[Spacehead]

Quote:

Originally Posted by hiker101 /img/forum/go_quote.gif That buffer is meant to be wrapped in a global feedback loop- that's why it's suggested as an option for the PiMeta. It's not clear how your circuit is set up in that regard. It's also not clear what you're doing about the ground, maybe that's the part of the circuit that's current starved. You said you had another class AB buffer working in the same circuit- was it a discrete buffer, or something else?

ground is buffered with OPA633, it has plenty of current capability. Sijosae Class AB discrete buffer (bc337&bc327) works very well in this circuit, which is Cmoy style, and buffers are inside feedback loop (wire is as short as possible)

I am trying to study Walt Jungs Diamond buffer article http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_2.pdf

My "input" resistors are now 10.9 ohm , emitter resistors 3.8 ohm, constant current resistors 620 ohm
between op amp and buffers there is 200 ohm resistor. I have decoupled buffers with 100 uF capacitor rail-to-rail.
I tried and added red leds to BD139 and BD140 base and another end to output, it didn't help.
What disturbs me most, is that BD139 and BD140 don't get warm at all. They are actually cold. I wonder if it would help to lower that current source resistor again to 200 ohm. But ?! hmm, could the JFETS (I used 2SK117) limit the current ?! If removing them would release the circuit so that there would be enough current for 32 ohm headphones (40-100 mA)?

 

[FallenAngel]

Not sure why you want to limit current there, you're essentially just setting CSS ("constant current" for Class-A operation), more will be drawn if needed in Class-AB.

 

[Spacehead]

Quote:

Originally Posted by FallenAngel /img/forum/go_quote.gif Not sure why you want to limit current there, you're essentially just setting CSS ("constant current" for Class-A operation), more will be drawn if needed in Class-AB.

I just wonder why at 7.5V rail voltage the buffer doesn't work properly, gives distortions to bass. Then at 10V one of the smaller chips, usually BC337 goes to thermal runaway and heats up until burn out.
It isn't stable then?

 

[FallenAngel]

Hmm... could be an issue with implementation?

 

[Spacehead]

Quote:

Originally Posted by FallenAngel /img/forum/go_quote.gif Hmm... could be an issue with implementation?

have you checked in simulation its frequency response?

 

[FallenAngel]

Unfortunately, no, that's too techie for me.

 

[Spacehead]

Quote:

Originally Posted by FallenAngel /img/forum/go_quote.gif Hmm... could be an issue with implementation?

I couldn't find any problems.

Then I built that. It works, but isn't as strong as Sijosae Class AB buffer.

All these lose to single Class A BJT or Mosfet.

I am trying to learn what controls the bias of transistor, but very little has come clear to me.

 

[cobaltmute]

To my understanding, the bias of the output stage is controlled by the CCS (JFET + resistor).

 

[Spacehead]

Quote:

Originally Posted by cobaltmute /img/forum/go_quote.gif To my understanding, the bias of the output stage is controlled by the CCS (JFET + resistor).

I just cannot get output transistors warm. I have tried many source resistors, 3K, which was worse than 620 ohm, 200 ohm, 80 ohm, 50 ohm, 470 ohm. None worked without distortions in bass. Transistors got hot when I connected gate directly to source on 2SK117 JFET but it also gave out distortions.

I cannot get it working. You try it!

 

[Spacehead]

Quote:

Originally Posted by Spacehead /img/forum/go_quote.gif

I did two of those for stereo. I used smaller resistors R2, R5 = 100 ohm, R3,R4 = 4.7 ohm, R6,R7= 3.3 ohm and changed Q3,Q4 to BD139,BD140

It took 2.5 hours.

First I had input and output shorted on other, and one BC327 burnt out. No more problems after changing that and clearing the short.

Those look cool and they run (too) cool

In Walt Jung´s article http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_2.pdf
R4 and R7 ratio is 2X , mine is now 1.42X
I don't know if want to bother with that, I don't have 2 ohm resistors and there isn't room for two, except underside.

These sound good and they are almost identical, transistors are matched.

 

[FallenAngel]

Very cool, congrats.

So... which buffer do you like most?

 

[Spacehead]

Quote:

Originally Posted by FallenAngel /img/forum/go_quote.gif Very cool, congrats. So... which buffer do you like most?

Thanks
This new one is absolutely best. Best bang for the buck is for sure basic Sijosae Class AB BUF634, it has clear sound and can deliver good currents for proper bass. Soundstage is definitely best on this diamond buffer. Sijosae BUF634 Diamond Buffer is third of these, it is slightly weaker. Then I have tried single mosfet and single bjt (Poor Man's Class A BUF634 - ÒôÏìDIYÂÛ̳ - HIFIDIYÂÛ̳ - Powered by Discuz! They are strong but lifeless. OPA633 is a diamond buffer too, but sounds too much like op amp, and like OPA2132 which I don't like. I couldn't yet get your diamond buffer working, maybe 75 ohm resistor needs to be 20 ohm, cool idea though using those CCS JFETs.
These buffers don't color sound that much as op amps so it is difficult to distinguish differences. Now I have AD8022 on my X-FI and AD8620 on this amplifier. That means very neutral sound.
I haven't tried Monofied Sijosae BUF634 yet, if I would get your thing working it would be it.

Now I am afraid that again one BC327 would get burn out, when I was having a nap other buffer had gone dark, fortunately there was no DC offset. I fixed that again and changed both input transistors, I hope it stays good now. It is PITA changing transistors and trying to figure out where there is a short if there even is. Transistors can come bad from factory, too? BTW. when transistors are pretty closely matched, DC offset is 00.00 mV under op amp feedback loop

Edit:
I had forgotten to isolate the dip8 socket from the protoboard so I guess it was shorting that transistor. I noticed that when I had trouble getting a good connection, leds didn't light up even I pressed it hard. I hope that electricity tape solves that problem and I would have a reliable and powerful diamond buffer.

Edit2:
As Suggested by Walt Jung I bypassed the buffers with 470 uF capacitors. Now leds are even colored and bass response is like it should be. I am quite happy to those buffers.

 

[Spacehead]

Quote:

Originally Posted by joneeboi /img/forum/go_quote.gif

Then I tried that, but with diodes and BD139/BD140, 2SK117, 620 ohm source resistors, and emitter 4.7 ohm. I omitted 220 ohm resistor as that is in my circuit already. Didn't work, distorted. Maybe some problems with shorts as I used old piece of protoboard (the one with FallenAngel's diamond buffer). I hate to do things that don't work. It feels bad.

 

[FallenAngel]

Want to try to build one of these?
discrete Diamond Buffer - Design

 

[Spacehead]

Quote:

Originally Posted by FallenAngel /img/forum/go_quote.gif Want to try to build one of these? discrete Diamond Buffer - Design

I would love to try it, but I am missing all those transistors. I am getting 2N5486 next week though and maybe I order 2N5088/87. MJE243/MJE253 can be replaced with BD139/BD140 I guess.

That Walt Jung buffer ran out of current with my 32 ohm Sennheiser HD438. It distorted at high volume much easier than pure class A BD139.
That is kind of weird because it has BD140 driving the load too. Again I don't understand enough to calculate maximum safe resistors values. Optimal would be if BD139/BD140 would dissipate about 1W in free air, then bias would be right.
Now these class A BD139 Poor Man's Class A BUF634 - ÒôÏìDIYÂÛ̳ - HIFIDIYÂÛ̳ - Powered by Discuz! have 82 ohm emitter resistor, which means (ohm's law&power) 1.21W dissipation. And they don't yet need heatsink.

I am getting PPA V.2 PCB next week too. It has a nice diamond buffer. But largest task is getting all the parts for it. Do you have any spares you could share? It is my first PCB project