output buffer topology

[syn]

Hi
   anyone have any good 3-4 transistor, no NFL or with NFL,  non inverting, output buffer topology, capable of driving at least 600ohms into class A? Level to handle up to 24~28dBu i'll be running it @ +/-28V probably more. i want to buffer output transformer from  an op amp that is providing gain. I don't want to use the "whole" op amp for this purpose as i think it can be done with less parts and simpler topology.. i don't mind heat.i don't mind many mAs that will eat, as long as  there is a good reason for it.
Thank you

 

[Gert]

Try the JLMhybrid.It's very goodsounding and class A biased

 

[syn]

thnx i'm aware of the "hybrid", however it is not what i'm lookin' for it's PP stage runs "inside" feedback loop.
sure it's good just, different purpose.

 

[tv]

I think that the only (reasonable) topo falling in your spec would be a sort of a diamond buffer...

4 tranistors would be the very very basic config, without any ccs's - but perhaps you could use bootstrapping instead.

Just a thought - perhaps you could use some darlington output pair, and a pair of diodes for biasing requirement? This would keep the circuit "mean and lean"..

 

[rotation]

thnx i'm aware of the "hybrid", however it is not what i'm lookin' for it's PP stage runs "inside" feedback loop.
sure it's good just, different purpose.

When saying feedback, are you talking about IC1b which can be used as a DC servo?

Here are some other examples, check "Class A MOSFET Follower":
http://headwize.com/projects/showfile.php?file=opamp_prj.htm

Miha

 

[syn]

@tv thank you i'll try to breadboard something like you proposed.

@rotation thank you very good reading i'll have to go through it a few times.about hybrid: it is good. i don't want to use ic's . joe is using it in NFL. i'm not talking about DC servo (which i don't want either), i'm talking that as i see it, hibryd is an opamp that needs surrounding circuit and to perform its best it needs NFL in that
surrounding circuit (see dingo or baby animal). i'm trying to stay "discrete" on this one (please, don't read this as i think ic's are no good).now ,if i ran it at unity gain how can i do it without NFL and still have it at its best?
i will edit my first post as it is not very clear. and the 3-4 transistor requirement does not include 2604

cheers

 

[syn]

rotation Class A MOSFET Follower looks very good, thank you. i don't know much about mosfets, do you have any favorites?
thank you

 

[rotation]

Hey

i didn't see reply, sorry for being late.. I don't have much experiences with mosfets, so i can't suggest them. I think i only tried 2N5457 until now (except for psu). This one is used for Hamptone HJFP, i tried ON-SEMI brand and another unknown brand. ON had much tighter spec, but i can't comment how their sound compare, i used ON because i didn't have to mess with bias so much. 
You might also look at Hamptone HJFP schematic, those two darlingtons are some kind of buffer. Mosfet acts as a voltage source. I'm not sure if impedance of output buffer is good for what you will have in front of it.
There is also Fetbloak, you built them, no? Check them for ideas.
I think i've seen what you are looking for, it was some time ago and can't remember where. Check HeadWize one more time, there are some good articles. Mosfet follower with op amp from "my" link was designed by PRR, ask him what to do. I remember there was some talking about it some time ago, try search.

Since you are my neighbor i can send you some MPSA14 and ZTX653 (also 2N5457) if you want to try Hamptone HJFP output stage. Check schematic and see if this is what you want. It runs on 24V and is class a. I'm not sure if it will be capable of 24-28dBu.?

Btw, are you modding a mixer? Something else?

Miha

 

[syn]

Miha

            you are very kind, thank you. Those transistors are j-fets (2n5457) mosfets are a bit different...
I've built all that you mention above and  i  stock all the transistors that you mention, no need to send anything,  but thank you on your kindness.
Yes it is a bit confusing. You have a buffer inside the op amp, which is the last stage, the output transistors, and those come in different configurations. Usually they provide no voltage gain, but they provide current gain, and low impedance, to make it easier for the rest of the op amp to  communicate with the world outside.
It is possible to have various sorts of buffers in between the sections of an op amp. like to buffer VAS from the output e.t.c. Now you could take the whole op amp like described above (including it's internal buffer) and make a buffer from it (Hampton, bloak, 2604, hybrid whatever). But then, for a buffer, i don't need V gain, so i don't need parts providing V gain. All i need is dumb and strong, without opinion (THD etc).I want this buffer to stop non linearities from the output transformer reflecting back and influencing NFL of the gain stage. Just to make it clear I'm not "rediscovering hot water" there are many designs out there that do just this, our beloved Neve in it's SE circuits is just one to mention.
I'm doing this because i like to experiment more than to just clone.
But I'm stuck at the moment, as no buffer topo that i tried up to now is “clean” enough, or it is clean and use many parts, or  they  insist on third and odd order harmonics which i don't want.Or it's sinlge rail.Yes i want too much...
I was hoping that someone will show me something fresh, and mosfets are indeed fresh for me as i never played with them. Off to on semi (my favorite transistors generally speaking) site to see what they have to offer .I'll Check HeadWize, i skipped it somehow. 24-28dBu spec is from "just in case" story...
Best
Milos



           

 

[tv]

One thing to consider:

With "single-ended" discrete buffers (most of the "follower" types) you will introduce a DC offset (usually one "diode drop" with BJTs or a Vgs drop with ***fets). You will have to correct for that within your circuit, which means either adding capacitors into the signal path or otherwise compensating for the introduced DC offset.

Diamond buffer OTOH, due it's internal characteristics, doesn't introduce significant DC offset (at least with some component matching). Meaning that a diamond buffer could be connected in a "capacitor-less" fashion between your opamp output and the load, without a feedback.


Something to think about.

 

[syn]

tv thanks, all true. but i was "ssscared" of the parts count. i  found recently a simplified topo of it,  that i'm about to breadboard. did a simo of it, nothing spectacular so i didn't jump on it straight away. i know simulations usually don't say much...

 

[tv]

If so, try to scale-down this one:
http://sound.westhost.com/project83.htm

some audiophiles swear by it...

...quite possibly you could adapt it to +/- pow.supply AND use a simple trimmer pot to adjust for DC offset == just one (input) cap in the audio path.


If you are scared of diamond buffer parts count, you could try to ONLY use a "half" of it (f.e. one PNP into one NPN) ... BUT use a CCS as active load. This COULD preserve the diamond's buffer benefit of canceling the DC offsets, if you are picky with parts selection and tweaks...

 

[syn]

thanks a lot, i'll play with both combinations .i'm not sure how i missed that esp article i was there many times.

 

[rotation]

Miloš,

SCA was offering a combination of AD797 opamp and BUF634 current driver. It was considered as a replacement for Jensen/Hardy 990. I know that BUF634 is not discrete, but it's capable of driving transformers at low impedance. I think i've seen it used as a headphones driver too. Check DiyAudio for more information, they might have some examples of discrete replacements for BUF634.
Project 83 from sound.westhost is overkill imo. But you might profit from mosfet sound. I don't know how they compare to j-fet, i really like j-fets pushed a little, they sound very nice this way. When i started to build Hamptone preamp i didn't think it will sound so different from transistorized circuits. So i was quite surprised when i heard them as a mic pre for drums, instrument pre for bass, guitars etc..

Miha

 

[syn]

Miha
          Project 83 from sound.westhost is overkill as it is, but scaled down it might be great.Thanks for BUF 634
another interesting find at first glance it looks like it draws it's root from the diamond config. Yes j-fets do sound
great i love them too.

Cheers

m

 

[tv]

All in all, no matter what, in the end, a "trimmed" diamond will introduce the smallest part count - and if you match devices well enough, there will be no caps, making the final footprint smaller than other solutions.

I'm afraid there will be no "magic dust" sound-wise....

I'm attaching a schematic of a headphone driver (from "Sijosae"). The devices are clearly insufficient for your purpose and voltage, so you would need to scale-up and probably use the BD139/140 or similar...

Pics:
http://www.headphoneamp.co.kr/ftp/sijosae/Gallery/Accesary/Buffer634-1b.jpg
http://www.headphoneamp.co.kr/ftp/sijosae/Gallery/Accesary/Buffer634-1a.jpg

Website:
http://www.headphoneamp.co.kr/ftp/sijosae/Gallery/

 

[syn]

tv
   that looks great. i'll see for some "magic" to happen before the buffer, the less it is noticeable the better.
thank you

 

[PRR]

Most good audio transformers hate DC.

None of the complementary or trimmed plans will hold DC low enough.

You must have either coupling-cap, or some form of servo (DC feedback or add-on).

Since you reject simple servoing (with a chip), it seems you are forced to a coupling cap.

A dumb emitter follower with resistor load is a fine buffer.

Except: it is very inefficient.

You say you don't mind heat.

+28dBm is most of one Watt. Resistor-coupled power amp can not exceed 6% efficiency. Therefore you are looking at 10-17 Watts of heat. Ouch!

CCS-loaded EF can do 25%. Only 4 watts heat.

If perfectly optimized!

+28dBm in 600 ohms is 27.53 Volts peak. Your +/-28V rails are about right. Less won't reach +28dBm, and more is just useless heat. You will have to do a little more to cover transistor and CCS losses.

+28dBm in 600 ohms is 45.9mA peak. Set your CCS for 46mA.

Rounding up a little: +/-30V at 50mA is 6 Watts (3 in transistor, 3 in CSS).

If your "opamp" output is at zero V DC, then the (NPN) emitter follower output is at -0.6V. This gives a small but very predictable bias for the output coupling capacitor (which "has" to be electrolytic because -3dB at 10Hz requires 27uFd, awful large for a film-cap, but a 100uFd electro is cheap).

The input impedance of any good small power transistor, emitter loaded in 600R, is well over 6K, so any opamp can drive it easily.

The roughly 1mA base current will keep the opamp's output stage in class A.

Assuming the opamp has low output impedance, the emitter impedance will be about 0.6 ohms, which is insignificant compared to 30-60 ohms copper resistance in the transformer. (In fact you should probably build-out with 10 or 30 ohms to isolate the emitter from winding capacitance and supersonic instability.)

 

[syn]

PRR
      thank you very much.