-
Capacitor quick questions:(Dissipation factor vs tan delta vs ESR vs Q vs GAHH?)
Posted by Powermax on 21 Jun, 2016 05:38 -
I made a really neat flyback plasma speaker of my own design utilizing a LM339 quad comparator. One comparator makes up a oscilator, another one is used as a proper comparator, and the third one is part of a fan control circuit. I'm using a FDP33N25 MOSFET for it's really good ratings compared to it's price. Very comparable to the classic IRFP250N, at less than half the cost. I am driving it really hard with a NPN PNP emitter follower totem poll config. I noticed that the MOSFET does periodically fail and analyzing it with the scope, I see these HUGE high voltage 400V (!) transient spikes!
A dissipative snubber would need to dissipate close to 90W just to reduce that to closer to a safe range. At first I could not believe I was dissipating that much power until I used a 100W 120V lamp as the resistor and it was glowing pretty bright! 
Luckily, I discovered adding a large capacitor across the drain and source effectively eliminated the overvoltage spikes and now the maximum peak voltage observed is only 150V.
It seems like the addition of the capacitor affected the switching speed of the MOSFET without affecting the switching losses within the MOSFET, and recycled that high frequency high power flyback energy into lower frequency lower voltage ringing which is more easily "digested" by the flyback transformer. However only a select few junk bin capacitors proved good enough for the job, most smaller film caps overheat, and on the scope, I can verify the MOSFET is switching on and off with a duty cycle of approx. 40%, but during the off time, the LC network between 2 or 3 old 0.47uF MMC film capacitors and my flyback transformer primary is some really strong damped ringing that maxes out at a peek of ~100V. I am unsure of the RMS current due to the non-sinusoidal nature of this waveform but it must be pretty high (given Xc and the high amplitude ringing), explaining why my 3, 1cm diameter 3cm long 0.47uF MMC film capacitors get pretty warm. The frequency of the damped ringing was measured with 3 0.47uF capacitors and was about 75KHz.
However, I also had these really tiny unknown 0.47uF capacitors, but they seem to be VERY good considering their size, most capacitors even 10 times bigger than them seem to want to bloat, smoke and die. with a 2S3P arrangement they still seem to overheat after several seconds to a minute or so but still pretty damn impressed they are handling that reasonably well. The overheating might simply be attributed to their low working voltage and no balancing resistors.
All my questions:
So I am looking to make this design into a kit for people to buy, but I want a decent small capacitor that is suitable for this, to save PCB realestate and cut production costs, but I am also looking to learn more about capacitors too.- So why do some film capacitor datasheets only give dissipation factor ratings at a few test frequencies (like 10KHz and 100KHz) while others give ESR ratings as well? Are the ratings used to determine the same things? Q factor seems to be simply the inverse of the dissipation factor or tan delta.
- Based on what I googled, I figured out dissipation factor is the ratio of ESR to the capacitive reactance. And since the frequency and capacitance is given it is not hard to calculate the ESR. So then why is this not conveniently already listed on the datasheet? Is it because the dissipation factor is what is directly measured? (like the total measured impedance of the capacitor at a certain test frequency)
- So far, every time I try to calculate ESR from tan delta at the 10KHz and 100KHz figures, I always find that the calculated ESR is LOWER at higher frequencies, even if tan delta gets worse. Am I doing something wrong or is that normal?
[li]Should I choose "self-healing" Polypropylene MKP capacitors or the (smaller) MLCC capacitors? What dissipation factor should I look for for?
[/li][/list]
Sorry this post is long, it is my first one. I have a video of the circuit working below. Also you can see the pictures on my old question over at instructables.
http://www.instructables.com/answers/Dissipation-factor-vs-ESR/[/list] -
- So far, every time I try to calculate ESR from tan delta at the 10KHz and 100KHz figures, I always find that the dissipation factor is LOWER at higher frequencies, even if tan delta gets worse. Am I doing something wrong or is that normal?
This sentence doesn't make any sense because tan delta is the same thing as dissipation factor. See: https://en.wikipedia.org/wiki/Dissipation_factor halfway down the page. -
This sentence doesn't make any sense because tan delta is the same thing as dissipation factor. See: https://en.wikipedia.org/wiki/Dissipation_factor halfway down the page.- So far, every time I try to calculate ESR from tan delta at the 10KHz and 100KHz figures, I always find that the dissipation factor is LOWER at higher frequencies, even if tan delta gets worse. Am I doing something wrong or is that normal?
Lol, I started to use the terms interchangeably because they are the same thing, and yeah I meant ESR, not dissipation factor. I'll edit the post to fix it. What I tried to say was that at higher frequencies, although the dissipation factor (tan delta) was a little higher, the frequency was an order of magnitude higher at that point and therefore the ESR was LOWER at higher frequencies instead of higher, as I would have expected. It just seems wrong! -
Below should be the schematic I am using. Notice C7 is the capacitor in question.
-
bump. (guess no one knows the answers to unlock the keys to the universe and flux capacitors?)
-
Using a capacitor across a hard switch like that is not a usual thing. I would expect it to cause high loss in the switch at turn on and high pulse current in the capacitor as you are effectively shorting it each switch cycle.
Have you tried an RC snubber? This is normally how ring and overshoot is handled.
-
Using a capacitor across a hard switch like that is not a usual thing. I would expect it to cause high loss in the switch at turn on and high pulse current in the capacitor as you are effectively shorting it each switch cycle.
Have you tried an RC snubber? This is normally how ring and overshoot is handled.
Yes, I used a 100uF capacitor with a 120V 150W lamp in series with it, and a power silicon diode (that looks like a bolt and is clearly meant for business) and under certain conditions I can get the bulb to glow half to FULL BRIGHTNESS, and STILL have excessive voltage spikes.
The capacitor actually seems to reduce the power dissipated in the switching transistor, whist increasing the thickness of the arcs drawn. However it seems like care needs to be taken, as it appears I am sort of operating in a class E ZVS mode, rather than a conventional flyback mode. I should note that I removed the spacers in my flyback transformer, effectively converting it into more of a normal transformer. -
Using a capacitor across a hard switch like that is not a usual thing. I would expect it to cause high loss in the switch at turn on and high pulse current in the capacitor as you are effectively shorting it each switch cycle.
Have you tried an RC snubber? This is normally how ring and overshoot is handled.
Yes, I used a 100uF capacitor with a 120V 150W lamp in series with it, and a power silicon diode (that looks like a bolt and is clearly meant for business) and under certain conditions I can get the bulb to glow half to FULL BRIGHTNESS, and STILL have excessive voltage spikes.
The capacitor actually seems to reduce the power dissipated in the switching transistor, whist increasing the thickness of the arcs drawn. However it seems like care needs to be taken, as it appears I am sort of operating in a class E ZVS mode, rather than a conventional flyback mode. I should note that I removed the spacers in my flyback transformer, effectively converting it into more of a normal transformer.
Class E operation is actually quite a strange oddity, and it is a configuration used in some tesla coils. Basically you setup a LC resonant circuit and you pulse a transistor at that resonant frequency, in such a way that the transistor is turning on when the oscillation reaches exactly 0V. It needs to be well tuned, but also damped correctly. If the LC oscillator is not damped enough then the oscillations can undershoot 0V (which will get clipped by the transistor anyway) and if it is overdamped, the oscillation never quite reaches zero volts, and power is lost in switching. You can see that if I finely adjust the frequency of the scope, i should be able to get the transistor switching when it reaches 0V exactly. So selecting the right capacitor is important. -
Yeah sounds like you accidentally created a soft switched circuit. Need to make sure the capacitor is the right size and has appropriate ripple current rating.
-
they give you ratings like that so you know you can probobly get away with using them in a op amp circuit or such. you need a VNA or similar machine and a amplifier if you want the details.
it is very useful to think in terms of energy/charge in addition to voltage and current in switching circuits. how much energy is being moved will sometimes tell you about what you want. -
Yeah sounds like you accidentally created a soft switched circuit. Need to make sure the capacitor is the right size and has appropriate ripple current rating.
Lol, yeah. I meant to start experimenting with the class E one to eliminate the need for a ridiculous size heatsink although I guess I got to that point a little early. However the problem is that the inductance of 3 turns on a core of the flyback does not have high inductance so a really big capacitor is needed. I am thinking about trying to use a motor run capacitor, I am sure the ESR has to be pretty good on those, as they are designed to handle significant current, but only at 60Hz, so the ESL may not be so great. -
Have you tried an RC snubber? This is normally how ring and overshoot is handled.
Yes, I used a 100uF capacitor with a 120V 150W lamp in series with it
I'm not surprised it doesn't work - the RC snubber needs to be quite low-inductance, so I guess a 120V lightbulb doesn't cut it. Use a small enough power resistor and short wiring. Actually calculate the power loss and choose the right capacitor value. Try something that dissipates only a watt or so; it can do quite a lot to prevent ringing. -
Have you tried an RC snubber? This is normally how ring and overshoot is handled.
Yes, I used a 100uF capacitor with a 120V 150W lamp in series with it
I'm not surprised it doesn't work - the RC snubber needs to be quite low-inductance, so I guess a 120V lightbulb doesn't cut it. Use a small enough power resistor and short wiring. Actually calculate the power loss and choose the right capacitor value. Try something that dissipates only a watt or so; it can do quite a lot to prevent ringing.
I used my largest 10W 10 ohm power resistor and it would overheat in seconds. I'm not sure how high inductance the lamp is, but that is why I added a bulk filter capacitor in series with it. I think I did figure out that some of those inductive spikes must be a result due to just the tiniest amount of stray inductance in the wire going from the drain to the primary, which the snubber was not snubbing. Guess the photoflash electrolytic capacitor used was also too high ESL for fast acting, but I think the majority may just have been the speed of that very large but old PN junction power diode, with unknown ratings. -
Also, why is it that I never get replies on this forum from the same ppl? or more specifically that they do not get notified of my replies, but why is that the default setting? Seems to be that way on many forums (notably the RCgroups forum as well), and it is annoying.
-
I used my largest 10W 10 ohm power resistor and it would overheat in seconds.
So you dissipated too much - but the resistor has nothing to do with it. You should really google for RC snubber. You can decide how much you want to dissipate, and you do it by calculating the capacitance.Quotebut that is why I added a bulk filter capacitor in series with it.
Any "bulk" cap in the simple RC snubber sounds horribly too big, and explains the high dissipation. Usually they are rated in nanofarad range. It depends on the frequency and voltage. Easiest way is to just decide some level of power (for example, 1W) you are ready to dissipate in the snubber and design for that. If it doesn't help, you need something else than just an RC snubber. -
So you dissipated too much - but the resistor has nothing to do with it. You should really google for RC snubber. You can decide how much you want to dissipate,
Any "bulk" cap in the simple RC snubber sounds horribly too big, and explains the high dissipation. Usually they are rated in nanofarad range. It depends on the frequency and voltage. Easiest way is to just decide some level of power (for example, 1W) you are ready to dissipate in the snubber and design for that. If it doesn't help, you need something else than just an RC snubber.
Yeah, I was trying to dissipate as much excessive flyback energy as required in the hope that it would have eliminated the voltage spikes, but it did not. I only reduced the voltage spikes by 50V or so, and still much over the rated voltage of the MOSFETs. I do need something else than a simple RC snubber. I was considering to use a TVS diode at one point, but then realizing how much energy the RC snubber was dissipating I decided that it would not have been practical. (I don't know where to get zener or TVS diodes in a TO-220 or better package) -
bump
-
Schematic, layout, and lots of pictures, please!
Tim