[G.Trenchev]

Hi,I've got some BUF634s,and I want to make an amp with them.Is this schematic ok ?
 

 
Should I bias the buffer in class A with the 150-ohm resistor,or it's unnecessary?

 

[Avro_Arrow]

I've never seen anyone try and bias a BUF634 into class A before.
I'm not sure it would do much good. The effect might only be noticeable
at the very lowest of listening levels.
 
Have a look at the PIMETA. It's pretty close to what you have drawn.

 

[jcx]

the 1st question about an amp design should be - for what load? - which headphones; Z, sensitivity
 
a output stage is only Class A for a certain range of signal and load - knowing both is necessary - for 25 Ohms shown the output is only
Class A for 1.2 Vrms - OK for sensitive iem - but then the gain of +4 may be too much
 
if you need more output V for your headphones this could be viewed as a variation on Self's "Class XD"
 
the LM4562 is very fast at  55 MHz GBW - you may need a local feedback cap and considerable layout care to keep it stable, R1 is about 10x too big too for good stability

 

[G.Trenchev]

The "Load"is on the schematic - 25 Ohms.
I've never had stability problems with LM4562.
 
~20 mA output current is enough.The gain is fine for me at 4.
I'm pursuing lowest distortion possible.
What about the WB mode on 634 ?
Any difference in audio ?
 

 

[AudioCats]

Quote:

I've never seen anyone try and bias a BUF634 into class A before.
I'm not sure it would do much good. The effect might only be noticeable
at the very lowest of listening levels.
 
Have a look at the PIMETA. It's pretty close to what you have drawn.

My little desktop amp has class-A bias for BUF634's. The design is a pair of opa627 driving four BUF634's, pretty much a basic data sheet circuit but I added 2mA class-A bias option for the opa627 and 30mA class-A bias option for the BUF634. The biasing is done using j-fet CCS connecting to the - rail.
 
the sonic effect of class-A biasing, in this amp:
-- 2mA bias at the opa627 (between the opa627 output and BUF634 input): sounds more hazy, flat. No way I could learn to like it, so I took the 2mA out.
-- 30mA biasing at the BUF634 output (done using 3 j-fets in parallel, each burns about 10mA): sounds is more open, less congested. So I kept it in the amp.
 
I think it will sound even better if I can bias it to 50mA, but the BUF I am using are dip-8, they got a bit too warm even at 30mA biasing. If I have to build a new one I'd use the BUF in TO220 package, parallel 4 per channel and add a 100mA CCS at the output.

 

[nikongod]

I would not bias the buf-634 like you did. If the BUF634 goes into thermal protection mode and cant supply the current to keep the output at 0V the amplifier may become extremely unstable, with the outputs possibly latching onto the negative rail. Just use the bandwidth pin. 
 
Good choice to use the TO220 version of the BUF634. It is mucho nicer than the DIP8 version

 

[AudioCats]

I do also have protection circuit (diode strings network, using diodes' forward drop to limit swing to about +/-3V max), though the actual purpose is more for controlling max volume. If the BUF's do go bad, the 3V max will not fry my phones (600 ohm Beyer and 100ohm MB Quart). In my case, the sonic benefit of class-A biasing is very worth while. YMMV.

 

[kevin gilmore]

These days there are a number of buffers that sound better than the buf634.
This is one of the better ones.
http://www.national.com/pf/LM/LME49600.html#Overview
NOT for battery use. Unless you want a 10lb portable.

 

[G.Trenchev]

Quote:

 
Good choice to use the TO220 version of the BUF634. It is mucho nicer than the DIP8 version

That's because TI send only BUF634T for free

.
The DIP8 seems more convenient to me,but whatever.
 
Well now...should I place a negative feedback for the opamp as well,like in the PIMETA(actually duplicate it),or the global FB on my schematic is enough?
Other suggestions ?
Will feed the amp via LM317/337.
 
 

 

[G.Trenchev]

OK,the mess is here,and it lives!
 

First thing,you guys were right.I've got some hiss with the LM4562.Temporarily solved with OPA2134.Should put some caps there,as jcx said.
It sounds good,but I'm not really impressed.That's probably because of the opamp(LM4562>OPA2134).
Compared to the LM4562 driving headphones alone,it is less open and clear,but the difference is minor.It is much more powerful with 10 times more current.
Here's some measurements:

How to lower the crosstalk ?

 

[G.Trenchev]

Nobody?I'll keep the monologue then

.
BUF634 combined with OPA2132
seems to be a nice headamp.
Adding some filtering capacitors reduced the crosstalk to tolerable -87 dB,so that's fine.The THD is now at 0.0011% with load and 0.0009% without load(Audiophile 2496 baseline).
Here's what it looks like right now:

I still can't tun it stable with LM4562 though.I just don't what to do,so...
Things tried:
-Bypass the 1k resitor with 22pF:
Remove hiss,low-level oscillations,bad performance
-Add 100k resistor from opamp's output to negative input:
Still bad performance,still low-level oscillations on one channel->

 
I can live with OPA2132,but I miss the control and attack of the LM4562.I think it will perform even better if I can run it stable.

 

[nikongod]

 
Quote:

I still can't tun it stable with LM4562 though.I just don't what to do,so...
 
 
I can live with OPA2132,but I miss the control and attack of the LM4562.I think it will perform even better if I can run it stable.
 

 
Quote:

the LM4562 is very fast at  55 MHz GBW - you may need a local feedback cap and considerable layout care to keep it stable, R1 is about 10x too big too for good stability

Id try changing R1.

 

[G.Trenchev]

Shortened it,no effect,increased to 1.5k,nope.
Something else ?
Increase gain ?

 

[cobaltmute]

Have you bypassed all the supply pins?
 
0.1uF to ground on all pos/neg supplies.

 

[G.Trenchev]

Yes,and close to the opamp and buffer pins.