[amb]

I think peranders is proposing to use a MOSFET for the Vbe multiplier and put that on the heat sink with the output MOSFETs. This provides a thermal feedback loop.

This sort of technique is common practice with class-AB power amplifiers to prevent thermal runaway, due to the positive thermal coefficient of BJT devices (quiescent current creeps up with increasing temperature).

Doing thermal sensing limits heat sink choices, because all three devices (both output transistors and the Vbe multiplier) must be mounted to a common heat sink to be meaningful. It also complicates PCB layout for the same reason, and we want to avoid the need to run wires from the board to the sink-mounted Vbe multiplier transistor.

In practice this is not necessary because bias drift is not an issue with this amp at all. The negative thermal coefficient of the MOSFETs will self-regulate the bias very nicely. This is one of the major advantages of MOSFETs. There is a period of time after cold turn-on where the quiescent current goes higher than the desired target idle value, but that actually makes for quicker warm-up times. Once warmed up, the quiescent current is rock steady. This, and the fact that MOSFETs have no secondary breakdown and SOA (safe operating area) concerns is why pro-audio amps nowadays are almost exclusively MOSFET designs.

-Ti

 

[amb]

On the subject of how much power dissipation, here is the basic rundown. With our push-pull class A topology, we want to run at least as much quiescent current through the output MOSFETs as one half the maximum output current. In fact, we want a bit more than that, to allow for headphone transducer reactance, etc.

So, for a 32 ohm load, if we want to achieve, say, 6Vp-p (3V average) without dropping out of class A, then that's 3V / 32 ohms = 94mA. One half of that is ~50mA. Give it a healthy bit of extra margin and let's say we run at 80mA, with +/-12V rails, then each output MOSFET will dissipate 12V * 0.08A = ~1W. This is for the left and right channels. Since the ground channel has to potentially sink or source twice the current return (such as in the case of a mono signal), then the ground channel should be biased to twice that amount. If you want to get even higher output swings into 32 ohm load or if you want to run even lower load impedances with the same voltage swing, then the quiescent current must be increased or you drop out of class A at high volumes.

For single-ended class A topology, the quiescent current (and therefore the power dissipation) must be doubled from the above calculation. This would be the case with the circuit using a current source at the output.

For high-Z phones, the situation is much easier since we're talking about mostly voltage swing and not much current output. So whatever bias we set to optimize for low impedance will work well with higher impedances.

-Ti

 

[aos]

Yes, I'm quite aware of the BJT thermal runoff and the advantage of mosfet's in that regard. My post was to note that while peranders said he wanted to use MOSFET in Vbe multiplier like in his monster amp, the schematics he provided doesn't actually use one. And the one he mentioned is not in a case that allows installation on a heatsink (unless there's a different version of BS170). Unless I managed to misinterpret his schematics.

 

[ppl]

if driving Headphones that require large voltage swings you might consider Bootstraping the op amp.
http://www.reed-electronics.com/ednm...ages/45890.pdf

Also discribed in the below artical is the Classic David White Mosfet power amp in addition to the Authors own circuitry.
http://wiredworld.tripod.com/tronics/mosfet_amp.html

 

[morsel]

PPL: Thanks for the links. We did discuss opamp bootstrapping recently. It adds a lot of complexity. If it came to it I'd settle for using a 36V opamp instead.

 

[xtreme4099]

itd be nice to have an all blackgated version of this amp, with silver wire throughout, and a labpower to see how it compares to my amp, i love the ubercaps, aos can attest to the fact the high quality caps can make a difference, he's heard my amp, but ive changed it since then, morsel, are the lead holes, as big as the ppa or bigger, it would be nice to add super giant caps to this.

 

[aos]

Caps make a difference but your lab power supply (by ppl) likely made the lion share of the difference. Only now a lot of DIYers are realizing the importance of a good PS.

 

[xtreme4099]

even so, good clean power is hard to get right, ive bought my quail cables from that group buy thread and they sure made a difference, more natural, focused highs and mids, very clean sounding, i wasnt originally a power cable guy but i kept an open mind, and sure enough it made a difference, clean power is underrated by diyer's it seems. Silver wire all the way. Wooh00

 

[morsel]

As you know, I think Black Gates and silver wire are a waste of money. I will not take heroic measures to make giant caps fit, but I believe there will be room for bigger caps on the M³ than are on the PPA.

 

[xtreme4099]

Quote:

Originally Posted by morsel As you know, I think Black Gates and silver wire are a waste of money. I will not take heroic measures to make giant caps fit, but I believe there will be room for bigger caps on the M³ than are on the PPA.

i do think there is a difference in tonal qualities between caps, ones that accent more detail versus smoothness, caps of bigger values do affect the overall sound, e.g. the big caps on my diamond buffer. Silver wire is known to be a better conductor than copper, i dont think im alone in saying that adding silver to the ppa has improved the sound for the better, clearer highs, deeper bass, biggersoundstage, some other ppa diy builders can chime in and attest to this as well, ratshack parts will only take you so far, come on morsel, experiment a little, if not for the fun of it, what else is there?

 

[amb]

No capacitors in the signal path is even better than an expensive capacitor... In the M³ the input, output and global feedback are all DC coupled. The only capacitors in the design are for HF compensation, the Vbe multiplier bypass cap, and power supply reservoir and decoupling caps.

The Vbe multiplier bypass cap is the only capacitor in our circuit that might warrant the use of something fancy. We specify a WIMA 1uF film capacitor for this purpose, which should be more than good enough.

Costly "audio grade" capacitors are good if a coupling capacitor must be used in the signal path, and it has been clearly demonstrated that the use of inappropriate types add distortion and smear up the sound. In the M³, we should not have that problem at all. Power supply capacitors are a whole different matter. I am not convinced that the spendy Black Gates are going to make any difference as a reservoir cap over the "standard issue" Nichicons, Elnas or Panasonics. For rail decoupling it is important to use a cap with good HF performance, such as a good film cap or NP0 multilayer ceramic. That's all that matters.

Of course, if you want to spend a lot of money and use Black Gates and such in the power supply, be my guest... it gives you bragging rights when you open the cover

.

-Ti

 

[digi01]

amb,
have you considered add zobels on the output?just like common 2 ch solid based amp doing.

 

[amb]

Quote:

Originally Posted by digi01 have you considered add zobels on the output?just like common 2 ch solid based amp doing.

I haven't seen too many headphone amps with a Zobel. In a speaker amp that is more common due to the types of load. Do you have any reason to believe that a Zobel is desirable in a headphone amp?

-Ti

 

[digi01]

thank you amb.

 

[peranders]

Quote:

Originally Posted by morsel Peranders, perhaps you could explain exactly what it is you are proposing and post a user friendly gif to go with it.

You could do something like this:
http://home5.swipnet.se/~w-50674/hif...0_bias_bjt.jpg
http://home5.swipnet.se/~w-50674/hif...ias_mosfet.jpg

Check the "double" transistor at the heatsink.
http://home5.swipnet.se/~w-50674/hif...2_overview.jpg