Got A Free Rattling Rigol DP832. Heatsink / Capacitor / LM317 Upgrade

[Jidouka]

After seeing Dave's review I ordered the DP832 power supply.  While I was taking it out the box, there was metal rattling in the case. It didn't instill a lot of confidence so I wanted to exchange it for another.  I received the replacement and asked how they want to go about returning the original unit, and they told me I could keep it.   Now that it's free, the warranty sticker was quickly removed to find that the toroidal transformer screws had fallen out.  A couple of violent shakes and the screws came out and were promptly reinstalled.  A programming header on the front panel also got smashed over, presumably from the transformer, so those had to be straightened out as well.

I had a lag between projects when I started this, so I thought I'd machine some copper heat sinks, replace the caps, and change out the single gauge LM317 with the double gauge version. I finished the top board and plan to do the lower board, but some unexpected projects came up so for now this will just be a top board upgrade thread.  I'll break up the heat sink and cap replacement into separate posts.  Let the upgrade begin!

[Jidouka]

Heatsinks:
What started this little adventure was issues with the replacement unit that I thought were supposed to be fixed.  I was just walking through the screens when it went into a reboot loop.  It rebooted 4 times within a 5 minute span.  It really pissed me off so I decided to tear apart the first unit and try my hand at upgrading it, and if it works out, tear into my virgin unit and do the same.

With the unit just idling in the lab, with an ambient temp of 65F, the main 5v heat sink that was the cause of the original reboot issue, got up to 110F. That’s not that hot, but I think we can do better.  I’ve been saving chunks out copper from the dumpster for years and thought this would a good use for them.   So I booted up the 50 year old mill and went to town.  As you can see from the pics, they’re taller, wider, and thicker than the OEM crap.  I increased the surface area and air flow with a combination of slots and holes.  Since the board is upside down, heat is going to rise to the bottom.  With the slot/hole combination, air should flow over and through rather than around.  The triac /rectifier area of the board has a bunch of cross flow slots which should be better than the original boxed in design.

 All of the OEM heat sinks together weighed in at 2oz vs. their copper replacements which weighed in at 1.2lbs.  I originally ordered 1.6mm screws/taps to bolt the heat sinks from the bottom, only to realize that I don’t have a chuck that can accept that small of a drill bit.  I wanted the heat sinks to be removable, so plan B involved soldering in mounting posts (14awg romex ground wire) that would fit into a receiver hole on the bottom of the HS and lock into place with 2mm set screws.  To ensure electrical insulation of the HS to their component as well as surrounding chips on the board, I put 3 coats of electrical insulating varnish around the bottom/length of the free standing pins.   Before installing the HS, I also put standoff plastic washers so the copper wouldn’t be sitting on the traces.  To help with heat dissipation, the single gauge LM317 regulators were replaced with their double gauge version and the crap watery heat sink compound was replaced with Arctic MX-4.  The only heat sink not replaced was the one on the other side of board.  I just reused one of the aluminum HS for the free standing LM317, which should be good enough but I may make copper replacements for those.

Results:
Idle temp of the main 5v regulator, which is all I can measure with the cover on, is now around 90F.  Not a major improvement but still much better.  I’m currently running a load profile to burn it in, and it has yet to break 100F; it actually gets better due to the fan speeding up.

[Jidouka]

Capacitors:
I wasn’t too keen on the 1000-2000hr Samyoung caps and wanted to replace them with either Nichicon or Nippon chemi-con caps.  Why do manufactures put shit caps in their products?  In quantity, caps cost nothing, and are the first failure point.  I would bet money that if Rigol offered a premium version of the DP832 with high quality Japanese caps,  all their Samyoung versions would sit on the shelf to rot.  The top board upgrade cost $13.46(single piece price).  If you buy in large quantities, it’s more than 75% less.  So why do they bother with other brands, I don’t get it.  We’ll actually I do, it’s not meant to last and they’ll do whatever it takes to get it out the door, that includes not installing the toroidal transformer screws.

Anyways, they only use 5 different types of caps on the top board.  I went to digikey and ordered a bunch of different sizes of caps based on voltage, ripple current, and life.  I took some measurements with my LCR meter and here’s what I came up with.  All the measurements were taken at 120hz.  This is just a comparison at one reading frequency to compare the difference. The data sheet specs are also just for comparison, since there’s all kinds of de-rating factors. (freq multiplier .4-.75 at 120hz for the low ESR caps)

LCR Measurements: (Rated Capacitance/Measured uf/ESR/Ripple/ Life Hrs)

-OEM Samyoung Capacitors (Ripple Current Rating@120hz)
-35v 100uF KMG x 4:  97.8uf/1.133ohms/168ma/1000-2000hrs
-16v 1000uf KMG x 1: 925.7uf/.2314ohms/680ma/same
-25v 2200uf KMG x 2: 2019.7uf/.1586ohms/1.297amps/same
-35v 1000uf KMG x 2: 962.9uf/.149ohms/1.023amps/same
-63v 10,000uf TDA x1: 10,355uf/.0885ohms/4.4amps/2,000hrs
                              
-Replacement Nippon Chemicon/Nichicon Capacitors: (Ripple Current Rating @100khz; .40 to .75 frequency multiplier@120hz)

OEM 35v 100uf: (Ripple Current Multiplier. 40)
   -35v 100uf KYB: 109.63uf/.4215ohms/540ma/7000hrs
   -35v 100uf KZN: 97.24uf/.4782ohms/700ma/7000hrs*

OEM 16v 1000uf: (All 25v 1000uf) (Ripple Current Multiplier. 60)
   -KY: 1019.1uf/.875ohms/1.91amps/10,000hrs*
   -KYB: 996uf/.0625ohms/1.65amps/10,000hrs
   -KZH: 1034.7uf/.0466ohms/2.25amps/6,000hrs
   -KZE: 989uf/.0446ohms/2.36amps/5,000hrs

OEM 25v 2200uf: (Ripple Current Multiplier. 75)
   -25v 2200uf: 2277.7uf/.0298ohms/3.66amps/10,000hrs*

OEM 35v 1000uf: (Ripple Current Multiplier. 60)
   -35v 1000uf KZE: 1027.3uf/.0346ohms/2.77amps/5000hrs
   -35v 1000uf KZM: 962.8uf/.0385ohms/2.48amps/10,000hrs
   -35v 1000uf KY: 937.3uf/.0482ohms/2.23amps/10,000hrs
   -50v 1000uf KY: 1039.7uf/.0417ohms/2.92amps/10,000hrs
   -50v 1000uf KZN: 1023.5uf/.0325ohms/3.81amps/10,000hrs*

OEM 63v 10,000uf: (Ripple Current @ 120hz)
   -63v 10,000uf Nichicon GU: 9465uf/.0727ohms/4.69amps/3000hrs
                           

After weighing the options I choose the ones marked with an asterisk.  I was kind of worried about the low esr value, not knowing how it would affect the voltage stability but I’ve been running a power profile on it and it seems ok.  Max output ripple is 1-2mv at 30v/2.75amps and the 5v logic power has the same 2mv ripple.  I think this is a literal example of "Gilding The Lily"

[Jidouka]

Preliminary Results:

5v Logic Supply:
-Ripple Voltage: 2mv
-Temp: not higher than 98F loaded or idle

Channel 1 Output:
I ran a power profile with my load tester at 250ma increments every 60seconds up to 2.75amps@30v.  I just ran a min max profile with a fluke 289 and the max voltage ripple was almost 0mv. It jumped up 2mv from 29.994 to 29.996 after 2.24min and stayed there for the rest of the test.  I ran the same test with the virgin unit and the max voltage ripple was 6mv.  I'm not saying this is conclusive proof that this upgrade improved anything but it's different.

[rs20]

I'm always suspicious when I see copper heatsinks, since the majority of the thermal resistance in a heatsinking system is in the heatsink-to-air and junction-to-heatsink interfaces -- especially in a passive system like this. So your little 10 degrees F boost might just be the arctic silver alone to be brutally honest.

Also, many linear regulator topologies require a certain minimum amount of ESR in their output caps, or else phase margin is compromised. So one should be slightly cautious when replacing with lower ESR caps.

[Jidouka]

What I really need to do is void the warranty on the other and do side by testing to see what changes make a difference.  A simple test for the heat sinking capability would be to just replace the heat sinking compound and see if that alone makes the difference.   As far as the ESR value, I have same concerns.  What I've read about it, some say it's matters more for ldo regulators rather than linear regulators and others say the opposite.  We'll see if it holds up, otherwise I may just leave the rest alone.

[JBaughb]

This was a bold thing to do. I am curious now that you've used it for a bit, what your thoughts are on your modifications. Also, have you done any further testing?

[Jidouka]

I guess I'm pretty happy with it. I haven't done any more real extensive tests on it as I got to finish up some other projects first.  But I have been using it and it holds rock solid; haven't noticed any oscillations or reboot issues like the first unit.  Eventually I'd like to get in there and do the bottom board and output terminal caps  as well as that glaring Capxon cap on the digital board.  On a side note, the first unit stopped rebooting, maybe that crap heat sink compound had to set in?

Was it worth it... probably not.  The cap upgrade sure, but the heat sinks maybe not.  It was really just a filler project and I thought it would be neat to probably have the only unit with 1.2 lbs. of copper in it.  In order to really know what improvements where made, the other unit would have to be voided as well.  One thing i'd like to know, is if changing the heat sink compound alone would yield the same improvements or not; I don't know?  From a couple things I've heard about their support services, I may not be losing much by voiding the warranty.  Maybe I'll be bold and tear into the other one  

[JBaughb]

The idea of it is really interesting. I mean, all this test equipment is build down to a price so the manufacturer can lower cost when building these units in bulk.  Once an end user has one maybe mild modifications could be warranted to increase stability or life. I mean, thats essentially what we do when we build our own items. Over-engineer them. Use components that cost more but with specs that are wayyyy better than what is needed at a minimum.

A better heatsink or higher quality caps might extend the life of these items.

[Jidouka]

That's really what I was after.  If mild improvements can extend the life and possibly, even if slightly, improve performance, then I think it's worth it.  I really do like this power supply, and plan to keep it for a long time.  Mods like this, as well as other quality tooling/test equipment, if over engineered and treated with respect, should be faithful servants for years to come.  I'm a firm believer in buying mid to higher end gear, even if I have less, because the build/component quality is much better, but it seems we all eventually succumb to that alluring low price point.