A superior VAS clamp and VAS current limit for the Blameless.

[GK]

Anyone out there who has tried out the Blameless with two-pole compensation may have noticed that the amplifier then develops horrible, asymmetric squarewave, clipping and clipping recovery performance.

This is due to a number of issues - here the mods and additions I have made to address all of them.

Firstly, a voltage clamp consisting of D3-D6 and transistors Q10 and Q11 is added which prevent saturation of the VAS transistor Q7 and its current source Q9. This comes at only a small cost to the output voltage swing.

Secondly, the LTP is provided with a 3-transistor current source and the VAS has an emitter degeneration resistance of 24 ohms. In this case the VAS standing current is 9mA and the LTP tail current is 5mA.

The result of this is that the quiescent, or DC voltage at the collector of each long tail pair transistor is equal.

This allows two back-to-back limiting diodes, D1 and D2, to clamp the maximum AC drive voltage to the VAS emitter follower symmetrically at +/-700mV.

This has two very important benefits. Firstly, it makes it impossible for the current mirror transistors to saturate. Secondly, the +/-700mV limited output swing automatically provides a current limit for the VAS, being equal to 700mV/24 ohms (R17), or ~30mA.

[GK]

Here is the result - perfect clipping performance at 20kHz!

[GK]

The next plot shows the collector voltages of the LTP transistors during clipping. The green trace is the collector voltage of Q1 and the pink trace is the collector voltage of Q2.

The red trace is the positive supply voltage. Note how D1 and D2 limit the voltage swing and how neither of the current mirror transistors can saturate.

[GK]

This plot shows the VAS current along with the output voltage. As can be seen, the VAS current is limited to a maximum of 30mA as predicted.

[GK]

And here is the 20kHz square wave response. Note that the overshoot inherent with 2-pole compensation is eliminated by the 500 ohm / 1nF input filter that limits the rise time of the input signal.

[GK]

And finally here is a teaser pic of a complete design using this 2-pole input stage and VAS. Using a pair of optimally biased output transistors (class B) wrapped up in a CFB output stage it is rated at 100W into 4 ohms and simulates as follows at 20kHz, 50W into 8 ohms:

Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 2.000e+04 2.828e+01 1.000e+00 -2.95° 0.00°
2 4.000e+04 1.939e-05 6.855e-07 179.75° 182.70°
3 6.000e+04 1.433e-05 5.065e-07 -150.49° -147.54°
4 8.000e+04 3.288e-06 1.163e-07 -10.90° -7.94°
5 1.000e+05 5.860e-06 2.072e-07 35.99° 38.94°
6 1.200e+05 4.653e-07 1.645e-08 -142.54° -139.58°
7 1.400e+05 3.085e-06 1.091e-07 -123.61° -120.65°
8 1.600e+05 1.149e-06 4.061e-08 -57.36° -54.40°
9 1.800e+05 2.595e-06 9.175e-08 -75.79° -72.84°

Total Harmonic Distortion: 0.000090%


The global unity loop gain frequency is 850kHz and the phase margin is 80 degrees.
I’m currently laying out the PCB for this one and will reveal the full circuit on my WWW when it’s up and running, along with a few other amps that are currently under construction.
This non-class-A one will be my cool running summer amp.

[pooge]

Originally posted by G.Kleinschmidt


Secondly, the LTP is provided with a 3-transistor current source

current mirror?

[GK]

Originally posted by pooge


current mirror?

Doh!

Yes, current mirror

It's after 11pm here.

[forr]

Hi G. Kleinschmidt,

Are Q10 & Q11 and R19 & R16 really useful ?

I tested a somewhat similar configuration more than twenty years ago, inspired by the DB Systems DB6 amplifier.

I think that, keeping Q10 and Q11, the current through them could be used to used for a clipping indicator.

I don't remember exactly what was my circuit, I think part of the current through Q10 and Q11 returned to the inverting input.

[jcx]

looks like an adaptation of Cordell's front end protections to the Blameless

http://www.cordellaudio.com/papers/MOSFET_Power_Amp.pdf

you might want to point out the care needed to keep 0 V across the diff pair collector calmp D1,2 and that even with even mV bias the diodes have an impact on openloop gain with the hi-Z buffered input VAS (Pease "Trouble Shooting Analog Circuits")

perhaps the (reverse)biased diode bridge clamp I've shown in my composite op amp thread could reduce the impact of small Vos across the front end clamp:

http://www.diyaudio.com/forums/showt...806#post512806