Here is a test waveform with an Exicon lateral MOSFET. The rise time test is an ECX10N20 driven by 0 to 2 volt (no negative zero crossing) square wave at 100khz , with a 470 ohm drain load from a 42 v power source. Scope is at 5v / div for detail. The tiny step at the start of rise is because there is no bias when signal source is at 0. I wanted to see the characteristics from a zero start. It is one fast device.
Attachments
Those were good parts. Unfortunately they fell a little short on current. The half amp rating was ok for hifi receivers and integrated amps, and of course hobby builds, but at least 1.5 amp rating would have been needed for professional amplifiers. They had the 8 amp devices, I never understood why they didn't make a driver with an amp or two rating.
Transistorlegacy, thanks for the good info!
For a moment I was scratching my head thinking how can they omit the isolators between mosfets and heatsink in the soundcraftsmen amp. Then I remembered that some amps with the lateral mosfets have no degeneration on the outputs.
For my build I am thinking of finding some surplus PC processor heatsink/fans and use one per mosfet, no electrical connection from the aluminum to anywhere so no parasitic capacitor. I will not need any isolation, I will build on a plank of wood.
Thanks,
Alex
For a moment I was scratching my head thinking how can they omit the isolators between mosfets and heatsink in the soundcraftsmen amp. Then I remembered that some amps with the lateral mosfets have no degeneration on the outputs.
For my build I am thinking of finding some surplus PC processor heatsink/fans and use one per mosfet, no electrical connection from the aluminum to anywhere so no parasitic capacitor. I will not need any isolation, I will build on a plank of wood.
Thanks,
Alex
In common source circuits using a live heatsink design, no isolation is need if there is a real case to heatsink connection, as with a TO-3.
The plastic case MOSFETs don't connect well enough to prevent oscillation, and ceramic isolators must be used, or undesirably high value gate resistors. The Exicon MOSFETs are faster and lower source resistance than originals. If you have any capacitor effect between case and heatsink with multiple devices, it will reliably oscillate. Even old MOSFET amplifier designs were marginally stable, hence how many need repair. I always add ceramic insulators on such repair job, unless it has good live heatsink connection like a Soundcraftsmen , the result is stable even with no zobel network.
I scored some of those in an auction! 😎
This was here in brazil of all places! They are almost impossible to find.
https://ibb.co/nbPL5CN
I think you mean source follower, aka common drain.
(Obviously the heatsink must be isolated from ground / chassis.)
I think everybody that read this got it already, but just to be clear, the case of the mosfets is the source terminal AND there's no source resistors in the design.
Last edited:
In reference to the ESP P101 amp build. Output osc. discussed by Rod E.
"Finally, the output section uses a conventional Zobel network. If desired, this can be followed (right at the speaker terminals) by a small inductor with a paralleled resistor, and finally a second Zobel network at the output. This is intended to decouple the amplifier from highly capacitive loads and ensure stability under all normal operational conditions. The use of ultra-low inductance (very high capacitance) speaker leads is not recommended, although the additional output network described will prevent oscillation in most cases. The inductor can be quite small (around 5-10µH is about right), and must be shunted by a parallel resistance of between 2.2 and 10 ohms (10µH gives a -3dB frequency of 71kHz with an 8 ohm load)."
Since I'm using double dye T03 Exicon outputs mounted on the heatsinks and hardwired (about 5 inches) to the boards, I thought I would use his suggestion and wire the parallel inductor and resistor (5uH / 5 Ohm) at the speaker terminal. What would be any downside to NOT using the second
Zobel. Duplicating R18, C8, and C10. I'm assuming that is what he meant by a second Zobel at the output.
Is there an online calculator for determining the -3dB cutoff freq. as Rod did?
"Finally, the output section uses a conventional Zobel network. If desired, this can be followed (right at the speaker terminals) by a small inductor with a paralleled resistor, and finally a second Zobel network at the output. This is intended to decouple the amplifier from highly capacitive loads and ensure stability under all normal operational conditions. The use of ultra-low inductance (very high capacitance) speaker leads is not recommended, although the additional output network described will prevent oscillation in most cases. The inductor can be quite small (around 5-10µH is about right), and must be shunted by a parallel resistance of between 2.2 and 10 ohms (10µH gives a -3dB frequency of 71kHz with an 8 ohm load)."
Since I'm using double dye T03 Exicon outputs mounted on the heatsinks and hardwired (about 5 inches) to the boards, I thought I would use his suggestion and wire the parallel inductor and resistor (5uH / 5 Ohm) at the speaker terminal. What would be any downside to NOT using the second
Zobel. Duplicating R18, C8, and C10. I'm assuming that is what he meant by a second Zobel at the output.
Is there an online calculator for determining the -3dB cutoff freq. as Rod did?
Hello sir
I'm looking for a way to get 3-4 pairs of 20N20/20P20 to work stably in parallel, and I want to keep the high-speed characteristics of the MOSFETs as much as possible, lose as little effective bandwidth as possible, and maximize the amount of negative feedback available.
They oscillate easily, as many have encountered. I've tried increasing the gate resistance, and while this solves the local oscillation problem, it also makes the MOSFET output stage too slow with a huge loss in distortion performance.
Until one day, I found this post😛
I was wondering, when you try to connect more 20N20/20P20 in parallel, do you also use the compensation values you shared?
If I want to not slow down the MOSFET as much as possible, is there a more optimal way of compensation and compensation values?
thank you very much
I didn't see this post until now... Yes the compensation values listed right in the start of this thread are for each device. I listed the values for single and double die devices. The recommended ceramic insulator is important. This solution has attained 100 v/us slew rate for me. I found this solution to be better than the alternatives, such as using a live heat sink and and bonding the source lead right to the heat sink (emulating the TO-3 mounting), or using large bypass capacitors from drain to ground plane at every device, or the dreaded heavy compensation where the reduced slew leads to phase lag problems, leading to needed more compensation in more stages.
After 3 decades I realized a certainty : once you slow down an amp, you have to slow it down a lot to keep it stable. That's why so many amps are under 20 v/us.