Motorola 1967 AKA P3a amplifier to be enhanced, but how?

[Hayk]

Before I retire from amplifiers, I am reviewing the great sounding amplifiers.
Bellow is the circuit of the original Motorola 1967 AES published and Rod Elliot's famous P3a.
This circuit can be enhanced to give either lower distortion than 0.04% or make it sound more dynamic by increasing the even harmonics.
I will go in both directions and select the one that has the best character.
First, this amp has the power transistors bias problem to be resolved.

 

[Hayk]

This is the output stage simulation of P3a. 1A peak 10khz triangular current is injected to the output to show the output impedance. This what a tweeter sees.
The Motorola has a mutual capacitor 2nf between the power transistors bases. Without this capacitor, at high frequencies or square waves, the outputs shoot through as the other doesn't unswitch fast.

 

[Hayk]

First trial. No more shoot-through. Bias 90ma instead of 25ma.

 

[Hiten]

Thanks for doing this. P3A is very popular DIY amplifier. Didn't knew it was Motorola based.

 

[Hayk]

I am persuaded that Elliot didn't knew neither. The P3 was derived from Douglas Self's blamless CFP version but with bootstrap, with P3a, he used PNP VAS instead of NPN. This inverts the phase of the second harmonic and makes it more vivid. This how he recreated without being aware, the Motorola 67 circuit.

 

[Hayk]

I followed the Motorola circuit but I added a simple current mirror.
For 27w it results 0.0025%THD for 1khz. The compensation is 2 pole roughly adjusted. Adding a CCS doesn't act much.

 

[Hayk]

The slew rate was catastrophic with compensation on the VAS base. So I had to innovate by using the current mirror as common base. Thus, I got now 16v/us falling and twice on rise.

 

[Hayk]

I replaced the input transistors with low noise high gain ones, now the THD is 0.0015% for 27w.
This is the final circuit, tomorrow I can start the PCB.

 

[Hayk]

I am persuaded that Elliot didn't knew neither. The P3 was derived from Douglas Self's blamless CFP version but with bootstrap, with P3a, he used PNP VAS instead of NPN. This inverts the phase of the second harmonic and makes it more vivid. This how he recreated without being aware, the Motorola 67 circuit.

I may be all wrong, he may copied the Motorola. There is a strange capacitor on the lower driver in both circuits, 50pf on Motorola and 100pf on P3a.
On simulation, it makes the fall edge go with spike. Probably the early complementary transistors didn't had identical speed, so it matches them.
By the way, the spice model of preceding input transistors had an issue, with these ones, it requires less compensation, so faster slew rate of about 50V/us.

 

[Hayk]

I adjusted the 2 poles compensation, The 10khz THD is 0.0035% while the 1khz became 0.0009%.

 

[Hayk]

Another way of biasing the power transistors is to use for each seperate bias so that it doesn't cutoff. The bias current follows the temperature of its transistor by a standard 10k NTC that can be screwed along the output transistor. This should result in finer high frequencies.

 

[Hayk]

Ready for PCB. It has 2 separate bias adjustment. With the main bias Rv1, zero, the power transistors are biased 60ma by Rv2, then increased to 85-90ma by the main bias. This 60ma will be the minimum current instead of cutoff.
The mjl21193-4 at low currents, have very low Hfe about 8-10 when measured by multimeter, is the case for most old power transistors. They must be slow transistors as they assure the stability of the CFP.

 

[Hayk]

Preliminary PCB.
The drivers and voltage multiplier are aligned with some back space to be mounted on a small heatsink.
I will name this amp, M67b.
5 units from JLCPCB will cost $4.

 

[Hayk]

The TTC/A004b are sold in pair and rather expensive. To reduce the cost, I replaced the voltage multiplier by BD139 and the VAS by KSA992.
Further adjustments, The THD,1khz 25w is now 0.0007%.

 

[Hayk]

I simulated an error corrector with this amp. The minimum possible THD is 0.00015%, instead of 0.0007% without.

 

[Hayk]

Spectrum of input and output, only noise.
I will implement the EC and let the user decide to wire it.

 

[Hayk]

I reviewed the adjustments of the input stage. I replaced the feedback network with Duo-beta. After replacing the VAS transistor, I didn't re adjust for its function. The decreased its load by increasing the bootstrap resistor from 1k to 2.7k. Increased the LTP current from 1ma to 3.5ma. P3a had input 1k resistor at the base, I suppressed it.
Now the THD is 0.0004% but at 22v out instead of 20v , with 1k generator impedance.
I will implement a servo with TL071, in case there happens offset drift. For now the offset is adjusted by R11 mirror resistor.

 

[Hayk]

Something wrong on positive falling slope.

 

[Hayk]

I got banned from engineering discussions on this website, so this thread is discontinued.

 

[Power Pop 23]

Per Amir,

4 transistor class A very low THD 40w amp.

I have designed long ago, a type of SRPP MOSFET output amplifier, that has ended after several models to https://www.eevblog.com/forum/projects/4-transistor-40w-class-a-0-0004-thd-amplifier/ How it works. The Q1,Q2,Q4 don't need explanation. Q3 is biased by a voltage across 470uf capacitor Vc...

www.audiosciencereview.com