I previously had a need for 12V to 33V boost converter, and to some extent I still have. Previously the problem was that while I had bought one for only 10 €, it wasn’t capable of what I needed it to be.
So now I went back to design and looked into using proper components and proper inductors and was amazed by how small inductors one can use when using high frequencies.
For instance this little Texas Instruments device can switch at up to 1MHz frequency and when you go through the calculations, first the duty cycle for 12V to 33V comes in at 66% and then if you type in the 500kHz switching frequency you get 170nH for minimum inductor size. That is ridiculously small both in inductance and in physical size and also in price.lm3478
For example this small Bourns inductor is 220nH and costs $1.22 in 10 quantity. And can handle 53 amperes of current.SRP1265A
And there are of course these larger inductors like the one as featured image, but that’s a $10 inductor. You could literally fill the PC board with these tiny ones.
So I began going through those calculations for different components (current limit, slope compensation) and at first I got negative values which made no sense. I was about to give up and deem this Texas Instruments part somehow, if not defective, then at least somehow unfitting.
But then I realized that since I used the formulas correctly, the only possible explanation must be that this part is not fit for what I was asking it to do. And then I realized that what if I changed some parameters (components) in those formulas and see what happens.
And that was the key.
The inductor I used (480nH) could not do what I was asking it to do, with the frequencies I was using it, and by changing the value of the inductor to a larger one, the calculations then became sensible.
And it maybe should have been obvious that 480nH inductor perhaps cannot boost 12V to 60V and 20A at 100kHz but to me it was these formulas which demonstrated it. So the part is perfectly viable for this purpose as long as components related to it are chosen to meet the requirements and other parameters.
And the 480nH can still be used if the switching frequency is raised from 100kHz to 1MHz, which is the maximum of this part.
And I believe 1MHz isn’t too much to ask from any ordinary MOSFET. It should be able to handle those frequencies if enough care and thought is put into laying the board and placing the parts.
Perhaps better option.snva168e
Looking into this and one very novel idea camo across: using MOSFET on-resistance as a shunt resistance. But because not too many (any?) manufacturer provide any information how accurate their measured on-resistances are, and because those vary with the temperature and current and Vgs and everything else it would require careful characterization of the system.
But that put aside, here is a great answer how to measure switched currents: https://electronics.stackexchange.com/questions/99185/measuring-average-current-with-uc-and-current-shunt
Also found some papers on it but it was all hardware based and that adds cost and uses PCB real estate which is limited so this would be something to try. It is not necessarily required for this setup but it wouldn’t hurt if it worked.
You understand the problem correctly: you need to get the “average” of the PWM, just like the meter you’re using for measurements.
You could use an RC filter on the A1,2,3 signals whose time constant is at least ten times your PWM period. That means if your PWM period was 10 microseconds then the RC time constant should be 100 microseconds. For example 10kOhms x 10nF = 100us
A better solution is to filter the signals digitally in the microcontroller ..
I don’t know why it is or if it is, but at least it allows more accurate measurement because the signal isn’t changed in any way, but this is good to hear because it means no need for extra hardware.
Another version with huge pin header to daisy chain these in SPI master-slave configuration.
See the via stitching. That is high current power plane so it has been stitched to increase the amount of copper. That copper will transfer the heat from the devices to the PCB copper and will then radiate it away. There is top and bottom copper on that plane and those are connected through this extensive stitching so the amount of copper is quite significant.
Here’s a good look at one of the power planes.
That is double sided (top layer hidden) and it will be stitched like the one already stitched so it will be capable of carrying huge amount of current. That trace alone is 200 mils and with double sided stitched it will be over 400 mils. Should be able to carry 20-30A amps.
Without taking vias into account 170mils (double sided 85mil) can carry 11 amps while rising 30°C above ambient.
Board was upgraded and soldermask was removed so that once those exposed traces are tinned, they are practically solid metal and should be able handle the absolute expected maximum of 32 amperes.
The outline is still bit thin, at about 110 mils per side for total of 220 mils plus the tinning and filled vias which should increase the effective width to 400+ mils. There are spots which are double that, so the overall resistance should be lower than that of 400 mils, and the heat should then spread around if parts experience more resistance than other parts.
Back side with some component placement showing.
In this version more soldermask was removed and some were modified slightly.
Received this heatsink today:
Initially figured it could house three 100W LEDs but not sure if it can dissipate that much heat. The size is almost exactly double what my current heatsink for single 100W LED is, and that heatsink could take little bit more. So with powerful fans this could cool 300W of LED or perhaps 250W.
Amazing how these old UPS can be reconditioned and reconfigured to do the work of 1000 € devices:
Check the whole series. Real hardware hacking.
Also the guy knows a lot about inverters and a lot of that seems to apply to UPS as well, and just like cheap inverters; cheap UPS too provide modified sine wave which can even be damaging to sensitive electronics and motors.
So if I ever want to get an UPS I will make sure to get one of these rack/professional units which are of higher quality.
To be considered for the LED driver board perhaps as a pre-boost stage.10.1.1.402.13
Linear Technology manufactures these extremely small devices they call μModule, which can deliver huge currents in small packages.
Downside is the price which for this particular model is about 30 €. So not exactly suitable for driving my 350W LED panel but whereas my barely-capable enough driver takes 20cm² this only takes 3cm² and delivers twice as much current. So great little devices.