Overclock.net banner
41 - 60 of 246 Posts
Discussion starter · #41 ·
I have to order the parts for the baseboard still and then reflow solder it. I also need to send off and have some 16 gauge sheet copper laser cut to fit the bus sections ... which are obvious on the board. Then I also have to send off the regulator board and solder it up ... which should be easy. At that point, I will have a working power supply to test and tune.

So yeah, many weeks still. I can only work on it at night once I'm home from my day job so it's slow going. Plus I have other projects in the mix.

Oh, and I need to get the copper rivets that go into 3 of the holes to move current from one side to the other.
 
Discussion starter · #43 ·
Regulator board that snaps onto the bottom edge of the baseboard went out yesterday. I should have it back in about eight days.





This board fits vertically between the baseboard and fan and helps prevent air from being drawn into the bottom section. Cooling airflow needs to come in from the other side and be drawn across the other components and heatsinks before being exhausted up and out.

You can see on this board that you will set the voltage limit on the three rotary switches on the top right. Then the regulator will use the signal from the PWM fan header to drive the output between 0.5V and this setpoint limit. The limit representing 100% on the PWM input on the fan header. If you have no signal plugged into the PWM header, the regulator will attempt to drive the voltage to the setpoint or what would have been 100% of the PWM signal.

First pass will be limited to 12V at 72 amps in, 60 amps out at up to 36V max. It won't do 60 amps at 36 volts, it is limited by what's available at the input. You have to have all three 8-pin PCIe power connectors plugged in on the input side to reach the 72 amp point. I'm limiting this to 8 amps per pin on those ... which seems reasonable.

There's a dual footprint on the output connector of the baseboard so that I can put on a 76 amp pluggable connector in the future and raise the output current limit. You're still limited to the input power (12V * 72A = 864 watts) multiplied by the efficiency of about 98% for 850 watts at the output ... max.

The output connectors will accommodate up to a #6 wire.
smile.gif


If it's running normally, the green LED to the left of "Vout Reg" will be lit. If your PWM signal is too low and would result in an output below 0.5V OR if the cooling fan gets stopped for some reason, the regulator will stop and the red LED to the left of "Halt" will be lit.

I've waited to send this out because I had a bunch of other stuff to send and wanted to save on shipping. So I've got four different boards from various projects coming back. This just happens to be one of them.
 
Discussion starter · #44 ·
I've got the four temp sensor ports on the mezzanine board now. Five if you count the ambient temp sensor that's on the I2C bus.

Do I need a flow sensor port as well? Two? It will be a squeeze assuming they are the same 100mil headers that are used for 3-pin fans.

I assume these are all pulse style sensors and the ones I can find seem to want +5v for power. It looks like the Aquacomputer High-Flow (non-USB version) is the best available.

Thoughts?

If I could fit two flow with the four temp sensors on board, one could theoretically monitor the heat transfer in watts through two separate loops.

On the temp sensors, I've jumped through some hoops and I think I've got a solution to measure down to 0.1degC from -15 up to 100. This is without oversampling ... which I will probably do in the ADC. With oversampling, a lookup table, and some interpolation I can probably get down to 0.05degC reliably. Once I oversample, I can probably stretch the +/- 0.05degC down to -40.

I learned also that Analog Technologies, Puck's thermistor source, has a new thermistor with a 0.1% accuracy. It's the same device/size but with a tighter tolerance.

ATH10K0.1_1R25
 
Quote:
Originally Posted by Master Chicken View Post

I've got the four temp sensor ports on the mezzanine board now. Five if you count the ambient temp sensor that's on the I2C bus.

Do I need a flow sensor port as well? Two? It will be a squeeze assuming they are the same 100mil headers that are used for 3-pin fans.

I assume these are all pulse style sensors and the ones I can find seem to want +5v for power. It looks like the Aquacomputer High-Flow (non-USB version) is the best available.

Thoughts?

If I could fit two flow with the four temp sensors on board, one could theoretically monitor the heat transfer in watts through two separate loops.

On the temp sensors, I've jumped through some hoops and I think I've got a solution to measure down to 0.1degC from -15 up to 100. This is without oversampling ... which I will probably do in the ADC. With oversampling, a lookup table, and some interpolation I can probably get down to 0.05degC reliably. Once I oversample, I can probably stretch the +/- 0.05degC down to -40.

I learned also that Analog Technologies, Puck's thermistor source, has a new thermistor with a 0.1% accuracy. It's the same device/size but with a tighter tolerance.

ATH10K0.1_1R25
Very cool!

Flow sensor is a nice bonus, but not a deal breaker for me...and I definitely wouldnt want one at the expense of a temp probe.

I'm sure I'm in the minoryt but I run dual pumps on anything with a TEC after a close call(and recommend the same to everyone!), so never have flow issues or even bother monitoring it. As long as I can see my res and know the pump is working, I'm fine lol
thumb.gif
.
 
Discussion starter · #46 ·
Quote:
Originally Posted by Puck View Post

Very cool!

Flow sensor is a nice bonus, but not a deal breaker for me...and I definitely wouldnt want one at the expense of a temp probe.

I'm sure I'm in the minoryt but I run dual pumps on anything with a TEC after a close call(and recommend the same to everyone!), so never have flow issues or even bother monitoring it. As long as I can see my res and know the pump is working, I'm fine lol
thumb.gif
.
Yeah, I'm liking that EK dual serial D5 arrangement myself. I've got an RD-30 that I may run but there's safety in the dual D5 Strong arrangement. I've also got an RD-20 but I'm not sure I'd ever use it. It's for testing purposes I suppose.
 
Yeah I run dual DDC pumps, I do monitor flow but it is mainly for curiosity, I do have a shut down alarm configured if flow drops below a certain point though, I monitor pump RPM as my main reference, but the flow rate gives me the option to shutdown if a tube blows off or splits etc. I use mayhems pastel coolant so I can also monitor if anything is clogging up etc. I'd say a flow monitor is optional but handy. However I easily use 8 temp sensors so that is something to consider, also mostly a curiosity thing though.

I find the thermistor that puck and I use for TEC control to be more than accurate enough, but we don't care about dew point
tongue.gif
insulate FTW
 
Discussion starter · #48 ·
Quote:
Originally Posted by LiamG6 View Post

I find the thermistor that puck and I use for TEC control to be more than accurate enough, but we don't care about dew point
tongue.gif
insulate FTW
It's tough to get that sensor to stay in the dynamic range of the ADC I'm using without a lot of gymnastics due to it's high Beta (25/85 of 3990 I think). Something with a B0/100 of 3450 would be flatter. Those are pretty expensive from Epcos though ... and Uber tiny. Half the diameter of the Puck/Foxrena sensor. They look about the same, glass encapsulated, just smaller.

What are you using to read the thermistor? AQ6 maybe? Some lab amplifier?
 
Actually I use that thermistor with my PWM TEC controller, I am not sure of the resolution of the controller, I think it is accurate to 0.1*c. I use basic 10k NTC flat type sensors with my aquaero AQ5LT just for monitoring but I have offset them to all have the same start point at room temp, I don't need the monitoring ones to be very accurate, just the control one to hold my cold plate at my set temp. My cold plate temp only fluctuates ±0.1*c

I think if you can get all this accuracy and resolution without much extra cost it is good to have but if it greatly increases costs I think it isn't necessary. Certainly is the bees knees of controllers though.
 
Discussion starter · #50 ·
It adds about $5 in parts approximately. Maybe a bit more since I'm using the ultra-tiny versions of some of this stuff to fit it all.

I was just excited about the prospect of getting a temp sensor with enough accuracy to measure the small incremental change on either side of the loop. With a decent flow meter, we could then see what kind of wattage we were adding at the TEC hot side or expelling at the radiators.

Where is your probe on the block? On the cold side plate or lower down on the pedestal that interfaces with the IHS?
 
that would be very cool to see what kind of wattage is moving through the system.

I have the control probe drilled into the edge of the cold plate, say about 2mm from the surface that the TEC touches, then I have my monitoring probe attached to the side of the pedestal that steps down to the IHS. Ideally I need a second high accuracy probe drilled into the surface 2mm below the cold plate and a third one drilled into the side of the pedestal step and monitor that in my aquaero to get a better idea of exactly what is going on, but for my needs this is more than accurate enough as is because the controller can quickly and accurately react to load changes and the monitoring stuff I rarely ever look at any more.
 
Not at all like the one Ryan uses, it is a tiny little PWM controller meant for TEC cooling laser's, optics etc. Very good quality, very stable, quite cheap, up to Vcc 25v and amp limit depends on adding a heat sink and fan and soldering wires on, not sure what the traces can handle but they cope fine with the 20 amps I'm throwing at it. The only thing that bothers me is the mosfet could break off because I can't anchor the heat sink to the PCB so have to be careful not to bump it or bend it too much.

http://www.ebay.com.au/itm/TEC-controller-for-Peltier-modules-with-PID-7-25V-max-20A-cooling-mode-/231704893134?epid=1883350912&hash=item35f2b016ce:g:fJQAAOSwiLdWA-wf
 
Discussion starter · #54 ·
Just did some math for looking at the wattage added / exhausted. The specific heat of water is ridiculously high which brings the deltaT down so low. If you're moving 1GPM over a water block (or exhausting out of a radiator) you will move 264 watts for a one degree Celcius deltaT. That differential will be too hard to measure accurately enough to make sense. Even if you could get +/-2% on the flow sensor you would have to be able to measure really tiny delta's on these NTC probes.

I only have a 12-bit ADC on the device I'm using. I suppose it's possible if one takes the time to calibrate the differential between two probes and only measures the differential in a narrow band.

I'll try to put one flow sensor port on this controller to at least be able to make the attempt. I think it would be nice to be able to measure the rejection of heat on the hot side (out of the radiators) on a direct die setup.
 
Discussion starter · #55 ·
Well, I think I'm going to sacrifice the accuracy of readings above 65degC to gain resolution in the mid-band from 20 - 65. That will allow the finer differential detail to come through in the area where most people's hot side water loops run. Then I should be able to generate a decent number to go toward the "Watts Rejected" measure.
 
Discussion starter · #56 ·
DHL dropped a box off today. Four sets of boards were in there ...



The skinny tall one to the lower right is the IQTEM regulator board ... CAD shot previously posted.

To the left is a Water Loop Commissioning and Maintenance Panel for running your water pump from an external supply without unplugging the pump.



Above that on the left is a double stack of Illudium Q-24 D5 Strong Modulator boards. The goal there is to take in 12V on a Molex and put out from 8V to 24V to run a D5 Strong or an RD-30 to their full capacity. That board is limited to 75 watts out and uses a fan header for control of the output.



Then to the right of that is a stack of the first build of Illudium Q-15 Fan Over Drive boards. These are limited to 100 watts and have an MCU, 2 temp sensors, a PWM fan input port, a USB port, and a connector on the back for a 128x96 1.2" full color OLED display. If they run efficiently enough, I'll bump them to 120 watts.



That, my friends, is a soul crushing number of surface mount components to place. Especially the 0402 parts on the IQFOD.
 
Discussion starter · #59 ·
It's the tweezing and placing through a magnifier that wears me out. They get soldered in the oven. Basically, you squeegee solder paste through a stainless steel stencil which results in the proper amount and shape of paste being put down on each pad. Then you place each item with tweezers onto its landing pads. Then you pop it into a reflow oven which activates the flux and then liquifies the solder. Once it cools, everything is soldered into place. Then you do the few through-hole parts by hand.

Several people are waiting on the IQ-24 D5 Strong Modulator so I will do a set of those first. Then the "Under Water" board and then the regulator for the IQTEM.
 
41 - 60 of 246 Posts