Simple Feed Forward for Power Amp

[Bonsai]

Here’s a simple feed forward (F/F) concept that reduces distortion by about 15dB at 20kHz. I’ve shown it used on a CFA topology amp here but don’t see that it would not bring similar benefits to VFA as well.

Distortion sim results for 20kHz at various output powers into 8 Ohms:-

Output Power 8 Ohms No F/F With F/F
6W 21ppm 21ppm
75W 67ppm 14.6ppm
225W 186ppm 30ppm
272W 237ppm 43ppm
300W 338ppm 72ppm

At 3 Ohm output load (640W 62V pk output) the distortion is 46ppm with F/F and without 355ppm – an improvement of 17.75 dB

At 300W out, the output is swinging to within about 2V of the rails – i.e. just entering clipping.

So, you can see it only does the job at high powers, exactly where its needed. Because the opamp is running in its small signal region – the voltage at Vb is about 2Vpk to pk - you don't really need a blindingly fast device for the job.

The sim just used an LT1057 high speed JFET input op-amp that was available in the LTSpice library. A more thoroughly worked design would use something better and probably run the F/F amp in class A for that last touch of refinement, but this needs more investigation.

The diodes between the opamp output and inverting input just ensure when the amp is overdriven, there is feedback around the F/F control amp it. This ensures a fast recovery as the power amp comes out of saturation (which on a CFA power amp is very fast) with no overhang caused by the F/F control amp.

When the amp clips with the current F/F control amp circuit, a large DC offset (3-4 volts) develops – I still need to do some work on that bit to sort it out. Under normal operating conditions, provided there are no DC offsets on the input, the F/F control amp will also act to remove output offsets.

Please note, this is a conceptual circuit, so not all the bugs are ironed out.
Attached Thumbnails

 

[knutn]

You are comparing the input voltage with a fraction of the output voltage, so this is feedback, not feed forward.

 

[Bonsai]

Yes, it could also be called feedback.

 

[sreten]

Yes, it could also be called feedback.

Hi, then its not "simple feedforward", rgds, sreten.

 

[Bonsai]

I am very happy with the c. 15dB of distortion improvement at 20 kHz

 

[lineup]

Bonsia you could post a a clearer picture of schematic.
It is too small to see the details.

 

[Bonsai]

Hello Lineup. Sure but it will be a bit later.

 

[MiiB]

Interesting....to use an op-amp to make corrections to a circuit..some good thinking in that...

 

[catalin]

Hello Bonsai ,congrat's for the design ,I just want to say that we are looking too much at the thd figures while the TIM distortion is more important than thd .
1% of thd is not audible but 1% of TIM is easy audible .
I have designed also an sziklai output amp with VFB+CFB with under 10ppm distortion at 20khz and the thd20 is not incresing more than 30ppm in all power levels.Also the CCIF IMD is bellow 10 ppm .
But I am not sure that this is the best sounding yet ,to reach milions per part thd .

 

[Bonsai]

Yes, I agree that low THD has to be considered in context. The circuit I posted above is really a kind of helper. Most amps start to show greater non linearity from about 6-9 DB blow full output voltage swing into a load which is to be expected, unless of course you have very high loop gains. The open loop of the power amp above is about 78 dB and the closed loop gain is 34 dB so not as high as some designs which are up at 50 dB.

(give another hour Lineup and I will post a bigger version up)

 

[Bonsai]

Here is a larger version of the circuit Lineup

 

[Technicalairgap]

I know its an old thread, but the use of non-inductive resistors in the CFB voltage divider with factorially improve both THD, and IMD of the real product.