Forums

Class B distortion

Started by Unknown December 20, 2014
meow2222@care2.com Wrote in message:
> On Sunday, December 21, 2014 6:02:25 AM UTC, bitrex wrote: >> miso <miso@sushi.com> Wrote in message: >> > meow2222@care2.com wrote: > >> > In any event, you can use any number of ways to make the circuit AB: >> > 1) VBE multiplier >> > 2) diode string >> > 3) Complementary class A buffers to eliminate the drop >> > >> > I'm trying to think of any integrated class B amps. Maybe motor drivers >> > where you wouldn't care about the notch. > >> LM386? > > Looks like AB: > http://i.stack.imgur.com/rHBqT.png > > > NT >
I guess it's technically class AB as I guess the output couldn't bias up to Vcc/2 at quiescense without some current through the output devices. It seems there is some difference in terminology when two diodes are used to bias the transistors right at the edge of of conduction so each is conducting exactly 180 degrees; I've seen it called either class AB or class B. Maybe this distinction doesn't matter. If it is class AB it is a very "lean" AB as the LM386 also has nasty crossover distortion. -- ----Android NewsGroup Reader---- http://usenet.sinaapp.com/
On Sun, 21 Dec 2014 20:07:48 +1000, David Eather <eather@tpg.com.au> wrote:

> On Sun, 21 Dec 2014 10:34:21 +1000, <meow2222@care2.com> wrote: > >> On Saturday, December 20, 2014 8:57:22 PM UTC, Jan Panteltje wrote: >> >>> + >>> | >>> |/ c >>> -------| NPN >>> | |\ e >>> in ---- +|\ | R | >>> | >------===----| >>> -- - |/ | |-------- out >>> | | |/ e | >>> | --------| PNP | >>> | |\ c | >>> | | | >>> | - | >>> -------------------------- >>> >>> For some value of R the cross over is eliminated as the opamp drives >>> the load for small signals. >> >> Yes, I tried that in spice. It seems the R has to be very low for an >> opamp though. 0.65v on 8ohm load = 75mA. 75mA 0.6v = 8ohms for the R... >> plus 8R load makes a 16R load. Can a random opamp drive that? >> >> >> NT > > Jan's circuit works. R is actually there for stability not to drive the > speakers (check it out 20ma output current into 8 ohms - a whopping > 3.2mW, into a standard speaker is below the noise floor in most indoor > locations). An R about 4.7k +/- 50% should be fine.
oops 3.2mW into a pretty standard speaker gives about 60db spl @ 1 meter - audible.
> > If you already tried this and crossover was really bad there must have > been something very wrong in your circuit. Even with a 741 the cross > over distortion is above 20k.
On Mon, 22 Dec 2014 07:50:00 +1000, <meow2222@care2.com> wrote:

> On Sunday, December 21, 2014 10:08:00 AM UTC, David Eather wrote: >> On Sun, 21 Dec 2014 10:34:21 +1000, <meow2222@care2.com> wrote: >> >> > On Saturday, December 20, 2014 8:57:22 PM UTC, Jan Panteltje wrote: >> > >> >> + >> >> | >> >> |/ c >> >> -------| NPN >> >> | |\ e >> >> in ---- +|\ | R | >> >> | >------===----| >> >> -- - |/ | |-------- out >> >> | | |/ e | >> >> | --------| PNP | >> >> | |\ c | >> >> | | | >> >> | - | >> >> -------------------------- >> >> >> >> For some value of R the cross over is eliminated as the opamp drives >> >> the load for small signals. >> > >> > Yes, I tried that in spice. It seems the R has to be very low for an >> > opamp though. 0.65v on 8ohm load = 75mA. 75mA 0.6v = 8ohms for the >> R... >> > plus 8R load makes a 16R load. Can a random opamp drive that? >> > >> > >> > NT >> >> Jan's circuit works. R is actually there for stability not to drive the >> speakers (check it out 20ma output current into 8 ohms - a whopping >> 3.2mW, >> into a standard speaker is below the noise floor in most indoor >> locations). An R about 4.7k +/- 50% should be fine. >> >> If you already tried this and crossover was really bad there must have >> been something very wrong in your circuit. Even with a 741 the cross >> over >> distortion is above 20k. > > I did sim it. Putting that R in made little difference, still heavy > crossover distortion. Maybe I need to go much lower with R. How well > your average opamp can cope with very low R I'm not so sure, lets hope > Spice models it ok. > > > NT
maybe you should build it
bitrex wrote:

> meow2222@care2.com Wrote in message: >> On Sunday, December 21, 2014 6:02:25 AM UTC, bitrex wrote: >>> miso <miso@sushi.com> Wrote in message: >>> > meow2222@care2.com wrote: >> >>> > In any event, you can use any number of ways to make the circuit AB: >>> > 1) VBE multiplier >>> > 2) diode string >>> > 3) Complementary class A buffers to eliminate the drop >>> > >>> > I'm trying to think of any integrated class B amps. Maybe motor >>> > drivers where you wouldn't care about the notch. >> >>> LM386? >> >> Looks like AB: >> http://i.stack.imgur.com/rHBqT.png >> >> >> NT >> > > I guess it's technically class AB as I guess the output couldn't > bias up to Vcc/2 at quiescense without some current through the > output devices. It seems there is some difference in terminology > when two diodes are used to bias the transistors right at the > edge of of conduction so each is conducting exactly 180 degrees; > I've seen it called either class AB or class B. Maybe this > distinction doesn't matter. > > If it is class AB it is a very "lean" AB as the LM386 also has > nasty crossover distortion. > >
I never used the compound PNP, but in theory the diode drop scheme should work the same. They probably don't bias it very far in Class A. The LM321 looks kind of class B, though they claim it is AB.
"Jan Panteltje"  wrote in message news:m74nre$bsd$1@news.datemas.de...

On a sunny day (Sat, 20 Dec 2014 19:59:46 -0000) it happened "Kevin Aylward"
<ExtractkevinRemove@kevinaylward.co.uk> wrote in
<roadnTcI5ZqxSAjJnZ2dnUVZ7sidnZ2d@bt.com>:

>>wrote in message >>news:9795867e-6869-495b-9078-bf72384ef4da@googlegroups.com... > >>Something I thought I knew, but am left puzzling... > >>>Take a simple class B output pair driven directly by an opamp. Result is >>nasty crossover distortion. Now, the nfb is taken from the circuit's >>output >>>terminal, not from the opamp's >output, so surely an ideal opamp should >>>correct for the 0.6v Vbe drops. Question is why doesn't it? > >>No real op amp is ideal. i.e. has infinite gain. That's why a real one can >>not reduce x-over distortion in a pure class B stage. A class B stage has >>gain, essentially, zero. i.e. the loop gain is still, essentially, 0 X Av >>= >>0. (open loop distortion) / loop gain, with loop gain -> zero is a big >>number.
+ | |/ c -------| NPN | |\ e
> in ---- +|\ | R |
| >------===----| -- - |/ | |-------- out | | |/ e | | --------| PNP | | |\ c | | | | | - | --------------------------
>For some value of R the cross over is eliminated as the opamp drives the >load for small signals.
The classic way to do this ahmmm.. err...correctly is the "Quad Current Dumper" by Peter Walker Kevin Aylward www.kevinaylward.co.uk www.anasoft.co.uk - SuperSpice
On a sunny day (Mon, 22 Dec 2014 20:00:27 -0000) it happened "Kevin Aylward"
<ExtractkevinRemove@kevinaylward.co.uk> wrote in
<dvCdnZQu2a7G5QXJnZ2dnUVZ8hCdnZ2d@bt.com>:

>"Jan Panteltje" wrote in message news:m74nre$bsd$1@news.datemas.de... > >On a sunny day (Sat, 20 Dec 2014 19:59:46 -0000) it happened "Kevin Aylward" ><ExtractkevinRemove@kevinaylward.co.uk> wrote in ><roadnTcI5ZqxSAjJnZ2dnUVZ7sidnZ2d@bt.com>: > >>>wrote in message >>>news:9795867e-6869-495b-9078-bf72384ef4da@googlegroups.com... >> >>>Something I thought I knew, but am left puzzling... >> >>>>Take a simple class B output pair driven directly by an opamp. Result is >>>nasty crossover distortion. Now, the nfb is taken from the circuit's >>>output >>>>terminal, not from the opamp's >output, so surely an ideal opamp should >>>>correct for the 0.6v Vbe drops. Question is why doesn't it? >> >>>No real op amp is ideal. i.e. has infinite gain. That's why a real one can >>>not reduce x-over distortion in a pure class B stage. A class B stage has >>>gain, essentially, zero. i.e. the loop gain is still, essentially, 0 X Av >>>= >>>0. (open loop distortion) / loop gain, with loop gain -> zero is a big >>>number. > + > | > |/ c > -------| NPN > | |\ e > in ---- +|\ | R | > | >------===----| > -- - |/ | |-------- out > | | |/ e | > | --------| PNP | > | |\ c | > | | | > | - | > -------------------------- > >>For some value of R the cross over is eliminated as the opamp drives the >>load for small signals. > > >The classic way to do this ahmmm.. err...correctly is the "Quad Current >Dumper" by Peter Walker
Yes I am familiar with that, a lot actually, used it in a low frequency transmitter in the UK. That was around 1982?
On Sunday, December 21, 2014 11:04:43 PM UTC, Tom Swift wrote:
> meow2222@care2.com wrote: > > > I think I follow. I can see possible issues, will spice it later and > > see > > Post your LTspice file so we can check it. > > > NT
I will - not been able to spice anything in last 24 hours tho, its probably not the best time of year. Will post it as soon as its simmed and have sorted the hardware issue. Computers! NT
On Saturday, December 20, 2014 8:55:42 AM UTC, meow...@care2.com wrote:
> Something I thought I knew, but am left puzzling... >=20 > Take a simple class B output pair driven directly by an opamp. Result is =
nasty crossover distortion. Now, the nfb is taken from the circuit's output= terminal, not from the opamp's output, so surely an ideal opamp should cor= rect for the 0.6v Vbe drops. Question is why doesn't it?
>=20 > The other question is what strategies exist to tackle it. > - bias the output trs to class AB > - helper resistor from opamp output to circuit output, so the opamp power=
s the output directly from -0.6 to +0.6v
> - low Vbe transistors, ie germanium > - a diode drop in the feedback path seems to reduce it to a degree >=20 > Any techniques I've missed? >=20 >=20 > NT
Thanks to everyone for all the help/explanations/ideas so far. I don't know= what I'll get time to do over the hols, but when I can I'll sim the sugges= tions and see what I can get to work. One of the issues with some circuits is that output tr Vbe drops to about 0= .35v at 175C (Tj). Diodes glued onto output trs would only rise to Tcase, r= esulting, in a basic implemetation, in runaway bias. Everything can be fixe= d with more bits of course, but I'm trying to do this one as rock bottom as= possible. Enjoy your xmas! NT
meow2222@care2.com wrote:
 
> One of the issues with some circuits is that output tr Vbe drops to about > 0.35v at 175C (Tj). Diodes glued onto output trs would only rise to Tcase, > resulting, in a basic implemetation, in runaway bias. Everything can be > fixed with more bits of course, but I'm trying to do this one as rock > bottom as possible. > > Enjoy your xmas! > > > NT
Um 175 dec C? You have ceramic packages? You have characterized a diode on the chip to determine Tj? Diodes on the output? Glue? This would go far better if you just stated what you are trying to achieve rather than discuss op amp design.
On Tuesday, December 23, 2014 8:45:40 AM UTC, miso wrote:
> meow2222@care2.com wrote: > =20 > > One of the issues with some circuits is that output tr Vbe drops to abo=
ut
> > 0.35v at 175C (Tj). Diodes glued onto output trs would only rise to Tca=
se,
> > resulting, in a basic implemetation, in runaway bias. Everything can be > > fixed with more bits of course, but I'm trying to do this one as rock > > bottom as possible. > >=20 > > Enjoy your xmas!
> Um 175 dec C? You have ceramic packages? You have characterized a diode o=
n=20
> the chip to determine Tj? Diodes on the output? Glue?
Standard plastic/metal trannies. Real life Tj will vary a good bit between = samples, since the specs of the parts used are either unknown or ignored. W= hether its 175 or 125, Vbe still shifts a long way.
> This would go far better if you just stated what you are trying to achiev=
e=20
> rather than discuss op amp design.
Like my other recent projects, getting a job done at the lowest possible pa= rts cost with labour as unskilled and ill equipped as possible, and using o= nly scrap parts with no data. I've got a simple class A amp, now want a B f= or battery apps. The point of the exercise is to provide designs that some = people in poverty might learn to build to sell. NT