MosFet Relays

[jan.didden]

Originally Posted by ChocoHolic

Sorry that I did not comment earlier.
No doubt, that's also a proper implementation. The relation of component count and performance is pretty nice.
It's more a question of personal taste that I am not in love with flyback converters close to audio signals. But that's silly, I know. Every tone that I intend to feed into my amp these days has already gone through kilometers of wires next to flybacks in modern audio processing and distribution systems....

Not silly at all, I believe that is a valid consideration.
It's just that I don't believe it has an impact in this implementation.
In the extreme you can even get excited about the ripple on the power supply of a closed mechanical relay.
In theory, that can modulate the ontact pressure and thus modulate the contact resistance.
How far you want to go.

jan

[ChocoHolic]

Originally Posted by Struth

Hi Guys

The first page of the thread has no THD info about mosfet relays.

londonpower.com

Well in posting 3 and 4 I tried to measure and had to accept that the distortions are far below any of my measurement capabilities.
In posting 3 I measured the THD at the output of amp with reasonably low distortion and found no difference with and without MosFet relay.
In post 4 I described that I measured the signal across the MosFet relay itself and found that there the distortion of the voltage drop across the MosFet is still unmeasurable for me, due to the noise when measuring small signals in the range of few mV, which limits this measurement to -60db.
The series resistance of the switch is approx factor 1000 less than the speaker resistance. At the same time the nonlinearity of this small series resistance is below -60db. Means the MosFet relay acts as a very constant series resistance with a variance of less than -60db.
Assume we have 8 Ohms speaker and put in a series resistor of 8 mOhm:
When the amp delivers 10V we will find 9,99000999V at the speaker.
Now let's calculate how the signal will be affected by the change of RdsON, which is less then -60dB of 8mOhms.
...means the RdsON could vary up to 8,008mOhms:
When the amp delivers 10V, now we will find 9,99000001V at the speaker.
Consequently the non linearity of the switch will distort the output signal about 0,0000098V.
Related to the speaker signal of 9.99V signal this variance of less 9.8uV translates to distortion levels below -120db.

As soon as my class D amps become better than THD<0.0001% I have to check the relay more detailed.

[ChocoHolic]

Originally Posted by janneman

How far you want to go.

That's the key question.
And I am convinced I would not hear the difference between your switch with a flyback and the same with a battery.
But I would feel more audiophile with a battery. That's the silly point.

[dhalbakken]

I want to build a simple MosFet relay for switching the speaker output from an amp with +-50 V rails using locally available parts. This is what I came up with.

LTSpice says the turn-off time from 50V to 0V is less than 100us.

Does this look like a reasonable design?

[AndrewT]

Less than 1/10th of a millisecond.
That is fast.
Are we talking 50times faster than a relay can release, if the contacts don't weld?

[Mooly]

How will you obtain the 9 volt "floating" DC source in practice ?

Yes its fast... whether in reality it performs the same as in simulation would have to be seen. The real time delays come in the drive and offset sensing circuitry that would drive the opto.

[Bonsai]

You can get some inspiration here: Solid State Relay with PCB Layout

[Bonsai]

NB at 65V peak out into 8 Ohms, the distortion is about -150 dB down. Even across the actual mosfets switches, the distortion is in the ppm range. Bottom line, SSLR distortion is negligible for all practical purposes.

[jan.didden]

Originally Posted by Bonsai

NB at 65V peak out into 8 Ohms, the distortion is about -150 dB down. Even across the actual mosfets switches, the distortion is in the ppm range. Bottom line, SSLR distortion is negligible for all practical purposes.

Exactly my experience. The distortion I measure on the amp output is the same as what I measure across the switch, indicating that the switch impedance is nicely resistive and linear.

jan

[dhalbakken]

I had intended to use a separate transformer and dc supply for the floating 9V, but I like Bonsai's approach using PV diodes instead of an external power source much better. I was unable to find a Spice model for the TLP191 that Bonsai uses, so I stacked some HCNR200's (which has an available Spice model) to attempt a similar emulation in LTSpice. Here's what I got. I trimmed R7 for the best turn-off time (at the expense of turn-on time, which is unimportant to me as long as we stay within the SOA of Q1 and Q2). I got a turn-off time of around 40uS. Bonsai, you say your circuit has a turn-off time of less than 250 uS. Have you tried using an external resistor to tweak the turn-off time of your circuit?