30W amp (triple emitter follower) with TIP3055/2955 outputs

Theoretically YES, you can do all of this, BUT:

a) Sim would have to be re-done, and I'm sure changes (e.g. in compensation) will be needed.
Triple output is likely to be sensitive to this kind of changes, and VAS stage is sensitive to op-amp changes.

b) From my experience with BD139/140 - when you but them you never know what it really is. There is plenty of no-name (unknown brand) BDs,
so the model you have in LT Spice may not match the actual devices you are using. So real build will behave differently than the sim.
It's much easier to get TT devices from reputable seller and with correct Spice models, rather than wasting hours fighting with oscillations later on.
If you buy known brand BDs from reputable seller, with correct models, most likely they will cost the same as TTs.

c) Burson v6 opamp, given its high price, seems like a overkill for this application. In my personal opinion you will not hear, nor measure,
any difference from 50 cent TL071

Conclusion - try the sim with different devices, and if everything looks good (Thd, fft, stability) give it a try.
Gottcha

BD139 and 149 are branded CDIL part and CDIL is a good Indian company that exports a lot.

I have started loading the basket for TTC and TTB in farnell.

Understood what you were trying to say with the sim model issue with non standard part.

I am wondering what to do with burson v6 , initial plan was to use it as a buffer for PGA2311.
 
has baker diode, but clipping looks poor.
Current limiting on the VAS would probably fix that. You end up with current spikes in the VAS darlington driver during clipping if you don’t. Needed if you add output SOA protection regardless so it’s just good practice.
On could also insert a resistor (low kohm value) in the collector of the VAS's EF (Q1). It comes with some Miller effect but only at clipping, when collector swing of Q1 increases.
 
TTA's and TTC's have only half the BD's CB capacitance. This matters for VAS and its ccs. But they have slightly better SOA which matters for the drivers. MJE243/253 would be better in this position.
The lower CB and input capacitance doesn’t matter as much in an EF3 as it does an EF2. You could put TIP41’s there. While it would not have this distortion performance it would work surprisingly well. Unlike the usual EF2 where you want to run screaming from the room when you hear it. Two little 6 cent TO-92’s make all the difference inthe world. MJE243 are NOT sustained-beta devices. Most BD139 are. This matters when you are chasing ppm-level distortion. The reason people don’t use them more is they’re only rated to 80 volts. Good ones are sustained beta AND have low quasi-saturation - low vceo is the price you pay to get it. That and nobody puts an fT rating but the regular sources make them at least 50 MHz, usually around 100. It’s simply decent by design. The MJE243 (and 171, 181) were designed for exceptionally low leakage. Presumably so they could be run right up to Vceo rating without issue - back in those days a 160V audio driver was hard to do (Motorola hadn’t figured out what Toshiba, Sanyo and others did yet). But the price paid is the usual mid-range beta hump reminiscent of the 2N3439.

Nothing wrong with TTA/C either in this design - it’s not like power dissipation is exceptionally high. Even with 2 ohms load. Vcc is low, which puts you in a favorable part of the SOA curve. Running three parallel outouts isn’t because 300+ watts of SOA is required - in this case it’s to get flat beta out to 15+ amps. That even out-strips a single MJL3281.
 
  • Like
Reactions: aditya
I tried that initially, but for some reason decided not to use it. Don't remember now why....
If you are chasing every degree of phase margin you don’t want ANY miller effect that you don’t need. Cob modulation also results. Trying to get distortion really low like this you have to beat down every contributor you can. Then it all comes down to layout…. and a new can of worms.
 
Steve, I see you moved the probe 'before' R13 connection. Is this way of measurement more correct? This way feedback loop is 'less' open than before,..
You noticed! 😁 I think so. After all, if it wasn't 2-pole, the miller cap "feedback" would not be included in the probed signal. Like normal compensation, it's another feedback loop. The voltage to R13 is normally the output voltage, not the stimulus voltage (created by the probe).
 
Yeah, but 1-pole miller cap is local feedback, this one here is global.
I'm not sure which way is more correct, but it makes a difference in PM/GM measurements..
With probe moved, good PM/GM is easier to achieve, which looks suspicious to me... I wish that was true 🙂
 
Last edited:
Hello Minek, why is this amp so conservatively rated (30 W) with no less than three output devices per half?

Is the fairly low supply of plus and minus 30 Volts the limiting factor? I would assume with a somewhat higher supply voltage ( +-35V) the output stage would be capable of delivering the 5 Amps required for 100W rms in 4 Ohms without any issues?
 
2N3055's are rated 60V, ie +/-30V, and there is the SOA problem. New BJT have better SOA and can handle 80V+, but not a great idea for ~40 year old parts. One pair could maybe survive 8 Ohms, but you need at least 2 pair for 4 Ohms (~100W) and 3 pair is insurance against recuring rebuilds. Today, if your amps are perishable, that's your choice, but professionals cannot afford a lot of returns. These transistors are cheap insurance. We do not push other parts to their limits so why power transistors?