uting the 3 phases, maximizing the trace thickness
We have a total of about 40mohm trace resistance accross the entire PCB, wi th about 25A rms currents, which translates to approx. 20W loss. Some addit ional resistance is due to connectors, but that is out of my hands
I am running about 20mm wide trace on 3 layers. Each layer has 25A current from a phase.
[snip]
It is a combination of 70u on outer layers (to allow for fine pitch devices ) and 105u on inner layers (total of 4 layers). So a mix of 2 and 3 Oz.
I think the jumpers may be a good idea. Need to find something cheap for th at. On the other side, increasing to 6 layers avoids the use of leaded asse mbly, which can be the same price as the cost of 2 extra layer
Wrong again. Air flow or coolant flow is a major consideration. If t is still, convection cooling is near zero no matter what the fluid. Other considerations are the heatsink area, placement, temperature of environment,etc.
With all three phases mostly on top of each other? If there's top room then I like the bus bars. put three is parallel one reaching down to each phase. yeah you lose copper for the hole to the inner layers. plastic barrier strips between bus bars if you need them.
Natural convection is driven by density gradients inside the fluid.
Hotter fluid is - almost always * - less dense than cooler fluid, so you g et gravitationally driven natural convection. Clearly, if the coolant can circulate in a bigger volume, natural convention can be more effective.
There's a dimensionless number - the Rayleigh number
formatting link
which you can calculate as indicator or what might be going on.
"If it is still" begs the question of whether natural convention is signifi cant in a particular situation.
I know what water gets less dense as it is cooled from 4C to 0C. There ma y be other situations where this occurs, but they are rare.
Something is rotten in Denmark. According tp a trace calculator the resistance of the 20 mm wide, 105 um thick inner layer traces should be
4.41 mohm and dissipate 1.05 Watts each. The outer layer trace should be 2.47 mohms and power is 1.55 Watts. I don't see how you can get 20 Watts on the PCB. Of course the connector resistance will be more than this depending on how many contacts you use and how well they work.
You can't get much cheaper than wire. 14 gauge is inexpensive and 2.5 mohm. Or you can go with one run of 10 ga per phase and get 1 mohm over
300 mm.
I guess one dumb question would be why are you running this current through the board rather than on a wire? I assume this is essential to the function of the board?
Is the problem the heat on the board or the loss of voltage across the connections?
Before you worry any more about the PCB traces, I think you need to find out why a 4 mohm trace is showing up as 40 mohms. That is your problem.
I'm not a PCB expert, but I've seen some processes that include a step where additional copper is electroplated onto traces after a board is etched. I suppose one could prepare a mask to selectively plate only the few high current traces where this is required.
--
Paul Hovnanian mailto:Paul@Hovnanian.com
------------------------------------------------------------------
Programmers don't die, they just GOSUB without RETURN.
routing the 3 phases, maximizing the trace thickness
r
nd
gth
is
le
ne
r
e
, with about 25A rms currents, which translates to approx. 20W loss. Some a dditional resistance is due to connectors, but that is out of my hands
ent from a phase.
LOL. There is always something rotten in denmark, 68% tax, over half of the working people employed in the government sector, 200% tax on cars.... I c ould go on
As for the trace resistance. If I sum it up, it equates to 20mm wide 2oz tr ace with a length of 3000mm. That's 36mohm, with a loss of 22W at 25A curre nt
You keep changing the specs. In one post you said the traces were 30 mm wide, then they were 20. Every post I have read indicated the board was
200 x 300 mm and the trace was 300 mm long, now you say the trace is
3000 mm! That's 10 feet or the size of the room I am in!
I never know which are typos, but if you are trying to run 25 amps 3 meters across a PCB you will need to rethink your design.
One of the web sites I used to calculate the resistance of the wire did an analysis of the current and voltage in use. For 240 volt power at 25 amps over 300 mm it said I should use 24 gauge wire, lol. They are obviously just looking at the voltage drop relative to the voltage being conveyed.
r routing the 3 phases, maximizing the trace thickness
o
her
find
ength
r
s is
r
ngle
.
gine
per
n
the
rs
ly
CB, with about 25A rms currents, which translates to approx. 20W loss. Some additional resistance is due to connectors, but that is out of my hands
rrent from a phase.
e
n
the working people employed in the government sector, 200% tax on cars.... I could go on
z trace with a length of 3000mm. That's 36mohm, with a loss of 22W at 25A c urrent
Sorry if it is by any means unclear.
The traces are on a PCB. Some places the trace is 20mm, others 30mm, some b roken/reduced by keepouts from vias etc. But, summing up all the tracs leng ths, it comes to about 3000mm and average of 20-30mm. Sadly the constraints of placing the power module is on the other side of the mains input, so tr ace length are very very long
I would upload a themography picture, but I am not allowed to do that
The bottom line is that I need to reduce the trace resistance by some way, keeping costs low at the same time
It's working and has passed EMC tests with plenty of margin (we are running the phase traces on each layer to keep loop areas low). We are just a litt he high in temperature, since the max ambient temperature is 50 degrees, so we are riding just below 90 degrees on big surfaces of the PCB (components primarily rated for 125 degrees, and the requirement specification allowin g degraded lifetime at 50 degrees ambient)
Yes. We are using some spade connector for external DC inductance. We are r unning 25A through that, so we have two spade terminals in parallel, just t o comply to the derating rules even though the connector is quite cool
It is specified for 10 degreee temperature increase at rated current, but o ur derating rules specifies 60% derating of current in the connector)
routing the 3 phases, maximizing the trace thickness
d
th
s
e
e
with about 25A rms currents, which translates to approx. 20W loss. Some add itional resistance is due to connectors, but that is out of my hands
t from a phase.
e
o
es) and 105u on inner layers (total of 4 layers). So a mix of 2 and 3 Oz.
that. On the other side, increasing to 6 layers avoids the use of leaded as sembly, which can be the same price as the cost of 2 extra layer
20mm x .1mm traces have a 2mm^2 cross-section area. To cut dissipation by a factor of 3 you'd need 6mm^2 Cu cross-section, e.g., 10mm x.6mm.
That looks like either custom bus bars, adding heavy bare jumpers, or splitting the board into a specialized extreme copper board and a fine-pitc h board.
Laminated custom bus bars are electrically excellent and can carry all thre e phases. Minimal loop area.
Adding bare jumper wires is cheapest. 2 x #14AWG (1.6mm) has a 4mm^2 cross- section, enough to cut your losses by two-thirds.
3m of path length is tough mechanically, and a mighty expensive bus bar.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.