Discussing audio amplifier design -- BJT, discrete
in [Basics]
|
Prev: Watch replacemnt (button) batteries?
Next: Low Power Satellite Based Laser / Imaging System Could Easily Track Activity Disturbing Cheap Reflecting Fibers
From: Paul E. Schoen on 23 Feb 2010 01:01 "Jon Kirwan" <jonk(a)infinitefactors.org> wrote in message news:9lj5o5dicqhmln4on3uq9rlk6ejii12hq8(a)4ax.com... > On Mon, 22 Feb 2010 10:03:41 -0800, I wrote: > >>On Mon, 22 Feb 2010 08:41:39 -0800, I wrote: >> >>><snip> >>>Again, looking at Self's chart (page 322 on his 5th edition) >>>I see a slight degregation into 4 ohms, going from about 20W >>>into 8 ohms to 15W into 4 ohms. I'm not entirely sure of >>>'theory' here, but I took this to suggest that at the higher >>>currents the drive circuitry's compliance coupled with the >>>likely somewhat lower gain caused by somewhat higher currents >>>now needed accounted for the droop. >>> >>>But his chart certainly doesn't suggest 1/2 rated power. >>><snip> >> >>Another thought crossed my mind, too. If the amplifier he >>was testing used a Vbe multiplier to achieve class-A >>operation, that won't be enough when faced with 4 ohms. If >>so, it will degrade into class-AB operation. Not sure that >>that means, yet. >> >>Need to think more on that, as well. > > Sorry to keep responding to myself, but even more crosses my > mind, including VAS loading. > > So I stopped letting things cross my mind and set up a spice > simulation to see what it tells me. (I hate doing this, > without applying theory, but I feel time is of the essence > and like cheating.. for now.) > > Class-A appears to deliver the same thing as class-B, at > least using a TIP3055 and TIP2955 output pair, and using an > idealized voltage source between the bases to set the class > of operation. With a 4-ohm load and the exact same drive > voltage (using again a voltage source as the VAS output), I > got 20.55 watts into 4 ohms with class-A operation and 10.31 > watts into 8 ohms. (Which is not a 4X but 2X phenomenon.) In > class-B, this was 19.66 watts into 4 ohms and 10.16 watts > into 8. Again, 2X. (I think I might have been just slightly > into class-AB with that last test, but I got it close.) > > So it maybe doesn't matter about class of operation. But is > about the quiescent current flowing via the vbe multipler and > what is available to _drive_ the output BJTs and perhaps also > some estimation about output drive impedance of the VAS which > hauls the output section around in real amplifiers that > caused the table entry values I saw with Self's book. The most important distinction, I think, is the difference in efficiency at power levels lower than maximum. Class A efficiency is about 35% to as low as 15% at full power, and it drops to the point where it is essentially just a heater when output is in the normal listening range of one or two watts for a 10W amp. See http://sound.westhost.com/class-a.htm. A class B amplifier should do about 69% efficiency at full power, especially if you can drive the output stage with a source that is higher than the rails for rail-to-rail output. An alternative is to create higher rails for the driver stage, or to use a bootstrap approach. I have attached a simple amplifier using MOSFETs that provides 14W into 8 ohms at 20 Hz - 20 kHz with a 35 VDC single supply, and it achieves 67%. It has a quiescent power of about 2.2W. It is not a "practical" design, however, as the biasing for the output MOSFETs depends on their Vto which may not be stable. I added a simple resistor to adjust bias, and I used logic level MOSFETs. The same basic circuit could be done with all bipolar as well, but I like MOSFETs. I'd like to see the ASC files for your LTSpice simulations. And it would be interesting to see the results of a frequency sweep using AC analysis. Paul ------------------------------------------------------------------------ Version 4 SHEET 1 1304 744 WIRE 736 -128 192 -128 WIRE 976 -128 736 -128 WIRE 736 -96 736 -128 WIRE 736 0 736 -16 WIRE 736 0 640 0 WIRE 848 0 736 0 WIRE 976 0 976 -128 WIRE 640 80 640 0 WIRE 736 80 736 64 WIRE 848 80 848 0 WIRE 928 80 848 80 WIRE 192 192 192 -128 WIRE 496 208 400 208 WIRE 736 208 560 208 WIRE 976 224 976 96 WIRE 1072 224 976 224 WIRE 1216 224 1136 224 WIRE 1248 224 1216 224 WIRE 976 304 976 224 WIRE 976 304 848 304 WIRE 848 352 848 304 WIRE 640 368 640 160 WIRE 736 368 736 336 WIRE 736 368 640 368 WIRE 800 368 736 368 WIRE 976 368 976 304 WIRE 1248 384 1248 224 WIRE 928 448 848 448 WIRE 736 464 736 368 WIRE 848 496 848 448 WIRE 400 512 400 208 WIRE 192 640 192 272 WIRE 400 640 400 592 WIRE 400 640 192 640 WIRE 656 640 400 640 WIRE 736 640 736 544 WIRE 736 640 656 640 WIRE 848 640 848 576 WIRE 848 640 736 640 WIRE 976 640 976 464 WIRE 976 640 848 640 WIRE 1248 640 1248 464 WIRE 1248 640 976 640 WIRE 656 672 656 640 FLAG 656 672 0 FLAG 400 208 Vin FLAG 1216 224 Vout SYMBOL voltage 192 176 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName V2 SYMATTR Value 35 SYMBOL voltage 400 496 R0 WINDOW 3 -75 232 Left 0 WINDOW 123 24 44 Left 0 WINDOW 39 0 0 Left 0 SYMATTR Value SINE(0 18 200 1u 0 0 5000) SYMATTR Value2 AC 1 SYMATTR InstName V1 SYMBOL nmos 928 0 R0 SYMATTR InstName M1 SYMATTR Value STD30NF06L SYMBOL nmos 928 368 R0 SYMATTR InstName M2 SYMATTR Value STD30NF06L SYMBOL pmos 800 448 M180 SYMATTR InstName M3 SYMATTR Value IRF7205 SYMBOL res 832 480 R0 SYMATTR InstName R2 SYMATTR Value 1k SYMBOL res 1232 368 R0 SYMATTR InstName R8 SYMATTR Value 8 SYMBOL polcap 1072 240 R270 WINDOW 0 32 32 VTop 0 WINDOW 3 0 32 VBottom 0 SYMATTR InstName C3 SYMATTR Value 10000� SYMATTR Description Capacitor SYMATTR Type cap SYMATTR SpiceLine V=16 Irms=1.6 Rser=0.05 MTBF=20000 Lser=0 ppPkg=1 SYMBOL res 720 -112 R0 SYMATTR InstName R9 SYMATTR Value 10k SYMBOL res 720 448 R0 SYMATTR InstName R1 SYMATTR Value 10k SYMBOL diode 720 80 R0 SYMATTR InstName D1 SYMATTR Value 1N4148 SYMBOL diode 720 144 R0 SYMATTR InstName D2 SYMATTR Value 1N4148 SYMBOL diode 720 208 R0 SYMATTR InstName D3 SYMATTR Value 1N4148 SYMBOL diode 720 272 R0 SYMATTR InstName D4 SYMATTR Value 1N4148 SYMBOL polcap 560 192 R90 WINDOW 0 0 32 VBottom 0 WINDOW 3 32 32 VTop 0 SYMATTR InstName C1 SYMATTR Value 100� SYMATTR Description Capacitor SYMATTR Type cap SYMATTR SpiceLine V=63 Irms=900m Rser=0.1 MTBF=20000 Lser=0 mfg="Nichicon" pn="UPH1J101MRH" type="Al electrolytic" ppPkg=1 SYMBOL diode 720 0 R0 SYMATTR InstName D5 SYMATTR Value 1N4148 SYMBOL res 624 64 R0 SYMATTR InstName R3 SYMATTR Value 2k TEXT -176 688 Left 0 !.tran .5 TEXT 40 696 Left 0 !;ac oct 5 20 20000
From: Jon Kirwan on 23 Feb 2010 07:20 On Mon, 22 Feb 2010 22:56:22 +1000, David Eather <eather(a)tpg.com.au> wrote: ><snip> >Wait till you start talking about speakers! Okay. Now you made me really worried. I just realized (okay, so I "get it" for a moment once in a while and then manage to forget it until the next time) again that speakers can be nasty in terms of phase shifts and loads. Cripes. Here I was about to embark on output stage design and this rears its head, again. At certain slew rates, a speaker can demand a lot from the amplifier's output and I will need to think about 'protection' -- especially if that takes place at low frequencies like 20Hz where the main stretch of the slope from peak to valley can last as long as maybe 20ms or so. Whatever the output stage is, it needs to handle peak dissipations as well as peak currents, perhaps, for that long. (I'm trying to keep in mind that current phase may either lag or lead.) Wait until I close the global NFB loop, too. And what about oscillation in local FB, too? (If FETs, that might be more [or less] a consideration... I have no idea, right now.) Jon
From: Jon Kirwan on 23 Feb 2010 08:13 On Tue, 23 Feb 2010 01:01:07 -0500, "Paul E. Schoen" <paul(a)peschoen.com> wrote: >"Jon Kirwan" <jonk(a)infinitefactors.org> wrote in message >news:9lj5o5dicqhmln4on3uq9rlk6ejii12hq8(a)4ax.com... >> On Mon, 22 Feb 2010 10:03:41 -0800, I wrote: >> >>>On Mon, 22 Feb 2010 08:41:39 -0800, I wrote: >>> >>>><snip> >>>>Again, looking at Self's chart (page 322 on his 5th edition) >>>>I see a slight degregation into 4 ohms, going from about 20W >>>>into 8 ohms to 15W into 4 ohms. I'm not entirely sure of >>>>'theory' here, but I took this to suggest that at the higher >>>>currents the drive circuitry's compliance coupled with the >>>>likely somewhat lower gain caused by somewhat higher currents >>>>now needed accounted for the droop. >>>> >>>>But his chart certainly doesn't suggest 1/2 rated power. >>>><snip> >>> >>>Another thought crossed my mind, too. If the amplifier he >>>was testing used a Vbe multiplier to achieve class-A >>>operation, that won't be enough when faced with 4 ohms. If >>>so, it will degrade into class-AB operation. Not sure that >>>that means, yet. >>> >>>Need to think more on that, as well. >> >> Sorry to keep responding to myself, but even more crosses my >> mind, including VAS loading. >> >> So I stopped letting things cross my mind and set up a spice >> simulation to see what it tells me. (I hate doing this, >> without applying theory, but I feel time is of the essence >> and like cheating.. for now.) >> >> Class-A appears to deliver the same thing as class-B, at >> least using a TIP3055 and TIP2955 output pair, and using an >> idealized voltage source between the bases to set the class >> of operation. With a 4-ohm load and the exact same drive >> voltage (using again a voltage source as the VAS output), I >> got 20.55 watts into 4 ohms with class-A operation and 10.31 >> watts into 8 ohms. (Which is not a 4X but 2X phenomenon.) In >> class-B, this was 19.66 watts into 4 ohms and 10.16 watts >> into 8. Again, 2X. (I think I might have been just slightly >> into class-AB with that last test, but I got it close.) >> >> So it maybe doesn't matter about class of operation. But is >> about the quiescent current flowing via the vbe multipler and >> what is available to _drive_ the output BJTs and perhaps also >> some estimation about output drive impedance of the VAS which >> hauls the output section around in real amplifiers that >> caused the table entry values I saw with Self's book. > >The most important distinction, I think, is the difference in efficiency at >power levels lower than maximum. Class A efficiency is about 35% to as low >as 15% at full power, and it drops to the point where it is essentially >just a heater when output is in the normal listening range of one or two >watts for a 10W amp. See http://sound.westhost.com/class-a.htm. Thanks. I'm seeing this well, now. >A class B amplifier should do about 69% efficiency at full power, >especially if you can drive the output stage with a source that is higher >than the rails for rail-to-rail output. By this, do you mean several power rails? (I'm trying to imagine a system that delays the output [equal time for all frequencies of interest, or a group delay = 0... haha] and "anticipates" the required voltages and uses an FFT [applied to an inverse FFT 'filter' determined at startup] to develop the appropriate signals that would automatically generate the right voltages at the right times. That should be 'fun', if it would work at all. Would have to be noisy as all get-out, I think. But might be interesting.) >An alternative is to create higher >rails for the driver stage, or to use a bootstrap approach. I know of the use of 'bootstrap' for other purposes, like stiffening the apparent input impedance and I plan to use it there. But what do you mean in this case? >I have attached a simple amplifier using MOSFETs Saved and observed. >that provides 14W into 8 >ohms at 20 Hz - 20 kHz with a 35 VDC single supply, and it achieves 67%. It >has a quiescent power of about 2.2W. It is not a "practical" design, >however, as the biasing for the output MOSFETs depends on their Vto which >may not be stable. I added a simple resistor to adjust bias, and I used >logic level MOSFETs. The same basic circuit could be done with all bipolar >as well, but I like MOSFETs. I'm not yet 'used to' them, except as switches. And as I understand things from reading people like Self, there are tradeoffs -- neither BJT nor FET is a certain win over the other. I feel a little more ready to face the details of a BJT output driver and probably need to get _one_ approach worked out so that I understand it better before taking on the other. (I'm assuming that you don't imagine using FETs all the way through.) In any case, I'll have some context, then, to fathom the pros and cons, later on. Also, high power FETs are "more expensive." At least, so far as I've experienced. And I can consider paralleling BJTs with emitter degeneration. I understand it and why it works. I'm not sure I'd understand, at this point, how to parallel FETs in linear operation. >I'd like to see the ASC files for your LTSpice simulations. And it would be >interesting to see the results of a frequency sweep using AC analysis. Okay: http://www.infinitefactors.org/misc/spice/behavioral%202-BJT%20output%2001.asc Hopefully, you can pick that up okay. Jon
From: Jon Kirwan on 23 Feb 2010 08:28 On Tue, 23 Feb 2010 05:13:15 -0800, I wrote: >http://www.infinitefactors.org/misc/spice/behavioral%202-BJT%20output%2001.asc I modify Vb to set the class of operation, using larger values (say, 2V, for a 4 ohm load for example) to make sure it is in class-A. I use 1.3V or thereabouts for class-B. All these depend upon load. Tweak as needed. I usually just observe the two emitter currents to get a bead on the operation mode and tweak the class-B to be just slightly on the class-AB side. For class-A, I jack it up so that the emitter currents don't show any visible "blunting" on their sinusoidal shape. Jon
From: David Eather on 24 Feb 2010 00:03
Joh, I'm leaving all the sniping to you because your the one who knows what you want to investigate further. Also, for today, I am using mostly using "perfect components" and theoretical efficiencies etc to make my life easier - they will show the point and I am talking about an amp spec'd for consumer audio. On 23/02/2010 2:41 AM, Jon Kirwan wrote: > On Mon, 22 Feb 2010 22:56:22 +1000, David Eather > <eather(a)tpg.com.au> wrote: > >> On 17/02/2010 2:28 PM, Jon Kirwan wrote: >>> On Thu, 11 Feb 2010 17:51:27 +1000, David Eather >>> <eather(a)tpg.com.au> wrote: >>> >>>> <snip> >>> >>> But a requirement to support short-term power levels is >>> really just a compliance requirement on the power supply >>> rails, isn't it? >> >> Yes. > > Okay. These concepts are slowly settling into my brain. > >>> So put another way, if I wanted a long-term average of 10W >>> output and I wanted the extra margins required to support the >>> worst case estimate of a factor of 8 for short-term power >>> bursts, then I'd need to design rails that support a voltage >>> compliance level substantially higher. >> >> The other ways around. The design will deliver ten watts maximum >> (disregarding clipping) but the average power output will actually be >> much lower - hence you can "skimp" a bit on the supply transformer and >> heatsinks - which wrt overheating have very long time constants relative >> the the peak output demands. > > Cripes! Really? So a 10W amplifier isn't designed to > actually deliver a full 10W steadily into a load? That's the > peak power capability? Cripes. No, that's the power of a 12.7 volt (peak) sine wave into an 8 ohm load. An instantaneous peak power figure would be (Vmax**2)/R or 20watts. > > Let me put this another way. I design a class-B output stage > with rails capable of 10W compliance into 8 ohms (roughly 13V > peak, so rails at maybe +/-17V or so?) With 10W into the 8 > ohm load, let's say this means the upper power BJT is > handling about 4-5W and the lower BJT is handling 4-5W, as > well. Call it 10W total dissipation inside the amp while 10W > are dissipated in the speaker. > > But I don't have to go find BJTs able to dissipate 4-5W, > because the 10W spec is just a max-unsustained case and the > real situation is more like 2W into the load, continuous? In > short, I need to find a BJT that only needs to dissipate 1W > for the high side and 1W for the low side? I could use two > PN2222As in parallel to do that! > > I can _cheat_ like that and call it a 10W design? It doesn't > actually _have_ to sustain 10W without burning up? Think of it this way: You build an amplifier that puts out a 10 watt sine wave into 8 ohms 100% of the time. For a power transformer you will need something like a 30 volt CT rated at 40VA for this design (this one is a realistic not theoretical estimate). Use this amp in a consumer environment and the customer is happy about everything except the cost. Can you lower the cost without damaging the output quality? Yes. As mentioned earlier the long term average output power of the amplifier will be about 2 watts and any transformer will have a very long thermal time constant compared to any other component in the amplifier, so there is no danger of overheating during a peak in the music output. 40VA x 0.2 = 8VA. An 8VA transformer is big enough (in real life you would use a little bigger because of the increased I**2 x R losses 10VA would probably be a good choice, (if you had more information you could make a better choice but the result would be very close to 10VA) A 10VA transformer costs a whole lot less than 40VA and all you have done is removed an unnecessary over-specification of a component and that will have zero effect to the consumer. You can do the same thing with the heatsink, but it is not so dramatic a change, and it needs more care. On a small amp like this where the cost of a heatsink is low I wouldn't bother - except as an exercise or if you were making hundreds of them. You won't be able to proceed here until you have a more finalized design. > > Okay, now I'm depressed. I go buy a 50W amplifier, stick a > sine wave signal generator on it and watch the thing toast > itself, bursting in fire soon enough? > >>> The parts would need >>> to withstand it, too. And because of the much higher rail >>> voltages that need to be dropped most of the time, the output >>> BJTs would need to have just that much more capacity to >>> dissipate. >>> >>> Or put still another way, assuming that my output swing at >>> the output stage emitters cannot exceed a magnitude of 15V >>> and that everything is sized for dissipating 10W, does this >>> mean the amplifier is a 10W amplifier that can support a peak >>> of 14W=(15^2/(2*8))? (Which isn't so good, considering your >>> comments above regarding "music?") >>> >>> What is meant when one says, '10 watts?' >> >> A ten watt amp delivers a sine wave producing 10 watts of output power >> into a specified load. Ideally this would be 10 watts for an infinite >> period of time but for audio amps, due to the nature of the signal, an >> "infinite period of time" in practical terms may be as short as a few >> seconds. > > Yeah. A few seconds. So... now I can go back with a > quasi-comp output stage and use a pair of those PN2222As for > it, without heat sinking! Nice little TO92 packages, too. ;) > >>> This gets worse when I consider the class of operation, >>> doesn't it? I mean, class-B might be specified as 10W into 8 >>> ohms, but wouldn't that be 20W into 4 ohms? >> >> 40W output. I**2 x R. The power supply voltage is >> approximately constant. > > I was looking at some actual measurements taken by Mr. Self > on an actual class-B amplifier when I wrote that. I didn't > do a theory-based analysis. Just read off the figures when > he was comparing a class-A with a class-B into different > loads. > > Now I'll do that. recheck everyone's figures > > I had then imagine it came from V^2/R and knowing that V^2 > remains the same for a given amplifier and only the R changed > from 8 to 4. Which makes sense then that it would double, > not quadruple, the power output. From an I^2*R perspective, > I get the same estimate because a smaller load does double > the current, but the R divides in half, so the combination is > still just twice, not four-times. > > Can you explain this 40W statement better for me? 20 Watts - you were right. My figure wrong. > >>> But if class-A, it's pretty much 10W no matter what? >> >> If class A, power is 5 watts out with 4 ohms. Current is held >> constant. > > Again, looking at Self's chart (page 322 on his 5th edition) > I see a slight degregation into 4 ohms, going from about 20W > into 8 ohms to 15W into 4 ohms. I'm not entirely sure of > 'theory' here, but I took this to suggest that at the higher > currents the drive circuitry's compliance coupled with the > likely somewhat lower gain caused by somewhat higher currents > now needed accounted for the droop. > > But his chart certainly doesn't suggest 1/2 rated power. > > I guess I need to delve into this a bit more to make sure I > understand. The class-B case seems easier for me to follow > (assuming I'm right, above, which I of course may not be.) You seem to be doing OK. Maybe we should brush up and compare notes on the meanings of average and RMS but that's about it I think. A class A amp say at 10 watts into 8 ohms will have an output stage with a constant current sink (or source) set at 1.59 amps. If the speaker load changes to 4 ohms the maximum current into and out of the speaker is still 1.59 amps. How's the power now? Self is talking about his practical results and if you dig around you will see/find he believes in over-biasing the output stage current source by 50% - 100% hence the apparent anomaly. > >>> I'm beginning to imagine amplifiers should be specified as to >>> their peak output voltage compliance into 8, 6, and 4 ohms; >>> instantaneous and sustained without damage to the unit. For >>> example, 35V into 8 ohms instantaneous, 15V sustained. Or >>> 80W instantaneous, 15W sustained. That way, someone might >>> have some knowledge about how well it might handle _their_ >>> music at, say, 15W average power. And could compare that >>> against another unit specified as 20V into 8 ohms, 15V >>> sustained. >> >> Your argument here is reasonable but ..... it is also the beginning of >> the PMPO fiasco - since no advertising department could agree on what >> constitutes "music" they used what ever figures looked best - and that >> led to the PFPO (peak fantasy power output) fiasco where you just put >> anything you like on the box. >> >> For a short time some (better) manufactures used a figure they called >> "headroom" which was the maximum possible instantaneous power output >> when the power caps are fully charged divided by the long term power >> output (10 watts in this case). It was always expressed in db - but was >> confusing to the customer - so it disappeared. > > Okay. Well, I can say one thing. I've learned that there > are output specs and there are output specs and what they > actually mean is yet another question, usually unanswered. > > As a consumer, I've become a little better informed even if > all that means is I'm a lot more suspicious than before. > >>> How does one know what they are buying? What a headache. >> >> Wait till you start talking about speakers! > > Hehe. Now I'm really scared. ;) In Sound Lounges no one can hear you scream.... > > Jon |