Hi,
I have a system with a first step-down stage (MAX1627) which generates 5.9V for the TPS650532 which then generates all necessary voltages for the OMAP and peripherals. DCDC1 gives 1.2V core voltage, DCDC2 gives 1.8V for RTC, USB and a CPLD; LDO1 gives 3.3V for SDRAMs etc. I've followed the instructions given in data sheets for component selection. 2.2uH output inductors and 22uF low ESR input/output capacitors. Are smaller caps crucial at the outputs? I haven't got any.
My problem is, the system seems to draw quite a lot of current and has bad interferences from the switchers. Actually I think that the current has increased after first power up. Also, I can see decreased resistance between ground and the voltages which is a bit alerting.
First I had a ferrite bead between power ground and system ground. Can this cause voltage spikes that cause interferences, or even breaking the power chip? There shouldn't be any design flaws in the system to be powered because I have used it with lab power supplies for a long time and everything works fine. So it's just the power domain that's acting up.
What could be a normal current at startup with 6V input? I've set the supply power limiter at 200mA and it's being hit which seems much higher than expected.
Best Regards,
Harri
Each ground should be tied together at a single point and with no impedance between them. Your ferrite bead may have several Ohms of DCR, which is far too much to have between ground connections. Depending on how much current each ground carries, these Ohms of resistance can equal Volts of voltage which could easily exceed maximum ratings of the ICs. Do you have a schematic you could attach?
For input current, measure how much current the board draws on each of its rails. Then, see how much power this is at the 6V input. 200 mA is just 1.2W of input power, which is fairly low I would think. It shouldn't hurt to set your input current limit higher, unless something is shorted on the board.
Oh well. So I just might have damaged the processor with this simple mistake of putting the bead in? Or just the power chip? I did replace the bead with a jumper (that's the only point connecting to system GND) after realizing it might be a bad thing but of course the damage may have occurred on the first power up...
The rails don't draw much, something like 30-40mA from 1.2V and 3.3V on normal power up when there's nothing running on the processor. The 1.8V rail draws even less than that. That said, 200mA from 6V is way more which is sad news. I don't think there's anything shorted on the board although I haven't used this particular board at all before (the lab supplies are used with another one). Before powering up, I measured the resistances between every rail and GND and they were ok.
The schema is quite big and I'm afraid I'm not allowed to show it because it's part of a big project I'm working on. Maybe I could explain the parts you're interested in? Please ask what you want to know and I'll try to give the details.
It depends on what you had connected to each ground. The current going into, say, the processor core has to return to the source--input ground. If this ground is separated by resistance from the processor ground, then there will be a voltage drop across that resistance. This, added to the 1.2V, could overvoltage that node.
To find out what is shorted, turn on the power and see (with a thermal camera or by touch) what gets hot.
The bead is from MURATA and it's specs are
Nothing feels hot but the resistance from 1.2V and 3.3V to GND is only several hundreds of ohms which most probably means damage. That's really bad because there's little possibility to replace the processor, don't you think? I can remove the power chip to see if the resistances get back to normal but from what you said, I can be pretty sure it's the processor that's damaged. I don't know where I got the idea of putting a bead there in the first place, probably thinking it could suppress spikes. And in this case, it did just the opposite.
Do you think there's even the slightest possibility that the processor could stand the overvoltages caused by that bead?
An over voltage of the processor is just one possibility. The power ICs might be over voltaged instead/also. You need to debug the board and see what is damaged. One way to do this is pumping a lot of power into the board (increase your current limit) and seeing what gets hot.
Ok, I'll debug it a little more. I'm not going as radical as experimental overheating yet because I just might save some of the components on the board. I'll get back when I know more. Thanks a lot for your help so far!
Well, I just might have had some luck with this one. I removed the TPS650532 chip and it looks like it was the only damaged part. Feeding this board with lab supplies works perfectly! It consumes 100mA/3.3V and 140mA/1.2V when I run the RTC test for example. So I'm sort of back at square one but still have some hope. The power domain definitely is a problem and the question now is, is there anything how I could make sure that the bead wasn't the only culprit. Before I fire it up with the onboard power again, that is.
Any clues?
One good thing to try would be testing the power (TPS650532) without the processor. An easy way to do this is order the TPS65053 EVM and then test it with your components/values.
Yes, that could be a good idea. On the other hand, I've managed to test my power far enough to see that all voltages are spot on. I don't think experimenting with the EVM would shed any light to why I'm having spikes and interferences. Did you mean connecting the EVM to my board just like I'm doing with lab supplies now? One useful thing with a test like that would indeed be to see if I'd have cleaner power lines just because the switcher portion is a bit farther from the system board GND/VCC planes. I'll see if I can get an EVM quickly enough to do that test.
There's one clearly suspicious thing about the board. You see, our PCB guy didn't follow the design instructions and there isn't a completely separate ground plane under the power domain. The power domain is placed to the corner of the PCB but the grounds are (thick) stripes going around the switchers and there's even the system ground plane underneath. Do you think this will cause interference problems that are impossible to cure?
Before the current started to rise and I had to switch off, I remember seeing high voltage spikes on the power rails which exactly followed the primary DCDC-converter output FET pumping. I'm very concerned about this kind of crosstalk but the ferrite bead was present at that time so it could cause some if not all of it.
The EVM would allow you to use a known good layout to power your board. Once the EVM works, you could then replace its passives (inductor, caps, etc.) with yours and verify that that works. It allows you to debug one thing at a time--layout, then components, etc.
If you are having crosstalk like issues, then most likely it is the board layout.