Home brew DC Load design review

[Mr.B]

Hi all,

This is my first ever Power Project and my biggest ever single project.

Aims of the project:
1. Modular. One “control centre”, up to three “Power Modules”.
2. Each Power Module consists of four 10A capable channels.
3. Each Power Module capable of the highest maximums of the following: 120vDC, 40A, 400W.
4. 5mA resolution. (Theoretical 2.5mA is achievable).
5. Cooling is provided by Corsair H60 CPU water block cooler units. One on each side of the Power Module. Corsair have refused to supply me with the thermal characteristics of the cooler, so 200W per cooler may not be achievable. I might need to review the Power Module maximum down to 200W total.
6. Not to be a tight wad. This project is not based on any budget, cost is not really a consideration. For example, the controller section will be based on a 4D Systems uLCD-70DT… And they are not cheap.
7. Have fun.
8. Learn lots.

Key components with links to datasheets:
Murata Ferrite: http://nz.mouser.com/Search/ProductDetail.aspx?R=BLM18PG221SN1Dvirtualkey64800000virtualkey81-BLM18PG221SN1D
IXTH30N50L2 Linear Power FET: http://nz.mouser.com/Search/ProductDetail.aspx?R=IXTH30N50L2virtualkey65120000virtualkey747-IXTH30N50L2
Low noise Op-Amp: http://nz.mouser.com/Search/ProductDetail.aspx?R=OPA4188AIDvirtualkey59500000virtualkey595-OPA4188AID
Current sense resistors: http://nz.mouser.com/Search/ProductDetail.aspx?R=WSR5R3000FEAvirtualkey61300000virtualkey71-WSR5R3000FEA
Shift registers: http://nz.mouser.com/Search/ProductDetail.aspx?R=74HC595PW-Q100virtualkey66800000virtualkey771-74HC595PW-Q100
Thermal sense BJT: http://nz.mouser.com/Search/ProductDetail.aspx?R=PZT3904virtualkey51210000virtualkey512-PZT3904
Fan management controller: http://nz.mouser.com/Search/ProductDetail.aspx?R=AMC6821SDBQvirtualkey59500000virtualkey595-AMC6821SDBQ
Logic level FETs: http://nz.mouser.com/Search/ProductDetail.aspx?R=SI2366DS-T1-GE3virtualkey61360000virtualkey78-SI2366DS-T1-GE3
Back EMF diode: http://nz.mouser.com/Search/ProductDetail.aspx?R=SS1H10-E3%2f5ATvirtualkey61370000virtualkey625-SS1H10-E3%2f5AT
Voltage reference: http://nz.mouser.com/Search/ProductDetail.aspx?R=ADR380ARTZ-REEL7virtualkey58430000virtualkey584-ADR380ARTZ-R7
DAC: http://nz.mouser.com/Search/ProductDetail.aspx?R=MCP4922-E%2fSTvirtualkey57940000virtualkey579-MCP4922-E%2fST
ADC: http://nz.mouser.com/Search/ProductDetail.aspx?R=MCP3204-CI%2fSTvirtualkey57940000virtualkey579-MCP3204-CI%2fST
DC to DC converter: http://nz.mouser.com/Search/ProductDetail.aspx?R=MTU1D1209MCvirtualkey58010000virtualkey580-MTU1D1209MC

Eagle files:
Eagle Schematic: https://drive.google.com/open?id=0B3IHlyNdvrjoTzdfb0lhUFR4Tm8&authuser=0
Eagle Board: https://drive.google.com/open?id=0B3IHlyNdvrjoc2xrRmRRMGtoMG8&authuser=0

High resolution images:
Image of Schematic: https://drive.google.com/open?id=0B3IHlyNdvrjoejdrM2hxaWtEd3M&authuser=0
Image of Top, no flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoNmVsLVJNSHVCaDQ&authuser=0
Image of Top, with flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoUmJmNjhFQUkyTjA&authuser=0
Image of Bottom, no flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoWjlzRFN6dElFUzQ&authuser=0
Image of Bottom, with flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoSTd5V0h1WXJnaVk&authuser=0
Image of All, no flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoOWxEOXVvTUszTWM&authuser=0
Image of All, with flood fill: https://drive.google.com/open?id=0B3IHlyNdvrjoWFFQeHNMajVVTU0&authuser=0
Image with annotation: https://drive.google.com/open?id=0B3IHlyNdvrjoSFdCZVJEQTFZYjA&authuser=0


I would really appreciate review of the Power Module section.
Any and all critique greatly appreciated.
Many thanks.

Teaser image:

[void_error]

5. Cooling is provided by Corsair H60 CPU water block cooler units. One on each side of the Power Module. Corsair have refused to supply me with the thermal characteristics of the cooler, so 200W per cooler may not be achievable. I might need to review the Power Module maximum down to 200W total.

Hmm... if this beast of a graphics card takes an input power of roughly 500W and converts it (mostly) into heat I guess a Corsair H60 should be ok up to more than 300W although a nowadays normal heatpipe based air cooler would also do. I really hate all the bullshit surrounding the PC industry, especially around the 'enthusiast' (poor bastard we the corporation suck  money off) grade stuff.

[Mr.B]

Yes, it should be OK. Might be a bit of trial and error though.
A number of the standard heatpipe coolers seem to perform better than the water coolers.
It is just that I picked up the water coolers for about 10% of their normal retail price from a receivership sale.

[nctnico]

You better run some numbers on the cooling solution. That small radiator is not going to move 500W. My experience with consumer PC hardware like the one pictured above is that it can not dissipate nowhere near as much energy as advertised. Temperature based speed throttling is what is keeping the videocard cool.

I'd build a setup using a resistor and measure the steady state temperature while dissipating a certain amount of power in the resistor. From there you can calculate the thermal resistance of the cooling solution (you also need to measure the room temperature).

[Mr.B]

Thanks nctnico.
I am not planning to use the one pictured by void_error and agree with you on the manufacturers not telling the truth about thermal ability.
Corsair would not tell me anything about the thermal characteristics. After I told them what I was going to use it for they may have been worried I would discover the truth.

I plan to use the below model H60.
It is supposed to be rated for top end CPUs, so should easily be able to handle 125W.
I am hopeful to be able to push it to 200W, but I will need to run tests first.

[Bazillus]

Hi!

Although your principal design looks fine, your plans are really challenging.

According to my experience from the past, I would start with a single channel/module version with reduced power (e.g. 20 volt, 5 amps, 100 watts) built around a standard heatsink. Do some simulation on regulation behaviour and compare simulation data to measured data from the first prototype. From this knowledge, I would increase the power capabilities, start paralleling modules etc. You learn much more if you start simple and measure a lot. It's frustrating sitting in front of a highly complex design with a microcontroller finding out about basic analogue misbehaviour.

Do you have an appropriate power supply to test the unit at either edge of the specification? 120v/3.3a, 10v/40a?

Michael

[TiN]

Why need complexity of liquid cooling just for 200W? You can dissipate that power without much trouble with something like Xeon 604 copper heatsink and high-CFM fan.

Also your FET is rated 0.31 °C/W, which is not a best in class, while this is really important if you want maximize your power rating without blowing things up.
Since your project is not really going after cheapest possible way, i'd recommend using something in ISOTOP (SOT227) package, which is really nice to bolt on your thermal solution directly. Like IXFN420N10T comes to mind. I had using IRF FB180SA10P before, not cheap, but works a treat, taking 260-300W into aluminum sink without much sweat (TJ calculated ~90C with 200CFM 120x120x38mm fans). Later used IRFP90N20P as well, they worked well for 140-180W with Aluminum sink with copper base (Xeon 603)

Big mosfet photo

Also I'd ditch LMV324's, and replace it them to something more suitable, like AD8664, powered from +12-15V.
Always had ocsillations on some currents with LMV324.
And beware your grounds You will quickly discover that ground is not 0V anymore when you pulling 500W thru a load.
Now you don't have differential sensing for current shunts, so both channels will be sweeping back and forth (e.g. you set 20A both channels, you get 19A on one, 21A on second due ground offsets, component mismatch, etc).

[Mr.B]

@Bazillus
Thanks for the input. Yes, I appreciate my schematic skills could be somewhat improved.
I have done a mock up of a single channel version on perf board.
I was very happy with the results.
Untested complexity in this design is stuff like the closed loop fan controllers.
Yes, I have power supplies capable of testing.
I have a 32V 32A SMPS, a 64V 10A SMPS and access to two 48V forklift batteies (96V).
The forklift batteries have no current limiting of course, so I have to be confident my circuit is working properly before testing at 96V.

@TiN
Many thanks.
The only reason I have chosen the water coolers is because I picked up a box of 10 of them for NZD116, so at $11.60 each they are a cheap cooling solution.
I am not going to use the Corsair fans, instead using Scythe Gentle Typhoon 120mm, 4250rpm, 116.5CFM.
It is yet to be decided, but I may change the design to use something like Xeon 604 heatsinks.

That IXFN420N10T has impressive thermal characteristics!

Apologies for the bad habits in my schematic, I have used the LMV324 in the schematic to get the right footprint for the board.
It is my intention to use OPA4188AID - low noise, zero drift op-amp.

Regarding ground:
I have tried to design with the ground current issues in mind.
There are effectively 6 grounds on the board.
Grounds for FETs A0, A1, B0 and B1.
Analogue ground and digital ground.
Analogue ground is tied to digital ground through a surface mount ferrite.
The four FET grounds and the digital ground will be tied to a very heavy duty copper bus bar via short lengths of 6mm2 stranded cable.
This should achieve a star grounding design.

Is this the right approach?

Should I implement differential current sense?
Do you mean measuring the voltage on both sides of the current shunt?

Thanks again.