Hi, little off the subject of THD impact of mosfet protection. Also, litte late...
The question is: in a case of output device short, the amplifier put all of its current capabilities on the speaker, specially the woofer coil or what ever coil there are in the crossover that can pass current.
When shutting off, Faraday's law implies a high back emf.
With normal mechanical relay usually one shorts the speaker to ground in fault occasion. It is for protecting the relay contacts from the arc.
Is it also a measure that should be taken to protect the speakers? Maybe it may harm the speakers to short this back emf?
I am asking as in a normal SS speaker protection this connection to GND measure can't be taken.
Is it safe for the speakers to leave this back emf un shorted?
For the mosfet themselves, what protect them agains this huge back emf? Internal diodes?
Last, I saw in many places that people write that mosfets are fast vs EMR but who care if most of the time is spent to sense the DC by the LPF and then activate?
Maybe the main issue is that the relay might not opened at all (only after fuses are burnt) with high voltages.
Thanks
The question is: in a case of output device short, the amplifier put all of its current capabilities on the speaker, specially the woofer coil or what ever coil there are in the crossover that can pass current.
When shutting off, Faraday's law implies a high back emf.
With normal mechanical relay usually one shorts the speaker to ground in fault occasion. It is for protecting the relay contacts from the arc.
Is it also a measure that should be taken to protect the speakers? Maybe it may harm the speakers to short this back emf?
I am asking as in a normal SS speaker protection this connection to GND measure can't be taken.
Is it safe for the speakers to leave this back emf un shorted?
For the mosfet themselves, what protect them agains this huge back emf? Internal diodes?
Last, I saw in many places that people write that mosfets are fast vs EMR but who care if most of the time is spent to sense the DC by the LPF and then activate?
Maybe the main issue is that the relay might not opened at all (only after fuses are burnt) with high voltages.
Thanks
Just to add one thing, as in any case the coil will charge into maximal amplifier PS capabilities (it has enough time because of the DC sense LPF) the back emf is huge specially with mosfet switch.
The dt in d(phi)/dt in Faraday's law is tiny compare to EMR.
The dt in d(phi)/dt in Faraday's law is tiny compare to EMR.
We usually set up TVS protection diodes rated at the upper end of the MOSFET’s max voltage.
OK, thanks.
I wonder of this introduces distortion.
I guess not because of the lower RDSon.
I wonder if mosfet if it really needed. I see at Bonsai website that he doesn't use it.
I wonder of this introduces distortion.
I guess not because of the lower RDSon.
I wonder if mosfet if it really needed. I see at Bonsai website that he doesn't use it.
I think the TVS is a useful addition to a mosfet output relay. You could alternatively run reverse biased diodes from the output side of the SSR to the amp supply rails so any inductive energy from the speaker load is dumped back into the supply rails. This is a little more inconvenient than a TVS for a stand alone SSR, but if you are integrating the SSR on your amp modules, it just means placing the protection diodes that you would in any event place across the output bipolar devices after the SSR. mosfets come with drain-source body diodes that offer back emf protection but not if you are worried about speaker back emf and you are using a SSR. In that case, if you opted for diode protectio, you would still have to add the. After the SSR to the supply rails.
The distortion introduced by SSR’s is negligible - Mark’s measurements attest to that as are the other measurements done by DIY audio members. It’s easy enough to see why. If your SSR is say 5 milli Ohms and it’s it’s 1% non linear, that’s 1% of 5 milli Ohms and 5 milli Ohms is .0625% of 8 Ohms, so the actual worst case effect is .000625 ie 6.25 ppm. But, when turned on properly, mosfet Rdson variation with load current is much less than 1%, so the distortion contribution is in effect much lower than 6.25 ppm.
The distortion introduced by SSR’s is negligible - Mark’s measurements attest to that as are the other measurements done by DIY audio members. It’s easy enough to see why. If your SSR is say 5 milli Ohms and it’s it’s 1% non linear, that’s 1% of 5 milli Ohms and 5 milli Ohms is .0625% of 8 Ohms, so the actual worst case effect is .000625 ie 6.25 ppm. But, when turned on properly, mosfet Rdson variation with load current is much less than 1%, so the distortion contribution is in effect much lower than 6.25 ppm.
Last edited:
No added distortion- measurements essentially show no difference. If anything, there is less distortion with SSR. I think there is a reason for this, as it was also observed by tomchr in his SSR protection.
As function of power:
As function of frequency with and without:
Thanks for your answer.
Of course I have diodes but to protect the bipolar outputs so before protection.
If I understand you correctly, the mosfets are safe by its internal diodes.
So putting the TVS for the speaker?
The question: The speaker could be damaged by back emf? In EMR a practice (to protect relay) is to short to GND in fault scenario.
I wonder if it is not more harmful for the speaker to give it a way to pass this large current.
I don't know.
Of course I have diodes but to protect the bipolar outputs so before protection.
If I understand you correctly, the mosfets are safe by its internal diodes.
So putting the TVS for the speaker?
The question: The speaker could be damaged by back emf? In EMR a practice (to protect relay) is to short to GND in fault scenario.
I wonder if it is not more harmful for the speaker to give it a way to pass this large current.
I don't know.
Thanks. Also Nuerochrom has nice graphs.
I have already build SSR protection and indeed no added distortion.
I did not put TVS and recently I was thinking maybe it is a mistake.
I started to think about that after I understood that the speed of the mosfet is not obtained becuasd it is the DC sense delay that dominant. That the reason the inductance charge fully and then one obtain a huge emf because the shut down is now fast by mosfet 😊 .
Then I started to think of I need go protect the mosfet .
I think a promotion to mosfet protection is not the speed but other thing.
Maybe to say that EMR sometimes does not open at all!
Yes, completely logical.
I wonder if back emf is harmful for speaker? Maybe it makes the situation even worse with clamping from the speaker point of view (not the mosfet).
Many articles can be found on the internet giving answers to a lot of basic questions regarding MOSFET relay for speaker protection.
https://sound-au.com/articles/mosfet-relay.htm
https://sound-au.com/project198.htm
Just two quick examples.
Being able to measure the actual distortion of the relay itself, without being masked by that of the amplifier, is the topic of this thread.
Cheers,
Patrick
https://sound-au.com/articles/mosfet-relay.htm
https://sound-au.com/project198.htm
Just two quick examples.
Being able to measure the actual distortion of the relay itself, without being masked by that of the amplifier, is the topic of this thread.
Cheers,
Patrick
Thanks.
I think I did not see Elliott put TVS. He mentioned the issue of back emf at cut off but I did not see he took a a measure except for EMR where speaker out is connected to GND when relay off.
I will look again, thank you.
I think I did not see Elliott put TVS. He mentioned the issue of back emf at cut off but I did not see he took a a measure except for EMR where speaker out is connected to GND when relay off.
I will look again, thank you.
Several years ago I measured the THD (10W@4 Ohm) of the power amplifier with and without MOSFET relay protection made by my colleague (@VEC7OR). I could not detect any difference in distortion with and without the relay made using modern (=low Rds) MOSFET switches.
Last edited:
Solid State Relays are great for loudspeaker protection, and this is why I built such a protection as well.
However, mine shows some weird behavior that totally puzzles me: It adds significant distortion, which is unexpected.
Here is the relevant part of the schematic with some DC measurements:
Nothing special really and all operating points are as expected:
There is 21mA current through the LEDs of the photovoltaic MOSFET drivers, which results in 14.5V gate to source voltage at the MOSFETs, thus they are deep in saturation for lowest on resistance and maximum current carrying capability.
Here is the spectrum I measured without the SSR protection circuit in place:
THD is 6ppm, which is nice.
And this is with the SSR in place:
Now 230ppm, which is much higher than expected.
I checked the voltage across R60 for oscillation (you never know...), but there is none.
What else could I measure?
Any ideas what I could have done wrong?
However, mine shows some weird behavior that totally puzzles me: It adds significant distortion, which is unexpected.
Here is the relevant part of the schematic with some DC measurements:
Nothing special really and all operating points are as expected:
There is 21mA current through the LEDs of the photovoltaic MOSFET drivers, which results in 14.5V gate to source voltage at the MOSFETs, thus they are deep in saturation for lowest on resistance and maximum current carrying capability.
Here is the spectrum I measured without the SSR protection circuit in place:
THD is 6ppm, which is nice.
And this is with the SSR in place:
Now 230ppm, which is much higher than expected.
I checked the voltage across R60 for oscillation (you never know...), but there is none.
What else could I measure?
Any ideas what I could have done wrong?
You might try replacing R63 and R64 varistors with TVS protection diodes set at voltage limit of MOSFETs. For simple test, try removing the resistors and see if distortion goes away. If a varistor is active within the audio path it will introduce nonlinear behavior which leads to distortion.
Good point, thanks.
I can't remember why I chose MOV instead of TVS back then.
The MOV (B72214S0111K101) is supposed to have a working voltage of 150VDC, conduct at 180VDC.
In any case, this is a non-linear component in parallel with the on resistance of the SSR MOSFETs.
Back when I made the decision, I expected that the non-linearity of the MOV should not matter too much since the voltage across the MOV is very low, like 54mV in given load case.
I removed the MOV as an experiment, but did not find significant impact:
I would say this is well within typical day to day measurement variance.
I can't remember why I chose MOV instead of TVS back then.
The MOV (B72214S0111K101) is supposed to have a working voltage of 150VDC, conduct at 180VDC.
In any case, this is a non-linear component in parallel with the on resistance of the SSR MOSFETs.
Back when I made the decision, I expected that the non-linearity of the MOV should not matter too much since the voltage across the MOV is very low, like 54mV in given load case.
I removed the MOV as an experiment, but did not find significant impact:
I would say this is well within typical day to day measurement variance.
You have an unusual (for this application) monsters of a MOSFET. I wonder if they, for some process/technology reasons, doesn’t conduct equally well in both directions, which is required for an AC SS relay. If you have at hand some other suitable parts, check with them.
Indeed, well possible.
I will make up my mind about different MOSFETs and have a look at other designs.
The rationale behind the monster MOSFETs was that during turn-off, they are in linear region briefly and SOA can matter. No idea how fast the turn off could be in reality.
So I wanted MOSFETs that have plenty of SOA.
The package is huge, but my MOSFETs are not that much stronger than the BSC040N10NS5 that xrk971 chose.
Here I illustrated my thoughts:
Given the rail voltage is higher, and the load is heavier, like 4 Ohm or even lower, there may be 20A or even 40A at maybe 80V to turn off.
I don't believe this could take as long as a few hundred micro seconds, but I have no idea how the turn off circuitry inside the photovoltaic driver works actually and wanted to be safe.
Probably the real issue is something totally stupid. Like I'm dealing with stupid mistakes for months now.
I still can't believe I forgot to wire the "enable" signal to the protection module and was wondering about stupidly high noise floor and terrible performance during FFT measurements all the time until I figured out that the SSR are off and the signal must pass through the off capacitance and body diodes of the MOSFETs. Since the capacitance is huge, quite a lot comes through given the load impedance is high enough, like 1Meg Ohm of the scope or a few kilo Ohms of the ADC.
I will make up my mind about different MOSFETs and have a look at other designs.
The rationale behind the monster MOSFETs was that during turn-off, they are in linear region briefly and SOA can matter. No idea how fast the turn off could be in reality.
So I wanted MOSFETs that have plenty of SOA.
The package is huge, but my MOSFETs are not that much stronger than the BSC040N10NS5 that xrk971 chose.
Here I illustrated my thoughts:
Given the rail voltage is higher, and the load is heavier, like 4 Ohm or even lower, there may be 20A or even 40A at maybe 80V to turn off.
I don't believe this could take as long as a few hundred micro seconds, but I have no idea how the turn off circuitry inside the photovoltaic driver works actually and wanted to be safe.
Probably the real issue is something totally stupid. Like I'm dealing with stupid mistakes for months now.
I still can't believe I forgot to wire the "enable" signal to the protection module and was wondering about stupidly high noise floor and terrible performance during FFT measurements all the time until I figured out that the SSR are off and the signal must pass through the off capacitance and body diodes of the MOSFETs. Since the capacitance is huge, quite a lot comes through given the load impedance is high enough, like 1Meg Ohm of the scope or a few kilo Ohms of the ADC.
Meanwhile I tried to optimize the control circuitry because I was suspecting issues there. Not sure whether there were any, but going through everything in detail and questioning every single component value was certainly not a bad idea.
I was suspecting something stupid and I believe I found something stupid:
THD may increase five fold dependent on the Speakon connector.
I can vaguely remember that the 6ppm measurement was taken with SSR in place, but back then I had four pin Speakon connectors installed and later switched to the two pin Speakon + 6.3mm combo variant. Maybe the plug does not fit this connector as well? Using a cheap Chinese knock-off plug for the measurement cable probably isn't too clever either. All PCB connectors are original Neutrik however.
During optimization of the control circuitry, THD went up and down without showing a clear pattern. Probably just because each time the connection was differently. Dependent on how well the connection is, best I can achieve right now is 50ppm, worst is 230ppm roughly.
Just wanted to let you know...
I was suspecting something stupid and I believe I found something stupid:
THD may increase five fold dependent on the Speakon connector.
I can vaguely remember that the 6ppm measurement was taken with SSR in place, but back then I had four pin Speakon connectors installed and later switched to the two pin Speakon + 6.3mm combo variant. Maybe the plug does not fit this connector as well? Using a cheap Chinese knock-off plug for the measurement cable probably isn't too clever either. All PCB connectors are original Neutrik however.
During optimization of the control circuitry, THD went up and down without showing a clear pattern. Probably just because each time the connection was differently. Dependent on how well the connection is, best I can achieve right now is 50ppm, worst is 230ppm roughly.
Just wanted to let you know...