Krell anticipator circuits of the 1990s

It depends on multiple elements in the design.

The Bias point the amp runs at at idle. The speed of the output stage ( slew rate ) How long long do you hold the amp stage at a class A bias before returning to idle. There are multiple way to increase the bias but end the end it must be summed to the output to the bias circuit.  We use to design analog amps for or Seismic equipment for exploration. One method is to monitor or sample the signal, it will take have multiple stages of opamps is series then compare it to a the decision point then sum the decision to the bias to increase it current and voltage p. Of the speed of the gain decision must be quick. The purpose of was the  digital control of gain was to keep the ADC at 3/4 full scale. That is the point that an ADC is most linear. This works great as we would ranging the gain of 64 channels at a time.  I have always found to keep it simple it the best approach. I have left off details because to in theory become complex. But it should work fine if was implemented to meet the change in bias in the correct time domain.  Straight class A is is more simple of course.  I always thought my Krell KSA250 was a little harsh they must have improved but it did not have dynamic biasing.

That baloney works pretty good on the new XD lineup.

I currently have the KRELL 175XD stereo and it is supposed to be 175 watts in Class A. Though that max power in Class A is only available when demanded. Thus, the unit is not heavy and nor very hot since it is not wasting Class power when the music is not demanding it.

I went and demoed the CODA #16 amp since I was considering upgrading from the CODA #8. I felt the KRELL 175XD was better than the #8. When I heard the CODA #16, which has the first 100 watts in Class A, I could hear a very similar sonic characteristic between the 175XD and the #16. I thought the #8 sounded not as smooth and relaxed (a little grainier), while the other 2 sound smooth and relaxed. Maybe the #16 was a little clearer, like a Benchmark AHB2.

There is another amp from Japan (I think) that does something like the iBias. This is my question to the designer or distributer. The thread is a good read.

Re-imagining "Class A" Amplification | What’s Best Audio and Video Forum. The Best High End Audio Forum on the planet! (whatsbestforum.com)

Thanks for your posts.  I think so far 3.5 to 0.5 in support of my position.

@jew16384   You offer suggestions as to elements of the design that might be engineered with the objective of moving instantaneously from Class B to Class A bias.  (Put like that it sounds an impossibility).

You say ' Of the speed of the gain decision must be quick.'  Indeed it must.  What I asked is can it be quick enough to amplify a signal just heard in pure Class A.  i.e. to change bias instantaneously.  That is an impossibility.  The only way of doing it would be to buffer the signal and amplify it later. when the Class A bias had been achieved.  But that would entail all sorts of dither and clock distortions, turning an analogue signal effectively into digital when I wanted to listen to LPs.

A friend of mine has one of these Krell amplifiers, I think.  It has about 5 different ranges that are identified by lights on the front.  The example you give is very specific and it's almost always possible to find a specific example to expose a design flaw because every design has compromises.  The way I would look at it is that there could be something lost when the amplifier is first triggered to the next level and then anything within that level would be unaffected until it dropped down a level or was triggered to go to the next.  I would think that by watching the lights it would be possible to estimate the number of times that something was potentially lost during the initial instantaneous need to go up to the next level.  I think it's obvious that the sound would be expected to be something less compared to another Krell running wide open all the time, but there's reasons that make that design not ideal (though not released to the SQ most likely).

If it changed the bias in few milliseconds you will not know the difference. However can the power demand keep up, will it compress the dynamics in the outputs. The main reason for doing this is to save money in the power supply and lowering the weight and size of the chassis.
 

Example,  If a fiber is Cut on a long haul fiber network happen the system will switch to a backup fiber with in 50msec . ( telecom voice and data ) it has been proven that at most you will hear tick in that time period of 50msec to restore the path. The data is buffered and request a retransmission and it works.

In the long run you’ve added unneeded complexity to a simple amplifier to reduce cost those element I mentioned. Keep it simple stupid is the cry. But it can work, I never compared one type amp to another 🤷‍♂️