Snubbers

[dknguyen]

I have a question about snubbers. THe flyback diodes used in H-bridges, are those considered snubbers? I'm having a bit of trouble understanding why RC or RCD snubbers are still used even when flyback diodes are there.

THe things I have so far are that the RC snubbers can control rate of rise and damp ringing while RCD snubbers clamp and control rate of rise. Flyback diodes also clamp though, but they don't control the rate of rise. THe other thing I read about was that RC or RCD snubbers are used to clamp the spike produced by the battery-bridge inductance which the anti-parallel diodes across the bridge MOSFETs will not protect against.

Also, that really big capacitor that tends to go across the battery terminals- that is meant to try and nullify some of the battery-bridge wiring inductance right? Or to absorb the spike produced by the battery-bridge inductance when switching off the bridge? (though it might just be the same thing and only different because it is being viewed from two different perspectives). THis capacitor also makes me wonder why a RC or RCD snubber might still be needed if this giant capacitor does the job (or does it not do somethings that RC or RCD snubbers do)? Oh, and capacitor is needed to provide surge the high frequency surge current needed when the bridge is being switched on which a battery can't do.

I'd hate to have the PCB made and then realize I need extra snubbers (aside from the flyback diodes and giant bypass capacitor) and not have the pads for them.

 

[mneary]

Snubbers control the rate of rise, and damp the ringing. I've always considered the diode to be separate from the snubber. An RC snubber is a lot faster than a diode.

That really big capacitor does both; you are correct that it's the same thing and only different because it is being viewed from two different perspectives.

The really big capacitor does a complementary job to the diodes. Without the diodes, the spikes suppressed by the diodes would have no where to go (except the distant battery). There are two current spikes; when the bridge turns on and when it turns off. Ironically, the RC snubber across the load (motor) increases the current demand when turning on.

Snubbers absorb some of the high frequency energy that diodes aren't fast enough to catch. Someone took the snubbers off one of my designs and next thing you know a $55K machine failed FCC testing.

 

[dknguyen]

Do you know why the resistor is used rather than just a small capacitor across the device being snubbed? Or does is it just there so the capacitor itself doesn't fry during discharging.

I am also unsure what is this current demand "when turning on". I understand when turning off since you are cutting off current flowing through inductances everywhere, but current when turning on (unless you are referring to the motor stall/starting current or are you referring to every inductive load in general) ? Can't you just ramp up the current (or is this mainly referring to hard switching on?)

Here is a schematic of one phase for my 3-phase BLDC driver as well as the power supply. I'm thinking the snubber on the power supply is kind of really pointless because of the large capacitor as well as the snubber that already exists across each power MOSFET. Or is the big capacitor "too slow"- I'm not even sure I have to snub the voltage across the bridge when I have on across each MOSFET already. Do you see anything wrong with the snubbing here (particularily the phase-power and phase-ground snubber thing. I just kind of threw it in there. I couldn't find anything it was used in other than a flyback converter so I had to guess where it would go for something like a push-pull circuit).

Or is there supposed to be a simple R-C snubber across the motor? I thought it was supposed to be placed across the switches...or does it do the same thing regardless of where it is- controls the rate of rise of whatever it's connected across. I'm not sure if it's kind of like where the flyback diode should go across the inductive load, except it can't in an "H-bridge-like circuit" due to bidirectional currents so it is placed across the MOSFETs to form an alternate path.

And I suppose are the noise supression capacitors that are often placed across the terminals of a motor (mainly brushed) doing the exact same thing as a snubber? It's a bit hard to tell if they are doing the same thing but the goals are different, or if the difference is just size.

 

[mneary]

Can't follow the diagram. Is it possible to add R- and C- numbers, and suppress the background grid?

 

[dknguyen]

Does this help? I cut out the snubber and flyback diode sections of interest.

 

[mneary]

Thanks, that's a lot easier to read.

The resistors are required to absorb the energy (damp the resonance) of the capacitor and motor inductance. A pure LC circuit would ring at higher voltages, not lower.

When hard switching on, the snubber sees the rate of rise and draws some (smaller) current. It's usually not significant and I apologize for introducing second order effects before adequately explaining the prime factors.

The snubbers across every diode are IMO overkill. A simple RC snubber across each motor winding is enough. That snubber should be enough to absorb the energy available between the cutoff of the winding power and the turn-on of the diode.

General terms of the snubber: If the winding is carrying 1A at 12V, then the resistor might be 12 ohms (limiting the basic rise to [Vcc +12V], or 24V). If the diode turn on time is 12 nS, then the RC time constant needs to be less than that. This gives a capacitance of about 1nF. If this occurs every 1us, then the resistor average power is estimated by 12W*12ns/1us, or 144mW. Substituting a slower diode increases the resistor dissipation dramatically.

A snubber on the Vbat is counter productive. Better to have a low ESR ceramic directly across the 1000uF electrolytic.

Capacitors across brushed motors suppress noise because the leakage inductance is generally less significant with respect to the winding resistance.

 

[dknguyen]

The RC snubbers across all the 3 motor terminals wouldn't dampen ringing would it? It's too far away from the FETs? It would be nice to get rid of those extra diodes- they cost $1.50 a pop! I'm at like $30 worth of diodes so far.

I can see how pure capacitance would form an LC oscillator with the circuit inductance, but what about the giant pure capacitance across the battery?

And, for these snubber capacitors, I was hoping to use SMD X7R ceramic just because I'm using those for everything else and will have a bunch extra, but they don't seem to have inrush current ratings on them.

EDIT: So my diode recovery time is supposed to be 5ns and the RC constant of 1nF+1ohm is 6.28ns...seems awfully small isn't it? Then again...the diode time in reality probably is longer...better to be safe than sorry?

30A@30V -> 1ohm
5ns turn on time, 30kHz frequency
= 900W x 5nsx30kHz = 0.135W
=x5 for kicks, 0.675W

 

[mneary]

RC snubbers across all the 3 motor windings dampens the ringing, since the motor is the source of most of the stored energy. The FETs are only switching, and the stored energy related to that lies only in the inductance of the lines from the FETs to the winding.

The consequence of letting the RC snubber time constant be too fast, is more switching voltage and the resultant emissions. Not a big deal if your MOSFETs are derated and/or you don't have to deal with FCC testing.

The giant 'pure' capacitance across the battery (and the effective capacitance of the battery itself) aren't pure. The reactance is normally so low that circuit resistance provides critical or better damping.

 

[dknguyen]

SOoo much empiricalness! My MOSFETs right now are 40V, but the maximum battery is 21.5V (technically, I am using 25V as for calculations) voltage. Better make some rooms on the pads...I think I'll just leave space for the RCD snubbers across each FET and RC snubbers across the motors...I'm a bit paranoid.

I got a crapload of PCB space anyways. For a 10% increase in the price I Get like 3x the board area. Strange pricing at low quantities.

 

[ClydeCrashKop]

You are probably way past this but I found it interesting:
**broken link removed**
One of the videos on servos explained freewheel diodes in a 3 phase motor pretty well.

 

[dknguyen]

THis is kind of an old thread, but I just finished speccing out snubber capacitors for my ultrasonic transceiver circuit (no problem there since the currents are so low), but then I revisisted this motor driver, and it seems I've specced out 10nF C0G 1206 SMD capacitors for the snubber circuit...they seem kind of...inadequate as far as I can tell for snubber applications where the maximum current is 50A. Suppose they could handle 0.25W like 1206 SMD resistors- they would need an ESR of 0.1mohms to do so! Similar SMD capacitors seem to be rated 1-5s at the rated voltage at 50mA, but otherwise no ratings are given for the current capability of the capacitors. I doubt you can scale down the current pulse time in order to increase the current capability linearily (in that case then it would mean it would be able to handle a 50A for 1ms).

WHat capacitors do you use for something like this and still keep it costing $1 or less per capacitor? Cause I need 30 of them, and it's not cheap! Even the $4 mica capacitors seem to only be rated at around 1A. SO I do not know. I don't want to just get rido f the snubbers though and onyl use diode clamps because you never know what can happen...

 

[Hero999]

Don't you normally add a resistor to increase the rate of decay?

Ideally you want it critically damped so the current falls as quickly as possible.

Normally more power is disipated in the resistor. I wouldn't worry about the capacitor since it only conducts a high current for a short period of time.

 

[dknguyen]

Don't you normally add a resistor to increase the rate of decay?

Ideally you want it critically damped so the current falls as quickly as possible.

Normally more power is disipated in the resistor. I wouldn't worry about the capacitor since it only conducts a high current for a short period of time.

Yeah, but one of the resistors for the RC snubber was calculated to be 0.8R and the resistor for the RCD snubber doesn't limit inrush current to the capacitor since the diode bypasses the resistor, and the resistor is onyl used to burn off the energy afterwards.

EDIT: I found this though and it seems to bring the emphasis a bit more onto the dv/dt rating of the capacitor and bring it away (a little bit) from the switch current at the time of turn-off.
https://www.globalspec.com/goto/PDFViewer?pdfURL=http://www.microsemi.com/micnotes/APT0404.pdf

So the ceramics can't handle the dv/dt it would seem and are about 60% too small for the RMS currents.

MORE UPDATE: So I found some nice "affordable" Mica snubber caps that will do the job. But not I seem to be having problems finding 1R 2-3W power resistors that are not wirewound! The only one I found costs $12 a pop! So...maybe I will just snub the motor terminals and not each diode/MOSFET...then it will cost 1/3 as much...expensive.