Hi again all.
I'm working on my first PFC and have finally got it to work (somewhat). The design is supposed to take input 85-265Vac and output 380Vdc. After much tinkering and trying understand how it works, I got it to turn on without blowing a fuse or a FET. My avg. voltage across my caps is about 390Vdc right now, but there are a lot of high frequency voltage spikes, sometimes registering on the voltmeter at over its limit (1000V
). On the o-scope, I see ~400Vpp max (ie. 600 V on cap during spike)
I zoomed in on the o-scope and found regular noise that lasts about 100ns at a time, and it's constantly happening, I'm guessing it's when the FET fires?
It freaks me out and I only have the thing on for a few seconds at a time, and I have a fuse and circuit breaker on the input. The time I left it on for about a minute, the inductor got up to 35C and the boost diode got up to 40C.
Since it's my first time, I don't know what's normal. I've looked around a lot and read a lot of application notes and guides, but haven't seen exactly what to expect. Do you guys see what's wrong with this? Or maybe that's normal? My caps are rated at 450V with surge rating of 500V :|
Here's the schematic. I haven't implemented the EMI portion yet.
I'm working on my first PFC and have finally got it to work (somewhat). The design is supposed to take input 85-265Vac and output 380Vdc. After much tinkering and trying understand how it works, I got it to turn on without blowing a fuse or a FET. My avg. voltage across my caps is about 390Vdc right now, but there are a lot of high frequency voltage spikes, sometimes registering on the voltmeter at over its limit (1000V
). On the o-scope, I see ~400Vpp max (ie. 600 V on cap during spike)I zoomed in on the o-scope and found regular noise that lasts about 100ns at a time, and it's constantly happening, I'm guessing it's when the FET fires?
It freaks me out and I only have the thing on for a few seconds at a time, and I have a fuse and circuit breaker on the input. The time I left it on for about a minute, the inductor got up to 35C and the boost diode got up to 40C.
Since it's my first time, I don't know what's normal. I've looked around a lot and read a lot of application notes and guides, but haven't seen exactly what to expect. Do you guys see what's wrong with this? Or maybe that's normal? My caps are rated at 450V with surge rating of 500V :|
Here's the schematic. I haven't implemented the EMI portion yet.
Yeah, it's continuous mode (although I'm not 100% sure what that means, that's what the IC datasheet says). Does the inductance of the boost inductor matter a lot in relation to the spikes? The same design aiming for 1/3 lower final wattage (300W) had 10x the inductance by my calculations - which didn't make a lot of sense to me :|
I'll watch the FET gates and see if that is related to the noise. I'm trying to learn about snubbers for my SMPS too, so once I understand how to design them, I'll put the on here...
I'll watch the FET gates and see if that is related to the noise. I'm trying to learn about snubbers for my SMPS too, so once I understand how to design them, I'll put the on here...
If you have chosen that diode, you have to go a long way before you will be able to succeed. There are still many things to learn.
btw: The spikes that you see are just the huge electromagnetic interference resulting from hard switching with such a diode, the absence of EMI filtering and probably a bad layout.
btw: The spikes that you see are just the huge electromagnetic interference resulting from hard switching with such a diode, the absence of EMI filtering and probably a bad layout.
Eva said:
I know
That's why I'm here...Plus, I didn't "choose" this diode per se
just the only one I had available with a high enough voltage rating to start testing. I have ordered a couple different ones (STPSC1006D, STPSC806D), but haven't put them into the circuit yet because they're expensive and my calculations show that I need ~15A rating, so I'd need 2+ in parallel...I don't know yet the additional things to consider when putting diodes in parallel.It's good to know that one source of the problem is the diode
Easy to fix, eh?A couple questions:
-Does the EMI filter setup look correct?
-Do these diodes (STPSC1006D, STPSC806D) look like they'll work? I found a couple others with similar characteristics (C3D08060A, C3D10060A), but with a smaller leakage current...does that have an effect in this application?
-Board layout - still using a perf board, but will hopefully be doing a PCB this week!
Any general suggestions welcome.
Great, these are Silicon Carbide diodes, they don't store any reverse recovery charge although they are expensive (recently introduced on the market). Using two in parallel is a good idea to be in the safe side, although one could probably do the job because duty cycle is low when current is high (120V input) and higher when current is lower (230V input). If peak current is always below 40A, do the RMS current calculation to know if it's below 18A (for STPSC1006D).
There are also much lower cost hyperfast diodes like 15ETX06 that store little charge at the expense of a higher forward voltage, and there are conventional ultrafast diodes like MUR1560 that store 1uC to 3uC typicallty.
For layout you have to visualize the loops in which HF current flows and minimize their area. A continuous ground plane over these loops also helps, it acts as a shorted turn for PCB track inductance.
For EMI filtering consider that output capacitors have some series inductance (1nH per mm of terminal spacing) and some ESR, and thus, there is going to be some output ripple that may require filtering depending on the kind of load.
Input inductor also has some capacitance, and input capacitor has some inductance. This means that some HF residual from switching waveform leaks through it to the input side, requiring additional filtering.
Also, output ground and input wires are not exactly at the same RF potential, some RF ripple is created across the PCB and the components. A common-mode filter fixes that by allowing one side to float at RF and coupling the other side to a 0V RF reference.
btw: Leakage current shouldn't be a problem as long as it does not produce substantial dissipation.
There are also much lower cost hyperfast diodes like 15ETX06 that store little charge at the expense of a higher forward voltage, and there are conventional ultrafast diodes like MUR1560 that store 1uC to 3uC typicallty.
For layout you have to visualize the loops in which HF current flows and minimize their area. A continuous ground plane over these loops also helps, it acts as a shorted turn for PCB track inductance.
For EMI filtering consider that output capacitors have some series inductance (1nH per mm of terminal spacing) and some ESR, and thus, there is going to be some output ripple that may require filtering depending on the kind of load.
Input inductor also has some capacitance, and input capacitor has some inductance. This means that some HF residual from switching waveform leaks through it to the input side, requiring additional filtering.
Also, output ground and input wires are not exactly at the same RF potential, some RF ripple is created across the PCB and the components. A common-mode filter fixes that by allowing one side to float at RF and coupling the other side to a 0V RF reference.
btw: Leakage current shouldn't be a problem as long as it does not produce substantial dissipation.
darkfenriz said:
Yes, but diode current is also flowing through the switch during reverse recovery time, so it's not an accurate way to predict switching losses. Actually, the current that was flowing through the diode has to stop flowing and start flowing through the switch at a di/dt rate, then reverse current starts to rise at a similar rate until the diode stops conducting. At higher diode currents, forcing a faster recovery usually results in less losses even if it's at the expense of a higher reverse peak current. Note that peak switch current is Id plus Irr.
Switch dissipation during the whole process is high, something like P=fsw*Vcc*.5*(Id+Irr)*Trr. The .5 factor accounts for the triangular current waveform. "Trr" has to include positive current fall time plus negative current build up time.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.