Patchwork Reloaded: Circuit Optimization and Board Layout.

[MJL21193]

I have had several months of listening to the original amp . Although I didn't put it in a case, I used it every time I did any work in my lab. I have had a few different amps to compare to it, mostly DIY, but none have been it's equal, including my Yamaha htr.
This may be due to it's high power - I don't have anything else that can put out more than 100 watts. The difference is mainly in the bottom end - there is more authority to the bass, it seems more solid. This could be a delusion on my part, but I'm usually tough on my creations, not giving credit where it's not warranted.

In any case, I have a six channel amp project planned and I thought I would put my effort to use by using this design for the amp modules. I set about refining the circuit, subtle changes to improve performance, lower distortion. I have changed the output devices to MJL21193/94, as I have many of these and would need 12 of each for the six modules. Other changes that I'll be going through and looking for advice on.

The new schematic:

[MJL21193]

The changes...

I started by not using the LM394 super-matched pair. As nice as these are, they are pricey. It would have been more than $60 alone for 6 of these. I went with the BC550 and kept the rest of the front end intact, except for some resistor values.
I played with multisim trying different things to either improve performance or reduce component count. I recently learned how to run the fourier analysis which will extend the precision of the THD figures. This has made a big difference.

Here's a look at it as it was before. This very good performance was proven on a real distortion analyzer last January.

[MJL21193]

In the front end I increased current by changing R27 from 20K to 12K. I increased R2 and R3 to 360 ohms and lowered the emitter resistors to 33 ohms.
This has ~4.3mA through the front end.

A slight improvement as shown below, about .0001% less. Every little bit helps, right?

[MJL21193]

Further changes were to more than double R4 from 7K to 15K. I deleted R6, increased R5 to 4.7K up from 1K. I deleted the emitter resistor of the VAS (R7 in the old schematic) as unnecessary. Interestingly, removing this 10 ohm resistor made an impact on the THD.

A further overall reduction:

[MJL21193]

Here's one I didn't expect. Changing the drivers load resistor from 55 ohm to 150 ohms reduced THD by .0002% alone.
R28, "bootstrap" resistor is now 450K - the best value. A reduction of .0001% at the test power (~56 watts into 8 ohms) and nearly .0003% at 20Khz - incredible really.

I also spent some time trying a previously attempted feedback to the cascode. This did nothing but increase distortion, so I guess I exercised prescient vision when I axed it before.
The result of these changes is an overall reduction in THD by more than .0003%. If this is accurate in real terms is debatable. Whether this reduction is audible is highly unlikely, but there were several improvements made beside the THD figure. Parts were eliminated - this is good. Current through the drivers was reduced without adverse effect (verified with the real amp, no problem into a low impedance load).

Thoughts on my changes? Ideas for more?

Last of these boring graphs:

[megajocke]

Why such a low voltage for the input stage cascode? Weird stuff might happen on input overload.

Output stage base resistors can probably be reduced from 10 ohms making current sharing less dependent on hfe.

[ferencz]

Nice design!

And nice THD graphs... But hey, 0.0006% vs. 0.0005% vs. 0.0004% tells nothing improvement. And the THD alone is not enough to desribe the quality of an amplifier!
I'd happy to see some square-wave & slew-rate & IM measurements, if they are available. Maybe there is room for a real improvement there!

[MJL21193]

Originally posted by megajocke
Why such a low voltage for the input stage cascode? Weird stuff might happen on input overload.

Output stage base resistors can probably be reduced from 10 ohms making current sharing less dependent on hfe.

Hi mega,
I have 3.87Vdc now from the 2 standard green LEDs. I went with this to avoid a zener in this position (noisy?).
What do you suggest to get the voltage up? Another LED maybe OR I have some high intensity blue ones that have a forward voltage of ~2.75. I will try those in the prototype.

I was going to experiment with deleting the base stoppers to see if the design is stable without them.

Do you have any other suggestions? My devices are not fancy, popular choices around here. My outputs are rather slow at 4Mz.

Originally posted by ferencz
Nice design!

And nice THD graphs... But hey, 0.0006% vs. 0.0005% vs. 0.0004% tells nothing improvement. And the THD alone is not enough to desribe the quality of an amplifier!
I'd happy to see some square-wave & slew-rate & IM measurements, if they are available. Maybe there is room for a real improvement there!

Thank you!
Lacking in depth knowledge, I use the THD figure to tell me if I'm doing something completely stupid. I figure if it goes lower, I'm doing something right.

I did some testing of the original amp in the first thread. Some of the results start
here.
A few square wave shots too.

When I get the prototype for this one road-worthy, I'll do some testing.

Having lived with the original for a few months, I can say that it is a fine sounding amp. Fast and accurate. Powerful and very stable. I have made some changes on this one, but nothing (I hope) that will undo my good results from before.

Do you have any suggestions for possible improvements?

[ferencz]

Do you have any suggestions for possible improvements?

Your amp is already a "state-of-art" design among the amplifiers using the "Lin" circuit topology.
But, as on every side of life there is always room for improvement!

The only mod I could recommend now is to change the drivers to MJE15032/33. The reason is that in practice I found that related the MJE15030/31 pair there is always a huge Hfe difference between PNP and NPN device.

PS:: I visited the quoted thread. You made excellent work! Excellent!
Anyway that's a great thread with many great informations!

[megajocke]

Originally posted by MJL21193


Hi mega,
I have 3.87Vdc now from the 2 standard green LEDs. I went with this to avoid a zener in this position (noisy?).
What do you suggest to get the voltage up? Another LED maybe OR I have some high intensity blue ones that have a forward voltage of ~2.75. I will try those in the prototype.

I was going to experiment with deleting the base stoppers to see if the design is stable without them.

Do you have any other suggestions? My devices are not fancy, popular choices around here. My outputs are rather slow at 4Mz.

I'd probably use zeners anyway as this noise will have a pretty hard time getting into the circuit and be amplified as the differential stage has very good rejection when voltage changes the same on both collectors. Another option is just using a resistive divider with a bypass cap. I'd probably like to have a little more voltage than possible with LEDs, at least 12-15V or so.

I'd probably reduce the output stage base resistors to 2 ohms or something similar. It might even be possible to remove them completely like you say but I believe the only thing gained is a lower component count.