MosFet Relays

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Also tired about struggling with old or low quality speaker relais?
May be this solution is more adequate for the 3rd Millenium.
It offers fully isolated speaker switching within few microseconds and distortion well below my measurement accuracy.
Don't bother about the huge PCB. The shown PCB is a trial PCB for multiple functional blocks which I intend to use in my next class D amp and also for getting familiar with the layout software...
 

Attachments

  • Switch1_PCB.JPG
    Switch1_PCB.JPG
    343.2 KB · Views: 4,070
The distortions are below my measurement possibilities.
Especially I am lacking a power amplifier, which would have a harmonic spectrum below -120db... :p
I adjusted one of my amps for class A operation, and
even with this the distortion spectrum remains unchanged when adding the switch. Here the spectrum of the class A amp with this switch delivering 10W into 5 Ohms.
 

Attachments

  • Distortion1.png
    Distortion1.png
    110.8 KB · Views: 3,712
Last edited:
How can I tell -120db from the above? From the above I cannot.
But I analysed the voltage drop accross the switch itself, which is approx. 1/1000 of the amplifier output voltage.
Analysing this drop showed nothing but the fundamental and the noise of
the mic amp, which is 60db below the drop of the switch.
So the distortions of the switch are below -60db of 1/1000 of the amp output.
Means referencing to the amp output the switch introduces distortions below -120 db.
 
Speaker Relay

...just for completing this monologue about a Solid State Speaker Relay...
Different from what some people might guess - such a MosFet solution can be very rugged. You have to settle a proper gate drive, which ensures a fast transition of the MosFets from isolating to conductive and vice versa.
Please note that the schematic shows the possibilty of multiple winding connections, furtheron the gate drive transformer has a myor influence. The types on the picture are not the best choice for this application.
If you ensure reasonable fast switching, i.e. voltage-current-overlap of 1us or less, then you can benefit of a huge MosFet SOA.
The shown circuit employs IRFB4321 types and I am not able to test it at its limits. My max. load that it can supply in a controlled way is around 75 ampere. In the attached picture you can see the switch operating to short circuit a power transformer, which delivers 125Vp unloaded and delivers about 75Ap during short circuit condition.
Please note that 125Vp is not recommended, because at this voltage the protection varistors do already start to clamp and generate continuous heat.
The shown circuit is fine for amps delivering in unbridged configuration approx. 85Vp, means OK for amps with rails up to +/-85V.
Generally this switch is suited not only for solid state amps but also for class D, in the range of 1kW...2kW and up to 3kW in bridged configurations with rails up to +/-85V. For bridged configurations you have to ensure that the switches for both Amp outputs are controlled with the same ON/OFF-control signal, otherwise you might overload the protection varistors.
And it is just a question of the MosFets+varistor to go higher.

White trace, measured with 10:1 probe: Transformer output voltage, 50V/div
Red trace, measured across 1 mOhm shunt: Current load, 50A/div
 

Attachments

  • 125V_75A.png
    125V_75A.png
    163.9 KB · Views: 3,436
Last edited:
The first schematic is the Rolls Royce, which might look complicated.
But we can simplify things to
2 MosFets
1k Resistor
9V Battery

For first real life trials you can even skip the protection varistor (BTW: Overvoltage protection could also be ensured by simple diodes from the MosFet drain to the amplifier rails).

The attached schematic shows all you need to start playing around.
Battery not connected ==> Speaker off
Battery connected ==> Speaker on
Feel free to start playing around and listen.

The MosFets should have low Rdson, the voltage rating should be as high as the rails of your amp.

How does the circuit work? If turned ON the N-chanel of the MosFet is able to conduct in both current directions! The only unpleasant thing is the body diode during the turned OFF situation. If you use only one MosFet, then you could block only on current direction. So I simply use two MosFet in series, connected in opposite direction. This is done since decades in bilateral CMOS-Switches. Modern MosFets allow such low Rdson that you can neglect the influence on distortion and damping factor.
 

Attachments

  • Principle.JPG
    Principle.JPG
    31.6 KB · Views: 3,386

Attachments

  • C ...-zzz-Emeko.ckt.pdf
    43.2 KB · Views: 679
Basically yes, but do NOT start playing like proposed in posting #6 in a high voltage circuit.

If you want use the MosFet Relay for this application you will have to use 600V MosFets with low Rds_on and use a protection varistor which is intended to operate at 250V or 275V mains.
Furtheron your driving transformer must ensure the isolation for such voltages.

The nice thing is that you can test the relay and ensure by measurements also for proper gate signals under low voltage conditions and then if everything is Ok, you may step to higher voltages.
In any case: Take care! Mains voltage applicatons can easily kill a homo sapiens and even easier a homo incautus.
 
Hi Chocoholic

You seem to have done a good job ensuring artifacts in the ON conditions are minimal.
But you might have overlooked the OFF state. Large MOSfets have pretty large and non-linear parasitic capacitances, and they could cause unwanted leakage into the output when OFF.
Choco1 shows the isolation vs. frequency of a circuit similar to yours. At 3KHz, in the region of maximum audibility, the isolation is ~60dB, which is clearly insufficient.
This applies to a single throw switch. The situation would much improve in multiple-throw configurations.
Choco2 shows the situation in the time domain, with a slew-limited square wave.
 

Attachments

  • Choco2.gif
    Choco2.gif
    36.7 KB · Views: 1,122
  • Choco1.gif
    Choco1.gif
    35.6 KB · Views: 3,315
Still - its at -80dB at 20khz, and the primary reason for protection relay is to protect the speaker from DC by catastrophic failure, and to prevent turn-on/off peaks.

Chocholic: Did you listen to the amp with this circuit, and does it work as inteded in real life?
 
If it is a protection relay only, the isolation is unimportant, so it should be OK.

Combining the battery option with a small circuit could improve the matters slightly for other applications: by reverse-biasing the G-S capacitance, the isolation can be improved by ~6dB (ChocoLv1).
ChocoLv2 shows an example circuit doing that (obviously, not all inputs should be implemented).
 

Attachments

  • ChocoLv2.gif
    ChocoLv2.gif
    30.4 KB · Views: 1,662
  • ChocoLv1.jpg
    ChocoLv1.jpg
    170 KB · Views: 1,537
Hi Ho !
That's nice, the discussion is started.
Off isolation was also one of my concerns, especially because of the protection varistor. First calculations told approx. 60db at 1kHz.
My measurements at 3.9 Ohms load show -63db at 1kHz and -44db at 10kHz.
As you already noticed for a speaker relay that's fine, for signal selector you would need the traditional arrangement with two switches.
One in series with the load, the second in parallel with the load and both switches driven with inverted control signal. You can of course do this arrangement also with the shown switch and theoretically then you will reach values in the 100db range.
But as you already noticed: For a speaker relay, the OFF-isolation of the simple arrangement is Ok.
This was also the result of the listening test.You can play really loud and then switch off - also perceptionwise it is off.
Also for the undesired pop, the isolation is perfectly fine.
And considering DC-protection, this switch is perfect anyway.

Regarding the artifacts during ON status: Yes, the artifacts are well below my measurement capabilties. The shown artifacts are fully caused by the amp itself, not at all by the switch.

Listening? ...listening to audio equipment? :confused:
Of course I do. :cool:
This switch is completely neutral to my ears.
I can listen for long time - fine.
I can compare, with and without switch - no difference.
Even when putting a jumper wire across the switch in status ON during listening - it simply does not make any difference, not any change in sound or level comes to perception.

So I decided to use this method in future for my amps.
 
Administrator
Joined 2007
Paid Member
This is a very interesting idea... as a big problem with relays under fault conditions (an output device going short etc) is that the contacts could weld together thus defeating the object of offset protection.
It would be interesting to test the switch at DC measuring the volt drop across it as current flow is increased, in both directions to see how symetrical and linear it is, perhaps up to 10 to 15 amps or so.
I am sure it does appear sonically transparent, unfortunately if it worsened the distortion of say an amp in the 0.00xxx range many would dismiss it.
Perhaps the "distortion" it does introduce is pleasing to the ear...

Interesting idea :)

Edit... and perhaps incorporate the amp feedback to include the switch in the loop.
 
Fast switch off at fault conditions with inductive load (speaker) can cause high voltage spikes (and voltage breakdown). Aditional protection clamping diodes at both sides of MOSFET switch are necessary..Internal body diodes are not sufficient.
 
Last edited:
Mooly:
That's a good idea. Especially with respect to the fact that all my meaurement methods so far were less good than the switch.
Unfortunately I cannot measure up to 10A DC, because I am lacking a
strong enough DC source. Also at 10A there will be heat, which is increasing the Rdson, so an unrealistic thermal distorsion would happen.
May be I can do pulsed measurements with the scope. I. e. DC-Pulses of 100ms...200ms duration, simulating realistic load distribution like drums and then seeing a realistic thermal distorsion.

Analog:
If the relay is high quality and in good condition, then I cannot measure and cannot hear any issue with a relay. Relays are fine, except old/defect/poor quality.
When I took one relay from my shelf I found an ugly high order spectrum with peaks in the range around -80db. Knocking a few times to its plastic cover 'repaired' it...
Similar thing I found with fuses in the speaker path, after some years in the fixture the contacts become poor, found 0.07% THD. Screwing them out and in again - repaired.
And of course I also had the situation that a relay was so obviously bad, that one speaker was fading, somehow distorted.
The above experiences are my motivation to move away from the relays.

BV:
Take your time and read the entire thread.
Your proposal with the diodes I mentioned in posting #6.
Furtheron in most cases you will only need them on the side of the speaker , on the side of the amp, these diodes are usually already existing in the amp.
Have a look to the first schematic, there I am using a varistor for protection.
Without that the heavily inductive test with 75A would not have been a good idea.
The simplified schematic is just for starting playing.
Please note: The internal body diodes do not clamp.
But for first playing around the avalanche capability of most MosFets will do the job. For a proper setup of course a varistor or two antiserial TVS-diodes or a bidirectional TVS acrross the switch are necessary. Or at least two diodes towards the rails ... or any other rugged clamping means...

Andrew:
IRF540 and IRF640 will work for first playing around, but due to their slightly high value of Rdson, both types are not the best choice. Of course paralleling will help. If the 100V of the 540 are sufficient in your case, you should prefer the 540 instead 640.
I used IRFB4321. Also a good choice would be IRF52n15d or for higher voltages IRFB4227.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.