I attempted to overcome the headroom issue with a CFP using lateral MOSFET output devices. The normal CFP circuit limits the driver excursion to the gates/bases of the output devices, but lateral MOSFETs have maybe upto 10V of Vgs which restricts the swing for a CFP arrangement. The normal CFP-with-a-little-gain arrangement dissipates quite a lot of power in the feedback resistors as they have to be low impedance (they go to the emitters of the driver transistors).
So I arranged feedback to bases of differential pair drivers, R6, R7 above - higher impedance and only one divider needed.
The emitter tails of the pairs need more current when the relevant output device is conducting more current - but the low transconductance of laterals means the Vgs is a reasonable proxy for this current, and this appears across the 150R CFP resistors R30, R33. So the voltage across these resistors is converted to a current by 10k resistors R1, R2, and bounced off current mirrors, which multiples the current about 45 times before being fed to the relevant emitters. Note that this method only responds to current, independent of the output voltage, so should handle reactive loads well.
This current feedback loop might be unstable unless the system is biased into conduction appropriately with R12 - at least this works in simulation (more analysis needed) - and this can be adjusted to set the output device currents - it would be better to have an adjustable current source/sink rather than resistance for this to be independent of supply voltage.
Diodes D1 and D2 prevent reverse biasing the transistors in the pairs during clipping, the 470pF C1 and C3 prevent oscillation (not sure how that works yet!), and R17 limits clipping artifacts.
The actual choice of transistor for the differential driver pairs needs to take account of their dissipation, I've must modelled with small-signal devices. And lots of component values can probably be optimized.
Here I wrapped an LTC6090-5 high-voltage opamp around as input and VAS.
As laterals are used there's no need for source resistors to further improve headroom.
Whether this is a workable scheme in reality I don't know yet, but I hope to build this up at some point. It probably already exists of course, nothing new under the sun, but I didn't seem to find much about CFP's with gain using laterals when I searched.
Big fan of your arrangement! In particular your kindof active LTP biasing. Ive tried many versions of TLP as mosfet drivers. Problems always occur when the drivers run out of current. Got any more plots and maybe fourier analysis?
Much cheers
Ruben
Much cheers
Ruben
A related BJT circuit that you may find interesting. Note that the THD is respectable while the OP is ~class-B. The cross-fed class AB drivers deal with need for driver current. This version requires a PNP power transistor with good gain, ie MJ2955 does not work well.
Alas I've not built the circuit in my original post so have no information on its real world behaviour, though maybe I still will get round to this.
However I have come up with another take on this general idea:
This idea stemmed from trying to arrange a CFP for lateral FETs that didn't compromize headroom like the naive approach, but turned into another differential-pair based output stage with gain, here a gain of 3. Each differential pair is independently fed from a current source. The biasing is odd in that increasing the bias voltage across R18 reduces the bias current. This I have actually built and measured its performance (limited by a 0.5A PSU though).
The emitter followers on the FET gates allow the rest of the circuitry to run at lower currents and not need heatsinked transistors. Q9 and Q10 might need heatsinking under heavy drive, but I'll need to drag out my big PSU to test that anyway.
Clearly the normal quoted figure of 100mA for lateral FET bias applies well as the yellow and violet traces show for 8 and 20 ohm loads respectively.
The 0.5A current limiting supply accounts for the steep rise in the 8 and 20 ohm plots. The x axis is the input voltage to the circuit, so the actual level at the load is 9.5dB higher than this.
The next experiment with this will probably to wrap a feedback loop around with an opamp run at +/-18V which given the 3-fold gain should be adequate.
However I have come up with another take on this general idea:
This idea stemmed from trying to arrange a CFP for lateral FETs that didn't compromize headroom like the naive approach, but turned into another differential-pair based output stage with gain, here a gain of 3. Each differential pair is independently fed from a current source. The biasing is odd in that increasing the bias voltage across R18 reduces the bias current. This I have actually built and measured its performance (limited by a 0.5A PSU though).
The emitter followers on the FET gates allow the rest of the circuitry to run at lower currents and not need heatsinked transistors. Q9 and Q10 might need heatsinking under heavy drive, but I'll need to drag out my big PSU to test that anyway.
Clearly the normal quoted figure of 100mA for lateral FET bias applies well as the yellow and violet traces show for 8 and 20 ohm loads respectively.
The 0.5A current limiting supply accounts for the steep rise in the 8 and 20 ohm plots. The x axis is the input voltage to the circuit, so the actual level at the load is 9.5dB higher than this.
The next experiment with this will probably to wrap a feedback loop around with an opamp run at +/-18V which given the 3-fold gain should be adequate.
Very interesting
CFP with feedback was a obsession of mine when
first starting for some reason.
And it seems like a very cool way of doing it.
specially reducing temp of feedback resistors.
Maybe even lower THD with differential
been awhile since I have done sims.
relaxed a bit on computer land.
But...this circuit is encouraging
to play around with.
Stability is the question.
But that is the fun.
Good one
CFP with feedback was a obsession of mine when
first starting for some reason.
And it seems like a very cool way of doing it.
specially reducing temp of feedback resistors.
Maybe even lower THD with differential
been awhile since I have done sims.
relaxed a bit on computer land.
But...this circuit is encouraging
to play around with.
Stability is the question.
But that is the fun.
Good one