In recent years, more and more hi-end audio manufacturers are turning to MOSFET relays for loudspeaker protection, most notable of them Accuphase.
Unless used in the shunt mode to short circuit the amplifier output, a loudspeaker relay is in the signal path.
If the contact resistance of the relay is not low, consistently constant over lifetime, and sufficiently linear,
it will be a source of added distortion between speaker and amplifier.
This is particularly the case for low-level signals.
While MOSFET relays do not suffer from those problems with a mechanical contact, the linearity of its “contact” resistance has not been reported in public.
Most measurements seen to date are those in conjunction with a resistive load.
If the distortion spectrum with or without the relay are seen to be identical, then the relay is declared distortion “free”.
This of course depends on the distortion of the base amp in the first place, with examples up to -90dB THD or higher at 1W 1kHz into 8 ohm.
The true performance of the relay is then hidden behind the base amp’s distortions.
I was told in private that this was actually how a well-known hi-end manufacturer measured their own design.
Not entirely satisfied with this, we set ourselves the task to try to measure the distortion of the relay itself directly.
As you can see in the article, we have successfully completed the task, and the DUT itself has a distortion below -80dB even at a current of 1Arms,
corresponds to 8W into 8Ohm.
PS Most sincere thanks to Jan Didden for carrying out all the measurements.
Patrick
.
Unless used in the shunt mode to short circuit the amplifier output, a loudspeaker relay is in the signal path.
If the contact resistance of the relay is not low, consistently constant over lifetime, and sufficiently linear,
it will be a source of added distortion between speaker and amplifier.
This is particularly the case for low-level signals.
While MOSFET relays do not suffer from those problems with a mechanical contact, the linearity of its “contact” resistance has not been reported in public.
Most measurements seen to date are those in conjunction with a resistive load.
If the distortion spectrum with or without the relay are seen to be identical, then the relay is declared distortion “free”.
This of course depends on the distortion of the base amp in the first place, with examples up to -90dB THD or higher at 1W 1kHz into 8 ohm.
The true performance of the relay is then hidden behind the base amp’s distortions.
I was told in private that this was actually how a well-known hi-end manufacturer measured their own design.
Not entirely satisfied with this, we set ourselves the task to try to measure the distortion of the relay itself directly.
As you can see in the article, we have successfully completed the task, and the DUT itself has a distortion below -80dB even at a current of 1Arms,
corresponds to 8W into 8Ohm.
PS Most sincere thanks to Jan Didden for carrying out all the measurements.
Patrick
.
Last edited:
Nice report Patrick.
I use mosfet based SSR’s on my commercial products, one of which does 7-8 ppm (this is the limit of my test gear) at 200 W into 8 Ohms. The SSR is in series with the load (speaker hot).
Mosfet SSR’s are the way to go. No wear out mechanisms, and no need to replace after a fault condition, which you should do with an EMR. They handle very high switching energies and can switch 60A in < 100 us.
I use mosfet based SSR’s on my commercial products, one of which does 7-8 ppm (this is the limit of my test gear) at 200 W into 8 Ohms. The SSR is in series with the load (speaker hot).
Mosfet SSR’s are the way to go. No wear out mechanisms, and no need to replace after a fault condition, which you should do with an EMR. They handle very high switching energies and can switch 60A in < 100 us.
There are a few details in our relay design which is unusual.
So the above measurements only apply to our particular case, straightly speaking.
Having said that, I think it is safe to say that they will not contribute to more than -110~120dB with a 8 Ohm load.
And MOSFETs fail by short circuit, which the user should know.
Patrick
So the above measurements only apply to our particular case, straightly speaking.
Having said that, I think it is safe to say that they will not contribute to more than -110~120dB with a 8 Ohm load.
And MOSFETs fail by short circuit, which the user should know.
Patrick
> Could you please post the schematics of the MOSFET SSR you tested?
Similar to Fig. 3 of :
ESP - MOSFET Solid State Relays
but different devices.
Patrick
Similar to Fig. 3 of :
ESP - MOSFET Solid State Relays
but different devices.
Patrick
Same here. I found that properly designed MOSFET based SSR’s for audio amp applications, contributed no extra distortion within the limit of the measurement gear (-130dB) or the amp under test. For most practical DIY amps, this is more than enough to say it doesn’t add any distortion.
I designed a mosfet DC protect circuit.
Sounded absolutely fine.
My ears dont go down to -80db anyway.
Modern mosfet's have such a low RDSon resistance its not worth bothering about. Not even a need for a heat sink on the mosfets even driving a couple of hundred watts RMS.
Some love tube amps with 0.1% THD.
As long as you are happy with what you have, all is fine.
I am happy with amps with -100dB to -120dB THD.
So I want my protection circuit to be a further 20dB lower.
As said, each to his own.
And also mentioned in post #6, MOSFETs fail by short circuit, so the relay is not 100% fail safe.
Patrick
As long as you are happy with what you have, all is fine.
I am happy with amps with -100dB to -120dB THD.
So I want my protection circuit to be a further 20dB lower.
As said, each to his own.
And also mentioned in post #6, MOSFETs fail by short circuit, so the relay is not 100% fail safe.
Patrick
Modern Trench devices feature Rds(on) values of 2-3 mOhms on 100V or 150V Vds devices. Wiring and PCB trace resistance is a bigger issue.
They are used in auto ABS systems, fuel injection and engine management systems i.a.o under the hood - all these applications are high temp, high stress and typically feature < 100 part per billion field failure rates.
If you size your mosfets right, you will not get failures in an audio amp.
Last edited:
The ones I chose (intended for the F5X power amp) are low voltage devices.
Datasheet says Rdson is ~0.75mOhm each, so 1.5mOhm total.
In actual measurement at the relay PCB pins, they measured 5mOhm total.
So all the Solder joins, PCB traces (70µm), etc, add up to ~3mOhm.
But I guess they are pure resistive and do not contribute to distortion.
Off capacitance is not really an issue for the low impedance that speakers have.
Quite a different story for line level impedances (say 10~100k).
Patrick
Datasheet says Rdson is ~0.75mOhm each, so 1.5mOhm total.
In actual measurement at the relay PCB pins, they measured 5mOhm total.
So all the Solder joins, PCB traces (70µm), etc, add up to ~3mOhm.
But I guess they are pure resistive and do not contribute to distortion.
Off capacitance is not really an issue for the low impedance that speakers have.
Quite a different story for line level impedances (say 10~100k).
Patrick
This thread has inspired me to dig out my prototype MOSFET speaker protect circuit and measure it.
The power amp is a Blameless style design with CFP output stage, I've measured it with and without 8 ohm resistive load, and before and after the protection unit:
The responses are the same before/after protection to within measurement variation I reckon. The levels on the x-axis are input to the amp, it has 27.7dB of gain, so its really -12dBV to +16dBV at the speaker terminals.
So pretty conclusive that it's performing nicely, given it uses IRLR3717 MOSFETs with 3.4 milliohms on-resistance it seems to show good linearity of that resistance (if it were non-linear it would provide upto a maximum of -65dB distortion under 8 ohm load, but its clearly well below that.
It borrows a technique from Doug Self for sensing, but I think I adapted it for opto-coupled MOSFET outputs. It doesn't need any special supply voltage, just the main rails and ground, so it ought to be fittable to any amp really. It lacks latching ability currently but its a proof of concept:
The power amp is a Blameless style design with CFP output stage, I've measured it with and without 8 ohm resistive load, and before and after the protection unit:
The responses are the same before/after protection to within measurement variation I reckon. The levels on the x-axis are input to the amp, it has 27.7dB of gain, so its really -12dBV to +16dBV at the speaker terminals.
So pretty conclusive that it's performing nicely, given it uses IRLR3717 MOSFETs with 3.4 milliohms on-resistance it seems to show good linearity of that resistance (if it were non-linear it would provide upto a maximum of -65dB distortion under 8 ohm load, but its clearly well below that.
It borrows a technique from Doug Self for sensing, but I think I adapted it for opto-coupled MOSFET outputs. It doesn't need any special supply voltage, just the main rails and ground, so it ought to be fittable to any amp really. It lacks latching ability currently but its a proof of concept:
Last edited:
For your interest :
https://www.diyaudio.com/community/...tage-perfectly-integrated.406024/post-7532667
Patrick
https://www.diyaudio.com/community/...tage-perfectly-integrated.406024/post-7532667
Patrick