What's a bored man to do on a sick day? Design a ridiculously cheap dummy load and order it from JLCPCB of course! Since the minimum order quantity is 5 boards I made them 2 ohms each for flexibility. Each board has 176 2W/22ohm resistors, arranged so as to make a 2ohm load in total. The BOM cost is $5 per board, so this is way cheaper than using wire wound power resistors, and has fewer capacitance and inductance issues I think? For cooling I'll probably put some aluminium strip on top of the resistors and point a fan at the boards.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Silly surface mount dummy load module
- Thread starter mcdn
- Start date
Old_School_Brad
Major Contributor
- Joined
- Sep 10, 2024
- Messages
- 1,453
- Likes
- 1,890
That looks like a great idea. But you'll probably be dealing with some serious heat on the PCB itself and some hot-spots on the surrounded resistors if you do not mount some sort of cooling plate that has contact with all of them. Which in itself may be problematic due to differing heights of the resistors.
Mounting the PCB in an air duct with a fan in the end might be the better solution.
Mounting the PCB in an air duct with a fan in the end might be the better solution.
ReDFoX
Member
SMDs have only 2 heat transfer contacts, nothing in between to actually transfer heat to PCB + spacing would be crucial here, since they're cramped so tight together...
I would add at least 1 hole in the middle to prevent PCB flex when you'll be tightening it to a heatsink. Something like high-power LED cooling solutions will give best results, IMO
solderdude
Grand Contributor
Perhaps a slab of silicone heat sink material (which can compress a little) and a heatsink on top could do the trick.
Then again that would require some holes in the PCB that would allow the heatsink to be mounted.
Another method could be to use lots of vias and a continuous bottom plane where to mount a heatsink on.
Again requires holes and heatsink compound. Might be difficult to solder the resistors onto as heat from the soldering iron will dissipate.
Also capacitance will increase.
350W is a lot of heat.
Then again that would require some holes in the PCB that would allow the heatsink to be mounted.
Another method could be to use lots of vias and a continuous bottom plane where to mount a heatsink on.
Again requires holes and heatsink compound. Might be difficult to solder the resistors onto as heat from the soldering iron will dissipate.
Also capacitance will increase.
350W is a lot of heat.
- Thread Starter
- #5
For $100 shipped for 5 boards this was mainly for amusement not serious, but yes, heat sinking would be important. These are all solved problems in the computer world at 250W/cm^2, so this application would be trivial.
In practice you'd be looking at 2 boards per amp load for a 4 ohm total, so each board only sinks 175W. Some aluminium angle pieces can be tacked on with some thermal compound or silicone material. That should give enough thermal mass for short term power measurements without a fan. For long term measurements you'd need a bigger heatsink and a decent fan.
I considered an exposed bottom thermal plate but the heat is generated in the resistors themselves so attaching to that side seems better. I also considered copper board instead of FR4 - It would have added $10 per board to the price and it's only available single sided, and I didn't like the look of the traces when done single sided.
@solderdude there's no soldering required, this is shipped fully assembled from JLC.
In practice you'd be looking at 2 boards per amp load for a 4 ohm total, so each board only sinks 175W. Some aluminium angle pieces can be tacked on with some thermal compound or silicone material. That should give enough thermal mass for short term power measurements without a fan. For long term measurements you'd need a bigger heatsink and a decent fan.
I considered an exposed bottom thermal plate but the heat is generated in the resistors themselves so attaching to that side seems better. I also considered copper board instead of FR4 - It would have added $10 per board to the price and it's only available single sided, and I didn't like the look of the traces when done single sided.
@solderdude there's no soldering required, this is shipped fully assembled from JLC.
Last edited:
solderdude
Grand Contributor
That would help.
- Thread Starter
- #7
So long as you stick to the LCSC parts list, it's astoundingly cheap and easy.
solderdude
Grand Contributor
Looks like a fun project... too bad I don't have any use for it. 
- Thread Starter
- #10
Yeah, throwing some aluminium sections on top would be the obvious solution. They are about $5/meter and one meter would do the job for 4 boards. The resistors are assembled on the board at the factory so will be very even in height. Hand-waving, I think that will be good enough for power measurements lasting 1 second or less. For longer term loads a fan would be required, and maybe a bigger heatsink in the form of some aluminium block.
The hand waving is because I am sick and can't be bothered with doing the math, but the math for heatsinks isn't very hard, and boils down to small heatsink+forced air is 10x better than big heatsink without forced air. Small fans are cheap, so just put one in there.
Last edited:
Cool idea!
What I would do, just to see what happens, is to leave the back side with various or a single huge pad (occupying most of the board) with bare copper and a lot of bias to a top ground plane with the resistors as separate as possible (price at jlcpcb doesn't start to increase until 100x100 mm) and solder the PCB to a large heatsink (needs to be done in an oven, obviously, maybe with the heatsink preheated). I believe you should go for 2oz copper anyways because of the traces that carry the signals. Not sure how the expansion will play, but if problematic, one could use smaller heatsinks instead of one big one. The resulting brick + a fan should take quite some heat.
Also, you could tune the total resistance with some jumpers placed strategically.
What I would do, just to see what happens, is to leave the back side with various or a single huge pad (occupying most of the board) with bare copper and a lot of bias to a top ground plane with the resistors as separate as possible (price at jlcpcb doesn't start to increase until 100x100 mm) and solder the PCB to a large heatsink (needs to be done in an oven, obviously, maybe with the heatsink preheated). I believe you should go for 2oz copper anyways because of the traces that carry the signals. Not sure how the expansion will play, but if problematic, one could use smaller heatsinks instead of one big one. The resulting brick + a fan should take quite some heat.
Also, you could tune the total resistance with some jumpers placed strategically.
- Thread Starter
- #14
Ordered, they will ship in the next 48 hours so should have them in my hands by Monday next week.
- Thread Starter
- #15
Nice ideas but too late, already ordered! Although I did spec 2oz copper - it was needed to carry the current with a 5mm trace width, and that put me into the next price tier regardless of PCB size. Also I think the right move is to attach the heatsink to the top of the resistors rather than trying to transfer all the thermal load to the bottom layer. The heat is surely better transferred directly off of the components rather than through vias etc?
you might be right, i would be afraid of accidentally shorting one. here below you can find very interesting information about power dissipation in smd resistors.
https://www.vishay.com/docs/53048/pprachp.pdf
Last edited:
That would need to be mounted on a heatsink if used for continuous operation.
- Thread Starter
- #18
Thanks @MCH for the link, it led me down an interesting Sunday morning rabbit hole!
Clearly some form of active cooling is required for operation for more than a few seconds at high load. To simplify things, we can consider 100W per board and a max temperature rise of 50C (e.g. operation at 70C in 20C ambient air). This is a 1% change in resistance based on the 200ppm/K spec. The airflow required for that is ~6m^3/hr, so 24m^3/hr for four boards at 400W total. Any 80MM PC case fan can easily deliver this, as they usually do 50m^3/hr.
So, getting the heat away using forced air convection will work. But we still need to get it from the board to the air. This paper gives a value of 100-200W/m^2K for the heat transfer coefficient of the top of a surface mount power device under forced air, and we can probably assume this for the board as a whole. Our board is ~0.01m^2 per side, K is 50, so in principle we can indeed move 100W from the board to the air without a heatsink thanks to its quite large surface area. Whether this will actually allow the chip junction temperatures and other hotspots to be cool enough requires harder maths.
Adding a heatsink would cool the chip tops directly much better, which this other Vishay paper shows is important, as the thermal conductivity of the resistor material is low compared with the junctions and even the FR4 board material. On the other hand it would restrict airflow to the board top surface, so is possibly more useful for use without a fan.
A version 2 of the board would therefore have:
Clearly some form of active cooling is required for operation for more than a few seconds at high load. To simplify things, we can consider 100W per board and a max temperature rise of 50C (e.g. operation at 70C in 20C ambient air). This is a 1% change in resistance based on the 200ppm/K spec. The airflow required for that is ~6m^3/hr, so 24m^3/hr for four boards at 400W total. Any 80MM PC case fan can easily deliver this, as they usually do 50m^3/hr.
So, getting the heat away using forced air convection will work. But we still need to get it from the board to the air. This paper gives a value of 100-200W/m^2K for the heat transfer coefficient of the top of a surface mount power device under forced air, and we can probably assume this for the board as a whole. Our board is ~0.01m^2 per side, K is 50, so in principle we can indeed move 100W from the board to the air without a heatsink thanks to its quite large surface area. Whether this will actually allow the chip junction temperatures and other hotspots to be cool enough requires harder maths.
Adding a heatsink would cool the chip tops directly much better, which this other Vishay paper shows is important, as the thermal conductivity of the resistor material is low compared with the junctions and even the FR4 board material. On the other hand it would restrict airflow to the board top surface, so is possibly more useful for use without a fan.
A version 2 of the board would therefore have:
- Thermal vias under each chip
- 4 layers of copper
- Use of layers 2 and 3 to do the parallel connections between banks
- Exposed copper on the bottom (electrically unconnected)
- Consideration of cheap heatsink sizes in the layout including mounting holes
Gruesome
Active Member
Instead of mounting a heat sink, I would just increase the spacing and make the board larger. But you seem to have a good idea of the air flow needed, and these are not semiconductors, so until you reach the solder reflow temp you should be ok.
- Thread Starter
- #20
You are right, but... The papers linked above are clear that junction temps must be kept well below 125C to avoid board delamination and solder crystallisation, which is much lower than the reflow temp. Device temps should preferably stay below 70C to avoid derating. Since junction temps are always lower than device temps the device temps set the limit.
What I'm hoping is that reasonable airflow might be enough to allow say 100W/board continuous power without a heatsink in this hacky version 1. With improvements in version 2 I'd be pretty confident in doubling whatever v1 does.