Push-Pull Follower

[TomWaterman]

Hey folks,

After a bit of reading and searching the web I found this schemo for Pavel Macura's discrete class-a push-pull follower based on the Hiraga stuff:

I am looking at the topology and its gonna take a while to sink in...however I'm looking for some application specific pointers.

I would like to use this style of follower as an output stage driving a 600ohm trafo - it appears in essence to be great for this (albiet a touch OTT!! as it stands) with low THD / IMD and ample current drive, plus I like the fact it uses garden variety BJTs.

I understand its unity gain but was wondering what would be the easiest way to add gain to it? Ideally I'd like a stage that does either unity or +10dB gain so it can be fed from a fader and be a booster output amp.

However it would need severe down scaling to improve efficiency.

As it stands the stage eats current to drive 8ohm speakers @ >1A. I'm still trying to figure out this discrete stuff and the way this stage is biased, however I'd like to keep max current draw to 50mA and still chuck decent level into 600ohm.

If I understand this right (forgive my innocence):

16V RMS would be +24dB, into 600ohm
16/600 = 27mA
Power = approx 0.5W

So 40-50mA would be plenty to maintain class-a into 600ohm with power @ 1-1.5W

What do you reckon - could this stage work well with a re-bias? Maybe Q5 & Q6 can be BD139 as well?

Cheers Tom

 

[Gus]

http://www.elecdesign.com/Globals/PlanetEE/Content/2800.html

http://www.national.com/an/AN/AN-227.pdf

http://www.groupdiy.com/index.php?topic=2491

 

[Samuel Groner]

If you really need 16 Vrms you have to use +/- 24 V supplies (or drive both ends of the primary windings). For a good class A drive into 600 ohm (do you really drive such low loads?) you'll need about 20 mA quiescent current for a push-pull circuit. If you want gain, you'll likely end up with using an opamp.

You might have a look at figure 13 (page 18) here as well: SGA-SOA-1_documentation.pdf
This design is easy to adapt to your needs.

Samuel

 

[TomWaterman]

Thanks for the links Gus, I came across everything but the National pdf before.

Samuel, I looked at your documentation the other day (nice work BTW)....I see the headphone amp is your SGA with a PP current booster in FB loop...looks good.

However I'm specifically interested in the diamond buffer idea to be honest. I guess I don't really need 16V RMS into 600ohm. Just overspeccing the output. In reality the 600ohm output TX will rarely be connected to 600ohm loads, more like >10k ADCs etc. If I understand correctly the reflected load with 1:1 600ohm will then be whatever the load Z is, so >10k??

Oh wait but the TX might need to be loaded with 600ohm (planning CMOB-2L) so yeah it would be 600//10k wouldn't it?

I guess I would really like is as much drive as possible but on +/-22V rails with no more Iq than 50mA, less is better like 20mA.

I suppose what I was asking was - do you think this circuit still has merits of performance when operating at low Iq driving a moderate load like 600ohm tx as opposed to headphones or an 8ohm speaker? Or is it too complex for the job?

Thanks again guys
Tom

 

[Samuel Groner]

I suppose what I was asking was--do you think this circuit still has merits of performance when operating at low Iq driving a moderate load like 600 ohm tx as opposed to headphones or an 8 ohm speaker? Or is it too complex for the job?

If you use it open-loop it is surely adequate, otherwise (i.e. with an opamp) it is--for my taste--overkill (and you might even run into stability problems).

Oh wait but the TX might need to be loaded with 600 ohm (planning CMOB-2L) so yeah it would be 600//10k wouldn't it?

Perhaps you could try a zobel network instead of the raw resistive load.

Samuel

 

[TomWaterman]

Hi Samuel, what would be the diff between a loading resistor and a zobel? Cinemag recommend a loading resistor only in the CMOB-2 pdf...

[edit] damn I think I answered my own ? - the zobel would have a much higher impedance and not load the output as much.....still I wanted to have a jumpered 600ohm load so that maybe the 'tone' could be affected [/edit]

If I ran it open loop, could I tack an input stage and a VAS on the input to get some easy gain or do you think that would be increasingly complex? I only want 10dB

Cheers Tom

 

[TomWaterman]

Actually thinking about the application I could get away with no gain at all so I'll forget about that....

Tom

 

[Samuel Groner]

So let me add a few notes on what you'd likely want to change for your power requirements:
* Q1-Q4: BC550C/BC650C (or similar)
* Q5/Q6: BD137-16/BD138-16 (or similar)
* R3/R4: probably 0 ohm for lower bias current
* R12/R13: probably around 4.7 ohm for lower bias current
* C3/C4: probably lower, say 100 pF (might lead to lower distortion, but check for stability)
* R1/R2/R9/R14-R16/C1/C5: probably unnecessary

Samuel

 

[TomWaterman]

Thanks again Samuel, thats very helpful.

My only concern now is output short cicuit protection, as my app would interface with the outside world...

This Jung circuit shows a diode clamp (D5 & D6), I've seen this mentioned by PRR as a short circuit safety in some of Tamas' amps...is this sufficient protection?

Many thanks
Tom

 

[Samuel Groner]

I think D3/D4 are the protection diodes (protection is somewhat different from standard sinlge stage p-p followers). The function of D5/D6 is not clear to me. You'll need to simulate or breadboard the short-circuit protection for an exact figure of maximum current flow.

It helps that you have a transformer after the buffer which likely offers enough resistance to protect the silicone in most practical circumstances.

Samuel

 

[sismofyt]

i'm the resident hiraga geek / collector. that circuit does not sound very nice. you really have to use those jap transistors which are now made of unobtanium

if you need gain, take a look at the original Lé Classé A 20W or 30W or perhaps the monster. i used that as a line amp some years ago

all my efforts (and everybody elses, afaik) with standard BTJs sounded quite poor and had a tendency to blow up

and it's very psu dependable

good luck

 

[TomWaterman]

Hey sismofyt, sounds a bit dissapointing - any ideas on why you think its so transistor dependant?

I'm going to play with the simulator today and try a few things out, i.e get my head around the damn software! Do you have any experience with other variations on the diamond buffer that you think are successful?

I will listen to it before I make up my own mind, but I did read of quite a few positive experiences with this circuit and also diamond buffers in general over on diyaudio.com...FWIW

Cheers Tom

 

[PRR]

> The function of D5/D6 is not clear to me.

Assume that Vbe of Q1 cancels Vbe of Q3. D5 clamps Q1 base to 0.6V above output. Q3 emitter is likewise clamped to 0.6V above output. No more than 0.6V can appear across R6. Max current in R6 is 0.6V/5ohms = 0.12 Amps.

> is this sufficient protection?

Murphy ensures worst-case loading. Worst case is probably a dead short output with large input. For DC drive, each transistor can dissipate (15V-0.6V)*0.12A= 1.8 Watts. For AC drive, half of that, 0.9 Watts. I've ignored the 10 ohm output resistor; it makes little difference.

The LEDs do the same thing, more directly, but allow about (1.6V-0.6V)/5ohms= 0.2 Amps, and higher dissipation.

You can't readily change the emitter resistor to change the current limit, because that also upsets the bias throughout.

Adding resistors in the collectors can also limit current, and drastically reduce transistor dissipation.

You can add gain with a chip opamp. If that is not acceptable, it becomes a Very Different Problem, even if you "only want a little gain".

 

[TomWaterman]

Sorry PRR I missed your post....

thanks for the info, I've decided I dont require any gain, so I should be ok.
I'm still trying to get my head around LTSpice, I can't figure out how to do FFT / distortion analysis....

also trying SiMetrix which I like a lot better, but the results I get are weird.

-Tom

 

[sismofyt]

i'll be back with some more notes soon, but you might wanna take a look at this:

http://www.bonavolta.ch/hobby/en/audio/monster.htm

 

[TomWaterman]

Hmmm thanks sismofyt.

I did a bit of a search on Le Monstre and it seems there are many opinions abound about the finnicky nature of this circuit. Seems to have a reputation similar to your description of the Hiraga stage.

Looking at it - its current feedback no?

Cheers Tom