Here's the diode-based solution. The only issue with that is that the clip is slightly asymmetric by about 0.6V. I will retrofit this into the amplifier I've already built so it doesn't blow up if I push it hard.
Last edited:
The diode based solution is the one commonly used in many NJM series operational amplifiers.
https://www.mouser.com/datasheet/2/294/NJM4558_NJM4559_E-218630.pdf
https://www.mouser.com/datasheet/2/294/NJM4558_NJM4559_E-218630.pdf
Yeah I like it, because it's dead simple.
I thought I was the only idiot who had forgotten to protect the VAS and did a quick search of other Blameless-style amplifiers. It's pretty shocking how many of them didn't have it, since it might cascade to blowing up a speaker.
Anyway, I updated the Github site in case anyone finds it and decides to try it.
Also in the TL071 family (the originals, not the CMOS versions). It's missing from the TLE2071, causing a substantial supply current increase during clipping to the negative rail.
Some of you asked for the .ASC file. Here it is. A couple of things:
1. The slew rate figure of 18-19V/us is taken before the output coil, since the coil will inevitably slow down the response. I think the amplifier might be stable without it, so it might be a moot point.
2. If you want to take accurate THD measurements and do a well-behaved FFT, you need to either short out C1 and C4 or set the delay_time parameter to >2s. Otherwise, there will be a DC component while voltages across the capacitors are settling. When just trying different things, I found shorting the capacitors to be accurate enough and avoids long simulation times.
1. The slew rate figure of 18-19V/us is taken before the output coil, since the coil will inevitably slow down the response. I think the amplifier might be stable without it, so it might be a moot point.
2. If you want to take accurate THD measurements and do a well-behaved FFT, you need to either short out C1 and C4 or set the delay_time parameter to >2s. Otherwise, there will be a DC component while voltages across the capacitors are settling. When just trying different things, I found shorting the capacitors to be accurate enough and avoids long simulation times.
I remember using the TL081 years ago. This was the standard FET device we kept in stock back then. Never looked much into the internals back then and didn't notice the diode. Hans Polak suggested using an LT1056 that seems without such issues. Hopefully it doesn't.
The TL081 has such a diode, just like the TL071.
The TLE2071 was meant to be an upgrade of the TL071, but apparently they forgot the diode.
The TLE2071 was meant to be an upgrade of the TL071, but apparently they forgot the diode.
IRFP240 datasheet has the body diode specs.
They look "good enough" following the idea of this design.
Regarding forward current, it looks the current would be drained by the body diode since it is lower and they are in parallel.
Body diode V=2V @ I=20A
External diode V=2V @ 5A.
If body diode drops 2V @ 20A, it will probably drop less in 5A, although I don't have the curve.
And regarding reverse recovery current, it doesn't help to parallel a faster diode.
Reverse current, if happening, will be mostly drained by the slower one (it would "hold" the current on itself).
In terms of robustness, the body diode is even better than the RFUH5TF6S.
I don't see a reason for the external diodes.
Does it make sense?
IRFP240
RFUH5TF6S
They look "good enough" following the idea of this design.
Regarding forward current, it looks the current would be drained by the body diode since it is lower and they are in parallel.
Body diode V=2V @ I=20A
External diode V=2V @ 5A.
If body diode drops 2V @ 20A, it will probably drop less in 5A, although I don't have the curve.
And regarding reverse recovery current, it doesn't help to parallel a faster diode.
Reverse current, if happening, will be mostly drained by the slower one (it would "hold" the current on itself).
In terms of robustness, the body diode is even better than the RFUH5TF6S.
I don't see a reason for the external diodes.
Does it make sense?
IRFP240
RFUH5TF6S
Not sure if we are talking about the same diode. It's adding a diode across C5 in post #73. Though not sure if the diodes you seem referring to as D7/D8 in that post are necessary as you suggest.
If the body diode comes on before the external one due to lower forward drop, the external one does nothing. In old school fast switching designs where freewheel diode Trr was critical, the solution was to put a low voltage schottky in series with the FET drain. The body diode can’t conduct, and it’s forced through the external path. Thats not really a good idea for an amplifier. IGBTs always have separate diodes, even if integrated into the same package. Newer FET designs have better body diodes, but those have such horrible Spirito effect that you would want to use them here. 1N4004’s are good enough in most amps anyway, and the IRFP240 body diode is better than that.
I might just lose the diodes then. Honestly, they are remnants of my initial BJT-based design that I just kept, because… why not? I’m going to try to keep the same form factor of the board, and real estate is getting scarce with the added drivers.
I see that the maximum supply voltage is +/- 40 V, so the voltage across the protection diode can get close to 80 V. Some brands of 1N4148 can handle 100 V, others are specified for only 75 V.
An easy fix for the first transistor current limit is a resistor in series with the collector. There are issues but it works well enough.
However, there are two hazards for the VAS. The one you have been addressing is about normal clipping but there is also the output fault condition where output protection shorts VAS current to limit the outputs, or perhaps the output is just shorted. In any case the active VAS transistor will try to drive an excess of current into the output. The approach that covers this is the VAS current sense transistor.
A Baker diode also ~prevents rail sticking and clipping instability.
However, there are two hazards for the VAS. The one you have been addressing is about normal clipping but there is also the output fault condition where output protection shorts VAS current to limit the outputs, or perhaps the output is just shorted. In any case the active VAS transistor will try to drive an excess of current into the output. The approach that covers this is the VAS current sense transistor.
A Baker diode also ~prevents rail sticking and clipping instability.
Any b
Thanks again!
Good point! The ones I have happen to be the 100V version, but there should at least be a note of it in the schematic that they need to handle >=80V. I like to have the 1N4148, because it's very bog standard. Looks like the voltage across it maxes out at about 70V, but that's getting a bit close for comfort! I think the 1N916B is 100V too. I would consider that bog standard too.
Thanks again!
Use the transistor type of VAS limiter if you’re worried about diode voltage. You can even use a Je ne sais quoi NPN TO-92. Make the board so it will take either?
