil-relay and then connect a high power load?

What happens with the output voltage when you increase or decrease the voltage and the relay switches?

I guess there are some nasty spikes...

No, no nasty spikes in both cases. A rough calculation of the time constant at 5A, 6600uF and about 40V before the series regulator will give you about 52ms. The control loop can respond way faster than that and the voltage before the series regulator never drops below the minimum value during the relays switch.

I'm not sure. Since the surge-y behavior of the current limiter doesn't seem to have changed with the clock speed, I suspect it's just digitally controlling an analog loop.

Yes it is a digitally controlled analog loop and with 2 x 470uF at the output you'll still blow your LEDs until you replace them with 2 x 1uF for example. I wonder how stable it is in this case. I will give it a try one day and let you know.

The flat top is interesting... I would like to know what's causing it.

It is the cap(s) discharging on the resistor while the loop gradually increasing the current to compensate for the discharge. When the current reaches the set limit, the flat ends.

Does anyone have the manual for a different Atten power supply with the serial commands? The connector J4 is the serial port and I would like to check if only the input port is missing or the comm module is missing from the software as well.

Thanks,
miti

[Zero999]

What happens when you set the output voltage close below the switching threshold of the transformer-coil-relay and then connect a high power load?

What happens with the output voltage when you increase or decrease the voltage and the relay switches?

I guess there are some nasty spikes...

That sort of tap changing scheme has been used for linear power supplies, long before MCUs were invented. It helps to save power and reduce the power dissipation and heatsinking requirements for the power transistors. There's nothing wrong with it as long as there's enough hysteresis to prevent oscillation and the power supply rejection and transient response needs to be good enough to suppress the spikes.

After replacing a suspect filter cap on the internal 12V supply (appears to have a puncture on its blowout seam, but no evidence of electrolyte leakage) I put everything back together and set up a simple test rig for the current limiter.

That's unacceptable, perhaps you should consider replacing all of the electrolytic capacitors? If they've used shitty capacitors throughout it's likely to cause problems later which could be time consuming and expensive to repair.

Looks to me like in each case, it takes a full second for the limiter to get where it's going. The curve looks like capacitor discharge or maybe an averaging routine if the uC is playing an active roll. The flat top is interesting... I would like to know what's causing it.

It probably is a capacitor discharging. For a start there's 940uF on the output of the regulator, the RC time cvonstant at 120R will be 112ms and there will probably be another RC delay in the current limiter circuit. Notice how the current limiter seems to kick in more quickly at higher voltages? This is because the current will be higher so the capacitor in the feedback loop will be charged more quickly.

[bobski]

Hmm. So the verdict is that this supply's analog section suffers from over-filtering?

[Zero999]

It's pretty normal to have this kind of time delay on the current limit. If it were instantaneous, you may complain about the voltage dipping under surge conditions such as those caused by a capacitor charging , large incandescent light turning on or a motor starting. Ideally, it'd be nice to have a PSU with a variable time delay on the current limit so it can safely be used for LEDs as well as motors.

[Miti]

Be very careful with this power supply, the firmware (and maybe the hardware) has a lot of glitches. Here are few that I've seen:
- Press the V-set button and the current display blinks
- Turned the knob to decrease the voltage and it went up
- Press the V-set or the I-set (I don't remember) and the output turned on - very bad

After about a week of ownership I think that it is well overrated at 5A. The main concerns are the rectifying bridge and the shunt wire resistor. At 5A they get pretty hot and I don't think there's a lot of air flow in the area of the bridge.
On the plus side though, it has an input for a temperature sensor and the fan goes to the logic board so it can be controlled but that part is not implemented in the FW.
Anyone up to the challenge to write a new complete and more responsive FW?

[JSnyder]

With everybody posting up atten PS tear-downs, I figured I'd post up my findings/mods to this power supply. I think you have to buy these supply's with the knowledge that they do have issues and it's a project rather than a perfect product out of the box. Theres 2 versions of this power supply, old and new. The old ones have hot glue instead of yellow adhesive over all of the connectors, a hand wound current shunt wire, metal wire jumpers instead of 0 ohm resistors. The new has the aforementioned yellow adhesive, and a mechanically wound current shunt wire (more on this later).

The first mod to this PS was to cut out the stamped metal fan guard and replace it with a proper chrome wire guard (PC) this increases airflow quite a bit in general. Second the very loud "Runda" "good quality fan"(printed right on it! lol) was replaced with a silent Enermax fan.
 

The second mod was more involved, I recapped all of the large electrolytic caps with name brands. The three large input caps (after the rectifier) were originally 2200uf @63v they were replaced with 4700uf@63v Nichicon caps. All of the others were replaced with the same value Panasonic FM and FC series caps.



I did the crystal mod from 8.00 to 16.00 Mhz as Bobski pioneered, unfortunately on this board it appears that atten has set the reset disable fuse, therefore I am unable to reprogram it.

As Miti pointed out the bridge rectifier gets worryingly warm with nothing more than a dinky heat sink to keep it cool. I de-soldered it and relocated to the chassis as seen in a number of mastec PS's. Problem solved, barely gets warm.


Also I think Bobski's issue with slow current limiting and slow/no updating of the voltage during CC mode stems from the current shunt wire, My supply showed the same behavior as bobski's with the hand wound shunt in place, when i replaced it with the newer machine wound shunt it updated the V display and switched to CC very quickly. Don't ask me why this works, it just did for me YMMV.

I plan on integrating the pc control part of this supply in the future, I was able to track down the software, all I have to do is build the usb  bridge with an FTDI chip.       

I did some tests to see about accuracy. The load was a 3w cree attached to aluminum angle run at 1A






 
       

[saturation]

@ JSnyder, great mods.  That's what its all about, save you the trouble of building from scratch and a few 'upgrades' will give you a decent PSU.  From your post it reads you've done the homework of reviewing the other PSU build qualities on eevblog.

[bobski]

I went ahead and did those cooling mods (quiet fan, proper grille, moved rectifier to bottom of chassis) and will play around with the caps next.

Theres 2 versions of this power supply, old and new.

Hmm. I wonder if there have been any changes to the uC code?

Also I think Bobski's issue with slow current limiting and slow/no updating of the voltage during CC mode stems from the current shunt wire, My supply showed the same behavior as bobski's with the hand wound shunt in place, when i replaced it with the newer machine wound shunt it updated the V display and switched to CC very quickly. Don't ask me why this works, it just did for me YMMV.

Any chance of getting a few measurements of the newer shunt? Wire diameter, diameter of the coils, coil spacing (wavelength)?

I was able to track down the software

Any info on that?

[JSnyder]

@ JSnyder, great mods.  That's what its all about, save you the trouble of building from scratch and a few 'upgrades' will give you a decent PSU.  From your post it reads you've done the homework of reviewing the other PSU build qualities on eevblog.

Thanks, you hit the nail on the head, I purchased it after looking at a number of power supply reviews, and wanted the project of bringing it up to standards.

Bobski:

I found there is no difference in the actual boards between the two revisions, some of the parts changed however, jumper wires are swapped with 0 ohm resistors, new current shunt wire, new rotary encoder, different physical size caps, silkscreen on the back of the analog board. I went ahead and dumped the hex from the newer power supply's atmega88p, I don't know how useful it is in that form. I have no reason to believe that there is any change in fw though.
I just find it odd that you had no problem flashing your "old" style PS with the new fuses, I had no problem flashing my "new" style PS, and my old style wont go into programming mode. I went so far as to hot air the atmega from the pre modified supply, I just can't bring myself to put it on the working board.


shunt specs:
37mm lead spacing
~ 1mm diameter wire (I'm almost certain you could rewind the old one)
~ 16 turns total
~ 2mm between turns
~ 5.5 to 6mm coil diameter

software can be found http://www.atten.com.cn/Article/IndexDown.aspx?ChannelId=2&G=3&classId=27
click on the d/l symbol next to this one PPS3005S??
the installer is in Chinese, however when starting the application you can change it to English.
The software seems decent, I can't comment as I don't have the usb bridge built. I can tell you the the serial port on the power supply does spit out data.

Having said that, Bobski is your location in your profile accurate? If so we probably don't live far from one another. Send me a pm if you want to see the current shunt in person.

[JSnyder]

bobski here is some close ups of the shunt, Your going to need a 3-4mm or so former to wind it.



I also took a vid of it switching to cc mode, can you confirm that this is better than what you have mentioned in the past? If so, I really do believe the new shunt should help.