feedforward vbe compensation in class-A

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
hi,

so, I'm working on a headphone amp, and I'm going to use a class-A emitter follower with no global feedback. philosophical issues aside, there shouldn't be any need for global feedback, since I can just waste a whole bunch of power getting the distortion low in the output stage and I'll have huge supply rails (like, +-32 V) for large emitter degeneration resistors in the input stages.

i'm designing this for the sennheiser 580/600/650 series by the way (300 ohms), thus the large supply rails and the lack of concern about noise from large bias currents.

what I'm concerned about here are nonlinear vbe and non-constant beta in the output transistors, as I see these as the main likely causes of distortion. with modern transistors the beta effect is small, and can be made smaller still by using an output triple or doubling up ouput transistors, but the vbe effect is fundamental.

anyway, what I'm thinking of doing is using a controlled current source in the emitter of the output transistor to make its collector current constant, thus reducing the effects of both vbe vs. Ic and beta vs. Ic. Is there a name for such a system? Has it been tried much? Is there a fatal flaw I'm not seeing? I know it's power-inefficient, but for a headphone amp, who cares...

tim
 
in the typical configuration, the current through the output transistor is the sum from the current source and the current driven through the load, so not constant.

for instance, say I'm driving 300 ohm headphones to +-24 V (about one watt, 126 dB spl with senn. 580) with a 100 mA standing current in the output transistor. the current through the load has peak amplitude of 80 mA, so the current through the output transistor varies between 180 mA at the positive extreme and 20 mA at the negative.

what I'm proposing is to actually measure the current through the load with a small series resistor, and subtract it from the current source setting. strictly speaking that introduces a feedback loop around the output transistor via loading effects on the voltage gain stage and the non-constant vbe in the output transistor, but the loop gain is small and the distortion reduction (if there is any) would come from the feed-forward effect instead.

I suppose you could ask why not just use global feedback if I'm going to do all that work...

furthermore it's very wasteful of power...I haven't crunched the numbers, but I really don't care if my headphone amplifier dissipates 20 watts to drive 1 watt into the load! it will just impress my friends more with how warm it gets and what giant heat sinks it has.

might be interesting just to try it and see if it makes any audible difference at all. most of the distortion from vbe and nonlinear loading effects will be second harmonic anyway, and all evidence shows that's the least offensive type.

oh, and of course there will be a global feedback loop to servo out the dc offset, but it will be way sub-audio.
 
The one and only
Joined 2001
Paid Member
The Sennheisers need so little voltage and current that
you probably can do the job with a dumb Mosfet like the
IRF610 operated as a follower with 13.8V (easy power
supply to buy) and biased by a 100 ohm source resistor.

You'll get a little 2nd harmonic, ohhh, too bad ;)
 
P=V^2/R, so the Sennheisers need lots of voltage but not much current.

anyway, I'm dead set on using BJTs, since (a) Doug Self seems to be convinced they're more linear (b) the square root (as opposed to log) voltage drop vs. output current for mosfets makes (a) logical and (c) I've listened to an utterly amazing BJT headphone amp that I'm basically trying to copy.

But yeah, a second harmonic at -60 dB is probably not going to be the end of the world. So perhaps I just won't bother.

Tim
 
oops! didn't see who I was replying to...I'm sure you know that P=V^2/R :)

oh, boy, and you must have kind of a rivalry with doug self since he's Mr. BJT and you're Mr. Mosfet. anyway, my foot's in my mouth!

by the way, I really dug that DIY op-amp tutorial you used to have on the website...is that still available somewhere? i used to tell everyone who was curious what an op-amp is to go read it...

later,

tim
 
I'm pretty sure you'll never hear from him again.;)

However, the answer to your question is that yes, the article is still available. It's on the Pass DIY website under preamps.

And, by the way, while I too know the relationship between power, resistance, and voltage, why do you say that you'll need lots of voltage ? Surely the headphones consume next to no power, don't they? Or am I missing something?
 
The one and only
Joined 2001
Paid Member
If the headphones are 300 ohm and you use a BJT with
a beta of 100, then the input impedance will look like 30K,
and anything should be able to drive that fine.

p.s. I was Mr BJT until about 13 years ago, and then I
converted for no particular reason. :cool:
 
So, my goal is to drive 1 watt into 300 ohms before clipping. For the Sennheisers, that makes 126 dB SPL, which is the 96 dB dynamic range of a CD plus an optimistic 30 dB of noise from the amplifier and listening room. That's 24.5 V peak, so to be safe I'm going to use +-32 V supply rails. There's also an annoying gap in available power transformers between 24 V and 36 V.

Don't try doing that calculation for your speakers or you'll wind up wanting a 10 kW amplifier!

Anyway, the amp that inspired me uses two cascaded driver stages for the output, presumably to reduce the effects of non-constant beta, and it's that output triple that I'm trying to avoid. From the looks of all the little ferrite beads and funny RC networks he had to throw in, there must have been some irritating stability problems.

All the transistors in that design are obsolete now, though, and now that I do the calculations it shouldn't be a problem to just do a normal two-stage emitter follower with modern transistors, which have more constant beta.

The whole problem with this is that the Sennheisers are just so good, you start to notice very tiny changes in the circuit easily. The guy who built the other amp spent a year and a half tweaking it...
 
:mad: I'm not planning on sitting around listening to sine waves at 126 dB. the point is, you don't need that much power continuously, but you need it to be available at low distortion for the transients. not for all music, but a big orchestra is probably at least that loud at full blast.

anyway, what's the point of having a 16 bit recording if you can only squeeze 14 bits worth out of your amplifier?

also, remember that the ears' sensitivity (and risk of damage) is highly frequency dependent. I remember reading once that 1/3 of the power in an orchestra comes from the bass drum...maybe that's exaggeration, but the basic point is that huge bass transients do exist in real life, and that none of the amplifiers most of us have ever listened to are capable of properly reproducing them. in fact, having an overpowered amplifier makes it safer for your ears, since less of the bass gets up-converted to more damaging frequencies.

plenty of people drive themselves deaf with the $.20 op amp in their walkman...

and to cope with the risk that something fails, I will of course have some protection system. I plan on actually calculating the real power delivered to the load using a current-sense resistor and analog multiplier, and shutting down the amp at a continuous power dissipation well below one watt. simple current limiting is no good, since headphones come in such a wide range of impedances. even if deafness were my goal, I wouldn't want to break my headphones!
 
oh, and i guess i should point out, there are probably precious few recordings for which this amplifier will have any advantage... the first couple least significant bits are probably noise anyway, and with the "level war" going on in popular music, serious dynamic range conpression is the norm.

but my feeling is, I want to design a headphone amp only once in my life, and whatever new recording technology comes along, I want it to be up to the challenge...at least power-wise.
 
Like my colleaguess I suggest to use a simple project, please concern a single bjt with a ccs as re as an emmiter follower (EF) that is steering direct from a cd player. 2Vrms from a cd is enough to drive 300 ohm through a EF?

Because the bjts like to have a large nfb maybe is it better to use a real re, not ccs, which will give 100% nfb? You will loose some damping factor which in case of 300 ohm do not seems to be important?

Probably it would be very hard to build an amp like you want and the result might be not worth of your efforts.

And please do not be afraid of the 2nd harmonic at -60dB: I hope that you will not hear it! Similar like nobody does! The worst are high order odd harmonics (5, 7, 9, 11.....) which even at -120 dB are very dangerous for making the sound unpleasant, sharp, metallic, etc...

In the end I recommend a use of a single mosfet DC coupled loaded by ccs in source follower mode. I doubt if is there something better and simpler?
 
Hmm.......

Its a long time since I've read such tortured logic to complicate
what is essentially a very simple design problem by adding in
engineering terms non-sensible requirements.

Having said that its perfectly feasible to connect 300 ohm head
phones across a power amplifier, and there are class A power
amplifiers dersignsespecially converted for headphone use,
with much reduced standing current in the output stage.

The musings on dynamic range are risible. If a high voltage
swing amplifier is being used and you are not using its swing
much you are simply compromising signal to noise, which will
be improved by using an attentuator at the output, basically
the "old-fashioned" amplifier headphone socket method.

Concern for dynamic range is fine, inconsistent logic isn't.
IMO 106dB maximum with an inaudible noise floor is better
than 126dB maximum with an all too audible noise floor.

Concerns for transients that simply don't exist on most
recordings is also fairly pointless. The very nature of
setting CD "average" levels means they will be limited,
except for the few brave souls allowing high peak to
average levels on CD, but this by definition means
average resolution is a lot less than 16 bits.

:) sreten.
 
Hearing loss and (by reading between the lines) estimates of dynamic range requirements for "natural" sound level reproduction are covered in this article:

http://headwize.com/articles/hearing_art.htm

while I’m not noticing any obvious clipping with op amp drive capable of 8 Vrms into my HD600 I don’t find the assertion of higher peak power requirement totally unreasonable – higher power (= more than standard op amp Vswing) does seem to be a differentiating feature of high end amps designed for 300 Ohm headphones


Your output stage ramblings seem to show a lack of understanding of the Vbe distortion cancellation of a class A push-pull complementary emitter follower stage – a very basic audio output circuit – as a discrete buffer many here and at the headphone forums make a fetish of the “diamond buffer” implementation of a complementary emitter follower

stabilizing bias for a low power class A stage is simpler than for high power class AB output but there are still many choices and subtleties

Gilmore certainly agrees with the high peak power requirement and has several class A output designs, his latest:

http://www5.head-fi.org/forums/showthread.php?t=71705&page=1

For all things headphone related you should cruise the content of both of these sites
 
:rolleyes: hmmm...

Your output stage ramblings seem to show a lack of understanding of the Vbe distortion cancellation of a class A push-pull complementary emitter follower stage – a very basic audio output circuit – as a discrete buffer many here and at the headphone forums make a fetish of the “diamond buffer” implementation of a complementary emitter follower

your offensive tone is sick, but THANK YOU.

just because I haven't ever designed an output stage, I'm not worthy of any respect? I consider output stages boring grunge, and I'd rather never have to build a power amp in my life, but whatever.

anyway, I don't have a book lying around that covers these things, so perhaps you could recommend one? I'm really sick of trying to find anything useful about power amps on google.

and what, if people had just read my original post and replied to that rather than dragging me around a tree over the voltage headroom I wanted, maybe I wouldn't end up having to ramble off?
 
Because the bjts like to have a large nfb maybe is it better to use a real re, not ccs, which will give 100% nfb? You will loose some damping factor which in case of 300 ohm do not seems to be important?

for sure, you're absolutely right. but I want to try the current source too just to compare, and as long as the current source is there along with the current-sense resistor for the protection system, why not try subtracting it? it will be fun just to see.

maybe I'm going overboard on the protection system...I just don't have good specs on the thermal handling characteristics of the headphones, so I'll want to play it safe. the protection system is going to double the parts count anyway...
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.