new amp... HARRY 77

[Ultima Thule]

Originally Posted by keantoken

Actually, I think C53 has a positive affect.

The voltage at the beginning of the OPS is nonlinear. When we use a Cdom cap as you describe, this nonlinear voltage injects nasty harmonics into the VAS input, when all we were trying to do in the first place was introduce linear HF compensation. The voltage at Q70's collector will still be nonlinear, but it will be 2nd harmonic in nature and won't inject so many HF harmonics into the feedback loop.

To optimize C53, it's best that the highest tolerable value be chosen for R147, as this resistor increases the effectiveness of C53. If this resistor is increased, C53 can be decreased, which will improve THD and LTP operation at HF.

- keantoken

Hi Keantoken,

have read your reply and can understand your points.
I'm not sure what the "non-linear voltage" at the base of the pre-drivers to the OPS means, maybe you are thinking of the current?
However the pre-drivers are normally biased in class A and their task while providing better drive capability of the output transistors and also to "isolate" the relative high impedive VAS output from the non-linearity especially seen in the OP transistors which normally works in class AB and have a non-resistive load to work with.

By letting the OPS to be "looser" and create higher order harmonics we have to realise the task will be taken over by the LTP to correct it anyway.
Now I'm not really sure about the following statement: "...won't inject so many HF harmonics into the feedback loop." Do you mean the local feedback loop around the VAS transistor?

The amplified Collector voltage being mostly of 2nd order harmonics is not what I tried to point out to be the problem but the non-linear intrinsic capacitance inside the transistor and by having a large Collector resistance value the voltage gain goes up and the intrinsic non-linear capacitance by nature appears to be so too and as the output of the LTP is not very low impedive it invites the distortion to occur. This is why I suggested to tie the collector to a firmer voltage node in which the designer can creates any suitable circuitry/solution as he/she prefers.

Cheers Michael

[Bigun]

Let me offer another way to analyze my proposed VAS.

The two devices are both in common emitter configuration so they will act like integrators. The first one has no gain to speak of because of the emitter resistor so it also acts like a buffer providing current to drive the 2nd VAS.

Since they both act as integrators we have a form of 'active pole splitting' which improves the overall distortion.

[keantoken]

I am writing a long post, I don't think "active pole splitting" is the only reason it lowers distortion.

- keantoken

[keantoken]

Originally Posted by Ultima Thule

Hi Keantoken,

have read your reply and can understand your points.
I'm not sure what the "non-linear voltage" at the base of the pre-drivers to the OPS means, maybe you are thinking of the current?

I mean that even though the drivers to the OPS reduce switching harmonics, they can't completely eliminate them. Cdom ignores OPS current, instead it follows VAS output voltage and injects local NFB from there. Any switching harmonics at the OPS input are injected through Cdom into the VAS, which we certainly don't want. We're only trying to introduce a phase shift to compensate the amp with. These switching harmonics the LTP has to deal with, and though the LTP works well it doesn't work so well at HF, amplified by the power needed to drive correction through the monster OPS AS WELL as Cdom; this puts an incredible amount of strain on the LTP. By referencing Cdom to the OPS input we are actually defeating the LTP's ability to correct for these harmonics because any such correction escapes through Cdom unnecessarily.

However the pre-drivers are normally biased in class A and their task while providing better drive capability of the output transistors and also to "isolate" the relative high impedive VAS output from the non-linearity especially seen in the OP transistors which normally works in class AB and have a non-resistive load to work with.

The source impedance into the OPS is generally very high, so even though the driver stage might decrease current draw, the net output impedance is not decreased (I don't see a problem here with the bootstrap CCS though). For example, a 10k resistor from OPS input to ground would make for lower source impedance because it provides a path for the base current of the drivers to go to ground if they don't agree with the VAS. Otherwise, base current caused by reactive AC characteristics of a speaker would "push" the input voltage of the OPS up or down, leaving the LTP to correct for this. We want the LTP to correct only what we can't correct ourselves with good design.

By letting the OPS to be "looser" and create higher order harmonics we have to realise the task will be taken over by the LTP to correct it anyway.

Sure, but the LTP isn't perfect at HF. Also, reducing switching harmonics increases stability (and reduces the "harsh" sound). I don't see how the OPS is "loose" though. Also, I believe we should make an effort to make all stages as linear as possible before adding NFB; because the LTP injects its own form of distortion - everything it has to correct is inadvertently added to the signal because of the varying base voltages with LTP current.

Now I'm not really sure about the following statement: "...won't inject so many HF harmonics into the feedback loop." Do you mean the local feedback loop around the VAS transistor?

Any interference between the stages is included in the feedback loop, and for stability we have to reduce NFB at HF, decreasing the LTP's ability to correct for HF interference injected by the OPS. I consider it "injected" because it is a nonlinear distortion.

The amplified Collector voltage being mostly of 2nd order harmonics is not what I tried to point out to be the problem but the non-linear intrinsic capacitance inside the transistor and by having a large Collector resistance value the voltage gain goes up and the intrinsic non-linear capacitance by nature appears to be so too and as the output of the LTP is not very low impedive it invites the distortion to occur. This is why I suggested to tie the collector to a firmer voltage node in which the designer can creates any suitable circuitry/solution as he/she prefers.

Cheers Michael

I wouldn't know how much of a problem this is, but it depends on the transistor. A lower Cob transistor would have less distortion, through the 5401 is already quite fast. In any case I think this distortion is more desirable than my aforementioned switching harmonics.

Imagine that we delete C8. Now the only way for voltage switching harmonics to enter the circuit is through Q7's Ccb. We have already eliminated half of our problem. The rest of the switching harmonics are current related.

Also, I think the standard usage of Cdom is somewhat ineffective. it is current draw, more than voltage output, which causes stress on the output stage. Cdom references voltage output, not current output, and yet the whole LTP and VAS is a current amplifier (meaning that any stability issues caused by more or less current flow, which affect the LTP the most, are ignored by Cdom since the output voltage will always be the same regardless of the load).

With Cdom referenced to Q7's collector, Cdom is more effective when more current is drawn. So say we put a 470n cap across the output. Output current will increase, so VAS current will increase, and therefore Ic(Q70) will increase, increasing the voltage swing at its own collector, increasing the amount of feedback Cdom provides to its base.

This technique works very well for class B amplifiers, at least in the simulator.

- keantoken