heat sink on a PC board

Do you pack your stuffs tighter than them? Each of these box is around 1000W.

These boards shouldn't be expected to move their own air - they should just provide low backpressure and suitable fins.

If your stuff is SMD, you could have fins one side, electronics the other.

Then you just use the catalog Rth for your fin shape and air volume.

CPU coolers aren't going to limbo under 1.75in, and all they do is scrub.

RL

A heatsink ON a PC board is less than useful, because the PC board doesn't shed heat very well; if you can arrange your cards to butt up against a REAL heatsink surface (obviously, opposite the grab-here-to-extract edge) then a screw or two will connect the hot parts to a chassis-mount sink.

Alternately, of course, you can move a few hot parts off the PC board entirely. This is as old a solution as the preamp/poweramp two-box stereo.

There's good reason to use enclosure fans AND a ducted airflow to ventilate the heatsink bits, but that requires you to guide that air, NOT let the host enclosure make the important decisions. Take a look at server computers' interiors, the airflow is a key to those rackmount enclosures, and not random 'use more fans and stir the air' bodgework.

Note here, there's a bank of seven small fans to make air go through the heatsinks

Rough estimate on that box is about 4 watts per cubic inch.

I'll have maybe 30 cubic inches available per module.

I worked with an electrical contractor in NY City. He found adjacent propeties where people drilled through walls to steal power from their neighbors, occasionally in both directions.

Most boards will be happy with the overall box air flow. But the higher the power rating of a dummy load board, the more sellable it is. So more fans might be worth it.

Are you trying to cool parts or just make an oven? What's the actual airflow though the enclosure?

Are you trying to be funny? I said it was a programmable dummy load module. They tend to dissipate power.

The rackmount box will have two fans, 60 cfm each, and that has to cool the kilowatt power supply, the control board, and up to 10 power modules. The switcher modules should be fine, but the load modules need all the cooling they can get. Cooler = better specs = more sales.

Some sort of ducting would help some, but that's a hassle to fabricate. A couple of local fans might work better.

An L-bracket down to the box bottom is a possibility too. We'd have to experiment to see how well that might work. It would make module installation a bit harder.

søndag den 6. juni 2021 kl. 00.08.39 UTC+2 skrev snipped-for-privacy@highlandsniptechnology.com:

:)

Just use non-smd real estate and air volume for more heatsink fins in the board plane.

Gives the system fans something to work with, not a squirrel's nest.

RL

No, I actually work equipment that dissipates lots of heat and requires cooling.

60CFM with no pressure, or measured?

So you want an oven, not a cooling system. What I'm getting at is cooling requires removal of heat, and not blowing it around some box with randomly placed corny fans.

If you just want to move heat a short distance, say to the back of a chassis, consider heat pipes. Noren is a good supplier. They also make a box that removes heat from enclosures without requiring an air inlet and outlet in the area being cooled.

In article snipped-for-privacy@4ax.com, snipped-for-privacy@highlandsniptechnology.com wrote: <SNIP>

Now you put met into imperial mode. How much btu is that again?

These days big electronic loads and battery cyclers usually feed power back into the AC mains, or into the DC bus that powers other units in the box, e.g. the load on the output of an inverter under test will feed power back into the power supply that provides the input to the unit under test, or the energy from flattening one battery goes into charging the next one.

e.g.

That was suggested, but the MeanWell bulk 48V power supply that we'll be using can't be back-fed into the AC line.

A dissipative isolated programmable load is nice and simple, but gets hot.

Yes. The one I posted the schematic of in another thread

was water-cooled. I was annoyed that nobody makes a compact water block designed for a single SOT-227B, so I had to make them myself. Machining and tig-welding pure copper is more annoying than I expected.

Is the device-under-test electrically powered, or where does it get its energy from ultimately? (Batteries? burning fuel?) If it is ultimately electrically powered then there may be no need to feed power into the mains, rather it could go into the DUT supply. You could use an off-the-shelf isolated DC-DC converter brick, design your "load" to dump its power into a power-zener, and put the input side of the DC-DC brick across this zener. Then somehow adjust the amount of load current the DC-DC brick feeds into the DUT input bus until your power-zener has negligible but non-zero current flowing in it.

Even if you can't feed the power into the DUT supply or the mains (because the DUT runs on burning fuel for example) you could still dump the power into an external resistive load, located somewhere more convenient.

My 1400 volt Pockels Cell driver is water cooled. I used an aluminum block and AlN insulators for the silicon carbide TO247 fets, assuming my customers would have a cold plate.

I can't get inside my users' various DUTs. One possible thing to load is an engine control computer; one is an alternator.

I could have a box-global load somewhere, just to allow my plugig cards to be dense and not have them get rid of heat themselves. That would sure complicate life.

My idea was pretty simple:

The FPGA will look at the voltage and current sensing isolated ADCs and play with the gate drive. We can do CV, CC, constant resistance, or constant power modes, AC or DC. We'll assume that we don't need to operate below maybe 1.5 volts, which is OK with my customer base; hence the bridge rectifier. I suppose an ideal (mosfet bridge) rectifier would be feasible. That would be an interesting circuit.

I'll ask my mechanical guy if maybe we could dump heat into the bottom of the rackmount box, which is a lot of aluminum. We'd have to experiment with a mockup to quantify that.

Is this the same module plug into the over-voltage-protection-less power supply? In this case, it would be simple to just insert an OV cutoff mosfet, driven by your FPGA fabric. I use ARM micro to do it, but the idea is the same. In either case, 3v driven mosfet would be ideal:

Diode DMN3020UTS TSSOP-8 14A Diode DMN2015UFDF DFN2020-6 14A Vishay SiA448DJ SC-70-8L 12A AlpOmg AO6404 TSOP6 8.6A Rohm RQ1C075UN TSMT8 7.5A ST STT5N2VH5 SOT23 5A

Are we talking about the mosfet bridge rectifier?

LT4320 is interesting but flawed.

Yes, you can integrate the logic into the bridge rectifier, or just another high side switch.

Easy enough to implement Forward, Backward or Off for the rectifier.

I think it needs four floating gate drivers (with n-fets) and some sensing stuff. The voltage ADC could sense the input polarity. High hassle level.

I could put two current limiters back-to-back with no rectifier, but that makes the thermals worse.