Prev: Watch replacemnt (button) batteries?
Next: Low Power Satellite Based Laser / Imaging System Could Easily Track Activity Disturbing Cheap Reflecting Fibers
From: George Herold on 29 Jan 2010 15:55
On Jan 29, 2:41 pm, "pimpom" <pim...(a)invalid.invalid> wrote:
> George Herold wrote:
> > On Jan 27, 9:51 pm, Jon Kirwan <j...(a)infinitefactors.org>
> > wrote:
>
> > "I'd probably replace the two diodes with
> > one of those BJT and a few resistor constructions I can't
> > remember the name of (which allows me to adjust the drop.)"
>
> > First Jon I know less about amplifier design than you do...
> > That said,
> > I would be careful about replacing the diodes in the push-pull
> > stage.
> > Way back in college I had a Sony stero amp that I had to fix.
> > It came
> > with a nice circuit diagram.   I seem to recall that the bias
> > diodes
> > in the push pull stage were thermally attached to the same heat
> > sink
> > that held the output transistors.  As the output transistors
> > warm up
> > their Vbe drop decreases.  You want the bias diodes to track
> > this
> > change.  Or else the whole thing could 'run-away' on you.  ...
> > degenerative emmiter resistors (as you suggest) will help some.
>
> I like the biasing scheme mentioned by Jon and use it for all my
> designs except the early ones using germanium transistors, though
> I don't know the name either. The biasing transistor can be
> mounted on the output transistors' heatsink for temperature
> tracking.
>
> I like it because it's versatile and a single transistor can be
> used to bias several transistors with their b-e junctions in
> series as long as they are mounted on a common heatsink.http://img691.imageshack.us/img691/2075/bias.png
>
> My personal preference is to place the bias adjustment pot R3 in
> this position rather than with R1. It ensures that any accidental
> loss of contact by the pot's wiper arm will reduce the total bias
> whereas placing it with R1 will have the opposite effect and
> could cause excessive quiescent current in the output
> transistors, possibly getting them to overheat.- Hide quoted text -
>
> - Show quoted text -

Ahh Thanks Pimpom, I've never done any temperature tracking type
calculations.

George H.
From: David Eather on 29 Jan 2010 16:49
On Jan 28, 12:51 pm, Jon Kirwan <j...(a)infinitefactors.org> wrote:
> On Thu, 28 Jan 2010 11:17:02 +1000, David Eather
>
>
>

Sorry Jon,
I'm stuck on google groups for a little while - I can't believe people
actually use it full time or that google could make an interface this
bad. (I suspect it is very fine for simple threads) anyway...

>
>
> <eat...(a)tpg.com.au> wrote:
> >Jon Kirwan wrote:
> >> On Wed, 27 Jan 2010 17:31:00 +1000, David Eather
> >> <eat...(a)tpg.com.au> wrote:
> >>> <snip>
>
> >>> My particular bias for an amp this size is to go class AB with a split
> >>> power supply. The majority of quality audio amps follow this topology
> >>> and this is, I think, I great reason to go down this design path (what
> >>> you learn is applicable in the most number of situations). I should hunt
> >>> down a schematics of what I'm seeing in the distance (which can/will
> >>> change as decisions are made) - some of the justifications will have to
> >>> wait
>
> >> I'm fine with taking things as they come.
>
> >> As far as the class, I guessed that at 10 watts class-A would
> >> be too power-hungry and probably not worth its weight but
> >> that class-AB might be okay.
>
> >> I have to warn you, though, that I'm not focused upon some
> >> 20ppm THD.  I'd like to learn, not design something whose
> >> distortion (or noise, for that matter) is around a bit on a
> >> 16-bit DAC or less.  I figure winding up close to class-B
> >> operation in the end.  But I'd like to take the walk along
> >> the way, so to speak.
>
> >10 watts / PPM thd? Mmmm... maybe more like .1 - .05 % are realistic and
> >a few detours to see what would help or harm that.
>
> Hehe.  I'm thinking of some numbers I saw in the area of
> .002% THD.  I hate percentages and immediately convert them.
> In this case, it is 20e-6 or 20 ppm.  Which is darned close
> to a bit on a 16-bit dac.  That's why I wrote that way.  I
> just don't like using % figures.  They annoy me just a tiny
> bit.

Sorry.

>
> Regarding .1% to .05%, I'm _very_ good with that.  Of course,
> I'm going to have to learn about how to estimate it from
> theory as well as measure it both via simulation before
> construction and from actual testing afterwards.  More stuff
> I might _think_ I have a feel for, but I'm sure I will
> discover I don't as I get more into it.

A little experience will get you into the right ballpark when
estimating what you could expect for distortion. It is basically the
same "rules" as you would see with op-amps - the more linear it is to
start with the better. Higher bandwidth stages generally mean you can
use more negative feedback to eliminate distortion - but the lower the
final gain the more instability is likely to become a problem. And bad
circuit layout can increase distortion (and even more so hum and
noise) easily by a factor of 10.

As for how low you need distortion to be one rule of thumb (I forget
the reference) is to be clearly audible the message must be 20db above
the background noise and to be inaudible distortion has to be 20db
below the background noise - which pretty much sets "low" distortion
for PA and similar uses at 1% or 10000 ppm. For HiFi the "message" has
a high dynamic range and you (allegedly) want a distortion figure at
least 20db below that. So a 60 db signal range 0.0001% (or 100PPM).
The you start getting into all kinds of trouble with power output /
dynamic range of the amp etc and you relies that it is all a
compromise anyway. You do the best you can within the restrictions of
the job description.

>
> But speaking from ignorance, I'm good shooting for the range
> you mentioned.  It was about what I had in mind, in fact,
> figuring I could always learn as I go.
>
>
>
>
>
> >>> The first step is to think about the output. The basic equations are
>
> >>> (1).....Vout = sqrt(2*P*R)
>
> >>> With R as 8 ohms for a common speaker and 10 watts that is 12.7 volts -
> >>> actually +/- 12.7 volts with a split power supply.
>
> >> If you don't mind, I'd like to discuss this more closely. Not
> >> just have it tossed out.  So, P=V*I; or P=Vrms^2/R with AC.
> >> Using Vpeak=SQRT(2)*Vrms, I get your Vpeak=SQRT(2*P*R)
> >> equation.  Which suggests the +/-12.7V swing.  Which further
> >> suggests, taking Vce drops and any small amounts emitter
> >> resistor drops into account, something along the lines of +/-
> >> 14-15V rails?
>
> >> Or should the rails be cut a lot closer to the edge here to
> >> improve efficiency.  What bothers me is saturation as Vce on
> >> the final output BJTs goes well below 1V each and beta goes
> >> away, as well, rapidly soaking up remaining drive compliance.
>
> >>> (2).....Imax = sqrt(2*P/R)
>
> >>> This comes out to 1.6 amps. You should probably also consider the case
> >>> when R speaker = 4 ohms when initially selecting a transistor for the
> >>> output 2.2 amps - remember this is max output current. The power supply
> >>> voltage will have to be somewhat higher than Vout to take into account
> >>> circuit drive requirements, ripple on the power supply and transformer
> >>> regulation etc.
>
> >> Okay.  I missed reading this when writing the above.  Rather
> >> than correct myself, I'll leave my thinking in place.
>
> >> So yes, the rails will need to be a bit higher.  Agreed.  On
> >> this subject, I'm curious about the need to _isolate_, just a
> >> little, the rails used by the input stage vs the output stage
> >> rails.  I'm thinking an RC (or LC for another pole?) for
> >> isolation.  But I honestly don't know if that's helpful, or
> >> not.
>
> >Mostly not needed, if you use a long tailed pair for the input / error
> >amplifier, but you might prefer some other arrangement so keep it in
> >mind if your circuit "motorboats"
>
> Okay.  I've _zero_ experience for audio.  It just crossed my
> mind from other cases.  I isolate the analog supply from the
> digital -- sometimes with as many as four caps and three
> inductor beads.  There, it _does_ help.
>
>
>
> >>> Are you OK with connecting mains to a transformer? or would you rather
> >>> use an AC plug pack (10 watts is about the biggest amp a plugpack can be
> >>> used for)? The "cost" for using an AC plug pack is you will need larger
> >>> filter capacitors.
>
> >> I'd much prefer to __avoid__ using someone else's "pack" for
> >> the supply.  All discrete parts should be on the table, so to
> >> speak, in plain view.  And I don't imagine _any_ conceptual
> >> difficulties for this portion of the design.  I'm reasonably
> >> familiar with transformers, rectifiers, ripple calculations,
> >> and how to consider peak charging currents vs averge load
> >> currents as they relate to the phase angles available for
> >> charging the caps.  So on this part, I may need less help
> >> than elsewhere.  In other words, I'm somewhat comfortable
> >> here.
>
> >Ah, then there are questions of what voltage and VA for a transformer.
> >So there are questions of usage (music, PA, PA with an emergency alert
> >siren tied in etc) and rectifier arrangement and capacitor size /
> >voltage to get your required voltage output at full load.
>
> I figure on working out the design of the amplifier and then
> going back, once that is determined and hashed out, with the
> actual required figures for the power supply and design that
> part as the near-end of the process.  Earlier on, I'd expect
> to have some rough idea of how "bad" it needs to be -- if the
> initial guesses don't raise alarms, then I wouldn't dig into
> the power supply design until later on.  The amplifier, it
> seems to me, dictates the parameters.  So that comes later,
> doesn't it?
>
>

Yes and No. All the published circuits are made by people who want to
sell transistors, not audio systems, power supplies or transformers.
As a result the power supply is often assumed to be regulated, which
is not true in this case, or the power supply is treated in a very
perfunctory manner that is not at all compatible with good design.

In this case you have the voltage you need for the 10 watts, plus
voltage drop for the driver circuitry and output stage , plus ripple
voltage, plus whatever is required for transformer regulation and
mains regulation. When you add it all up you might find that a chosen
transistor/component is actually not at all suitable for the job. Back
to the drawing board. Change this change that recheck everything again
etc.

If you do the power supply first you have the figures needed for your
worst case already. It saves time and makes a better result (no
tendency to comprimise to save all the calculations already done).

>
>
>
> >>> I should also ask if you have a multi meter, oscilloscope (not necessary
> >>> but useful)and how is your soldering? But it would be wise to keep this
> >>> whole thing as a paper exercise before you commit to anything.
>
> >> I have a 6 1/2 digit HP multimeter, a Tek DMM916 true RMS
> >> handheld, two oscilloscopes (TEK 2245 with voltmeter option
> >> and an HP 54645D), three triple-output power supplies with
> >> two of them GPIB drivable, the usual not-too-expensive signal
> >> generator, and a fair bunch of other stuff on the shelves.
> >> Lots of probes, clips, and so on.  For soldering, I'm limited
> >> to a Weller WTCPT and some 0.4mm round, 0.8mm spade, and
> >> somewhat wider spade tips in the 1.5mm area.  I have tubs and
> >> jars of various types of fluxes, as well, and wire wrap tools
> >> and wire wrap wire, as well.  I also have a room set aside
> >> for this kind of stuff, when I get time to play.
>
> >OK. Next serious project, I'm coming around to your place!
>
> You come to the west coast of the US and I'll have a room for
> you!
>
> >Your gear is
> >better than mine. I had to ask, rather than just assume just in case my
> >assumptions got you building something you didn't want to, and got you
> >splattered all over the place from the mains, or suggesting you choose
> >the miller cap by watching the phase shift of the feedback circuit - I
> >don't read a lot of the posts so I didn't know what you could do.
>
> To be honest, I can do a few things but I'm really not very
> practiced.  My oscilloscope knowledge is lacking in some
> areas -- which becomes all too painfully obvious to me when I
> watch a pro using my equipment.  And I'm still learning to
> solder better.  It's one of a few hobbies.
>
> >> Jon
>
> >Have a look at
> >http://en.wikipedia.org/wiki/Electronic_amplifier
>
> Done.
>
> >The bits on class A might be interesting as it says 25% efficiency and
> >50% obtainable with inductive output coupling (i.e. with a transformer)
> >which is what I said, not what blow hard Phil said.
>
> What I first see there is the amplifier sketch at the top of
> the page

I wasn't going to prompt, but it is close to the sort of thing, I
think, you should be aiming for . As someone has already noted (I
would attribute you if I wasn't on GG, I'm sorry) it has been drawn up
for a single supply, rather than a more common (for this size /
configuration) split supply.

(I don't really care too much about arguing about
> efficiencies right now -- I'm more concerned about learning.)
> The input stage shown is a voltage-in, current-out bog
> standard diff-pair.  First thing I remember about is that R4
> shouldn't be there

Correct. Theory says it does nothing. I practice the theory but have
the occasional heretical belief about that.

and better still both R3 and R4 should be
> replaced with a current mirror.  

This would provide more differential gain.

R5 should be a replaced with
> a BJT, as well.  

In the right configuration it would reduce the common mode signal gain
of things like mains hum and supply ripple (you mentioned power supply
isolation before).
Also, from another (what do you call it branch? thread?) you were
discussing boot-strapping R6. This is not done so much as amplifiers
get bigger but a BJT configured in the same way as the replacement for
R5 is very common. I'm leaving the details to you - perhaps there is a
way to reduce component count without affecting performance. (I am
hoping this is what you wanted "nutting it out for yourself")

I assume the input impedance of that example
> is basically the parallel resistance of R1 and R2, but if we
Yes.

> use split supplies I'd imagine replacing the two of them with
> a single resistor to the center-ground point.  
Yes, but you should probably think of a whole passive network to
filter out low and high frequency - (think what happens if you amp is
operated near a source of RF)
There's no
> miller cap on Q3,

Depending on transistors layout etc it might not be needed, but more
often it is the size that is the question.

I'd probably replace the two diodes with
> one of those BJT and a few resistor constructions I can't
> remember the name of (which allows me to adjust the drop.)

Vbe multiplier...

> The feedback ... well, I need to think about that a little
> more.  There's no degen resistors in the emitters of Q4 and
> Q5.
>

Why would/should you use them?

> Um.. okay, I need to sit down and think.  Mind is spinning,
> but I've not set a finger to paper yet and there is lots to
> think about in that one.  I could be way, way off base.

Not at all.

>
> Jon


Is there a way you could post a schematic of where your thinking is
and what you would like to discuss - there is no need for a complete
circuit.
From: Jon Kirwan on 29 Jan 2010 18:06
On Fri, 29 Jan 2010 13:49:16 -0800 (PST), David Eather
<eather(a)tpg.com.au> wrote:

>On Jan 28, 12:51�pm, Jon Kirwan <j...(a)infinitefactors.org> wrote:
>> On Thu, 28 Jan 2010 11:17:02 +1000, David Eather
>
>Sorry Jon,
>
>I'm stuck on google groups for a little while - I can't believe people
>actually use it full time or that google could make an interface this
>bad. (I suspect it is very fine for simple threads) anyway...

Cripes. Google didn't even show the thread when I'd looked,
a day ago or so. And it had been around for at least 24
hours by then. Used to be the case that google groups would
show the posts within an hour or so. Doesn't seem to be
true, anymore. If not, there is no possibility of having a
discussion very quickly via google. It would greatly
lengthen out the interactions. Maybe that's on purpose, now,
to cause people to find some other solution?

>> <eat...(a)tpg.com.au> wrote:
>> >Jon Kirwan wrote:
>> >> On Wed, 27 Jan 2010 17:31:00 +1000, David Eather
>> >> <eat...(a)tpg.com.au> wrote:
>> >>> <snip>
>>
>> >>> My particular bias for an amp this size is to go class AB with a split
>> >>> power supply. The majority of quality audio amps follow this topology
>> >>> and this is, I think, I great reason to go down this design path (what
>> >>> you learn is applicable in the most number of situations). I should hunt
>> >>> down a schematics of what I'm seeing in the distance (which can/will
>> >>> change as decisions are made) - some of the justifications will have to
>> >>> wait
>>
>> >> I'm fine with taking things as they come.
>>
>> >> As far as the class, I guessed that at 10 watts class-A would
>> >> be too power-hungry and probably not worth its weight but
>> >> that class-AB might be okay.
>>
>> >> I have to warn you, though, that I'm not focused upon some
>> >> 20ppm THD. �I'd like to learn, not design something whose
>> >> distortion (or noise, for that matter) is around a bit on a
>> >> 16-bit DAC or less. �I figure winding up close to class-B
>> >> operation in the end. �But I'd like to take the walk along
>> >> the way, so to speak.
>>
>> >10 watts / PPM thd? Mmmm... maybe more like .1 - .05 % are realistic and
>> >a few detours to see what would help or harm that.
>>
>> Hehe. �I'm thinking of some numbers I saw in the area of
>> .002% THD. �I hate percentages and immediately convert them.
>> In this case, it is 20e-6 or 20 ppm. �Which is darned close
>> to a bit on a 16-bit dac. �That's why I wrote that way. �I
>> just don't like using % figures. �They annoy me just a tiny
>> bit.
>
>Sorry.

Don't be. I was just explaining myself, not complaining
about your usage.

>> Regarding .1% to .05%, I'm _very_ good with that. �Of course,
>> I'm going to have to learn about how to estimate it from
>> theory as well as measure it both via simulation before
>> construction and from actual testing afterwards. �More stuff
>> I might _think_ I have a feel for, but I'm sure I will
>> discover I don't as I get more into it.
>
>A little experience will get you into the right ballpark when
>estimating what you could expect for distortion. It is basically the
>same "rules" as you would see with op-amps - the more linear it is to
>start with the better. Higher bandwidth stages generally mean you can
>use more negative feedback to eliminate distortion - but the lower the
>final gain the more instability is likely to become a problem. And bad
>circuit layout can increase distortion (and even more so hum and
>noise) easily by a factor of 10.
>
>As for how low you need distortion to be one rule of thumb (I forget
>the reference) is to be clearly audible the message must be 20db above
>the background noise and to be inaudible distortion has to be 20db
>below the background noise - which pretty much sets "low" distortion
>for PA and similar uses at 1% or 10000 ppm. For HiFi the "message" has
>a high dynamic range and you (allegedly) want a distortion figure at
>least 20db below that. So a 60 db signal range 0.0001% (or 100PPM).
>The you start getting into all kinds of trouble with power output /
>dynamic range of the amp etc and you relies that it is all a
>compromise anyway. You do the best you can within the restrictions of
>the job description.

Understood.

>> But speaking from ignorance, I'm good shooting for the range
>> you mentioned. �It was about what I had in mind, in fact,
>> figuring I could always learn as I go.
>>
>>
>> >>> The first step is to think about the output. The basic equations are
>>
>> >>> (1).....Vout = sqrt(2*P*R)
>>
>> >>> With R as 8 ohms for a common speaker and 10 watts that is 12.7 volts -
>> >>> actually +/- 12.7 volts with a split power supply.
>>
>> >> If you don't mind, I'd like to discuss this more closely. Not
>> >> just have it tossed out. �So, P=V*I; or P=Vrms^2/R with AC.
>> >> Using Vpeak=SQRT(2)*Vrms, I get your Vpeak=SQRT(2*P*R)
>> >> equation. �Which suggests the +/-12.7V swing. �Which further
>> >> suggests, taking Vce drops and any small amounts emitter
>> >> resistor drops into account, something along the lines of +/-
>> >> 14-15V rails?
>>
>> >> Or should the rails be cut a lot closer to the edge here to
>> >> improve efficiency. �What bothers me is saturation as Vce on
>> >> the final output BJTs goes well below 1V each and beta goes
>> >> away, as well, rapidly soaking up remaining drive compliance.
>>
>> >>> (2).....Imax = sqrt(2*P/R)
>>
>> >>> This comes out to 1.6 amps. You should probably also consider the case
>> >>> when R speaker = 4 ohms when initially selecting a transistor for the
>> >>> output 2.2 amps - remember this is max output current. The power supply
>> >>> voltage will have to be somewhat higher than Vout to take into account
>> >>> circuit drive requirements, ripple on the power supply and transformer
>> >>> regulation etc.
>>
>> >> Okay. �I missed reading this when writing the above. �Rather
>> >> than correct myself, I'll leave my thinking in place.
>>
>> >> So yes, the rails will need to be a bit higher. �Agreed. �On
>> >> this subject, I'm curious about the need to _isolate_, just a
>> >> little, the rails used by the input stage vs the output stage
>> >> rails. �I'm thinking an RC (or LC for another pole?) for
>> >> isolation. �But I honestly don't know if that's helpful, or
>> >> not.
>>
>> >Mostly not needed, if you use a long tailed pair for the input / error
>> >amplifier, but you might prefer some other arrangement so keep it in
>> >mind if your circuit "motorboats"
>>
>> Okay. �I've _zero_ experience for audio. �It just crossed my
>> mind from other cases. �I isolate the analog supply from the
>> digital -- sometimes with as many as four caps and three
>> inductor beads. �There, it _does_ help.
>>
>>
>>
>> >>> Are you OK with connecting mains to a transformer? or would you rather
>> >>> use an AC plug pack (10 watts is about the biggest amp a plugpack can be
>> >>> used for)? The "cost" for using an AC plug pack is you will need larger
>> >>> filter capacitors.
>>
>> >> I'd much prefer to __avoid__ using someone else's "pack" for
>> >> the supply. �All discrete parts should be on the table, so to
>> >> speak, in plain view. �And I don't imagine _any_ conceptual
>> >> difficulties for this portion of the design. �I'm reasonably
>> >> familiar with transformers, rectifiers, ripple calculations,
>> >> and how to consider peak charging currents vs averge load
>> >> currents as they relate to the phase angles available for
>> >> charging the caps. �So on this part, I may need less help
>> >> than elsewhere. �In other words, I'm somewhat comfortable
>> >> here.
>>
>> >Ah, then there are questions of what voltage and VA for a transformer.
>> >So there are questions of usage (music, PA, PA with an emergency alert
>> >siren tied in etc) and rectifier arrangement and capacitor size /
>> >voltage to get your required voltage output at full load.
>>
>> I figure on working out the design of the amplifier and then
>> going back, once that is determined and hashed out, with the
>> actual required figures for the power supply and design that
>> part as the near-end of the process. �Earlier on, I'd expect
>> to have some rough idea of how "bad" it needs to be -- if the
>> initial guesses don't raise alarms, then I wouldn't dig into
>> the power supply design until later on. �The amplifier, it
>> seems to me, dictates the parameters. �So that comes later,
>> doesn't it?
>
>Yes and No. All the published circuits are made by people who want to
>sell transistors,

A concern I care not the least about. My _real_ preference,
were I to impose it on the design, would be to use ONLY
PN2222A BJTs for all the active devices. One part. That's
it. Why? Because I've got thousands of them. ;)

Literally. Something like 22,000 of the bastards. I give
them away like popcorn to students at schools. Got them
_very cheaply_. So if I were pushing something, I'd be
pushing a 10W PN2222A design, use signal splitting approach
probably (because it's the only way I think think of, right
now), and distribute the dissipation across lots and lots of
the things.

What to go there? :)

>not audio systems, power supplies or transformers.

Got it.

>As a result the power supply is often assumed to be regulated, which
>is not true in this case, or the power supply is treated in a very
>perfunctory manner that is not at all compatible with good design.
>
>In this case you have the voltage you need for the 10 watts, plus
>voltage drop for the driver circuitry and output stage , plus ripple
>voltage, plus whatever is required for transformer regulation and
>mains regulation. When you add it all up you might find that a chosen
>transistor/component is actually not at all suitable for the job. Back
>to the drawing board. Change this change that recheck everything again
>etc.

In this case, though, there is nothing particularly
remarkable about the rails. Taken across the entire span,
even, doesn't exceed the maximum Vce of a great many BJTs. So
no real worry there. But I see some of where problems may
arise. Luckily, at this level I can side-step worrying about
that part and get back to learning about amplifier design,
yes?

>If you do the power supply first you have the figures needed for your
>worst case already. It saves time and makes a better result (no
>tendency to comprimise to save all the calculations already done).

Well, does this mean we should hack out the power supply
first? I'm perfectly fine with that and can get back to you
with a suggested circuit and parts list if you want to start
there. We could settle that part before going anywhere else
and I'd be happy with that approach, too, because to be
honest I don't imagine it to put a horrible delay into
getting back to amplifier design. So I'm good either way.

>> >>> I should also ask if you have a multi meter, oscilloscope (not necessary
>> >>> but useful)and how is your soldering? But it would be wise to keep this
>> >>> whole thing as a paper exercise before you commit to anything.
>>
>> >> I have a 6 1/2 digit HP multimeter, a Tek DMM916 true RMS
>> >> handheld, two oscilloscopes (TEK 2245 with voltmeter option
>> >> and an HP 54645D), three triple-output power supplies with
>> >> two of them GPIB drivable, the usual not-too-expensive signal
>> >> generator, and a fair bunch of other stuff on the shelves.
>> >> Lots of probes, clips, and so on. �For soldering, I'm limited
>> >> to a Weller WTCPT and some 0.4mm round, 0.8mm spade, and
>> >> somewhat wider spade tips in the 1.5mm area. �I have tubs and
>> >> jars of various types of fluxes, as well, and wire wrap tools
>> >> and wire wrap wire, as well. �I also have a room set aside
>> >> for this kind of stuff, when I get time to play.
>>
>> >OK. Next serious project, I'm coming around to your place!
>>
>> You come to the west coast of the US and I'll have a room for
>> you!
>>
>> >Your gear is
>> >better than mine. I had to ask, rather than just assume just in case my
>> >assumptions got you building something you didn't want to, and got you
>> >splattered all over the place from the mains, or suggesting you choose
>> >the miller cap by watching the phase shift of the feedback circuit - I
>> >don't read a lot of the posts so I didn't know what you could do.
>>
>> To be honest, I can do a few things but I'm really not very
>> practiced. �My oscilloscope knowledge is lacking in some
>> areas -- which becomes all too painfully obvious to me when I
>> watch a pro using my equipment. �And I'm still learning to
>> solder better. �It's one of a few hobbies.
>>
>> >> Jon
>>
>> >Have a look at
>> >http://en.wikipedia.org/wiki/Electronic_amplifier
>>
>> Done.
>>
>> >The bits on class A might be interesting as it says 25% efficiency and
>> >50% obtainable with inductive output coupling (i.e. with a transformer)
>> >which is what I said, not what blow hard Phil said.
>>
>> What I first see there is the amplifier sketch at the top of
>> the page
>
>I wasn't going to prompt, but it is close to the sort of thing, I
>think, you should be aiming for . As someone has already noted (I
>would attribute you if I wasn't on GG, I'm sorry) it has been drawn up
>for a single supply, rather than a more common (for this size /
>configuration) split supply.

I had assumed we'd be using a split supply.

I had assumed a speaker would be hooked up via a cap to the
output, so DC currents into a speaker coil would be removed
from any concern. But I was also holding in the back of my
mind the idea of tweaking out DC bias via the speaker and
removing the coupling cap as an experiment to try. And if
so, I'd pretty much want the ground as a "third rail."
(Playing just a bit upon the Chicago parlance about the once
dangerous rail in their transit system.)

>(I don't really care too much about arguing about
>> efficiencies right now -- I'm more concerned about learning.)
>> The input stage shown is a voltage-in, current-out bog
>> standard diff-pair. �First thing I remember about is that R4
>> shouldn't be there
>
>Correct. Theory says it does nothing. I practice the theory but have
>the occasional heretical belief about that.

Actually, I think I've read that theory says it is _better_
to be removed. The reason seemed pretty basic, as it's
easier to get close to a balanced current split; and this, I
gather, lowers 2nd harmonic distortions produced in the pair
-- notable more on the high frequency end I suppose because
gain used for linearizing feedback up there is diminishing
and can't compensate it.

In other words, it's not neutral. It's considered to be
better if I gathered the details. Then even better, the
current mirror enforces the whole deal and you've got about
the best to be had.

Of course, mostly just being a reader means I have no idea
which end is up. So I might have all this wrong.

> and better still both R3 and R4 should be
>> replaced with a current mirror. �
>
>This would provide more differential gain.

_and_ improve distortion because the currents are forced to
be balanced in the pair, yes?

>>R5 should be a replaced with
>> a BJT, as well. �
>
>In the right configuration it would reduce the common mode signal gain
>of things like mains hum and supply ripple (you mentioned power supply
>isolation before).

Yes, that's how I thought about it.

>Also, from another (what do you call it branch? thread?) you were
>discussing boot-strapping R6. This is not done so much as amplifiers
>get bigger but a BJT configured in the same way as the replacement for
>R5 is very common. I'm leaving the details to you - perhaps there is a
>way to reduce component count without affecting performance. (I am
>hoping this is what you wanted "nutting it out for yourself")

Yes! I don't want things handed on a platter. But I also
don't want to have to rediscover all of the ideas by making
all of the mistakes, either. This is the kind of "pointer"
towards something that I like a lot. It gives me a place to
think about something, but leaves me some reason to have to
do so and that helps me own it better.

One general truth about learning is that you don't present
someone with a problem so out of their depth that they have
no chance at it. Doing that means they fail, they feel like
a failure, and it causes a student to just want to go away.
They lose motivation, usually, in cases like that. On the
other hand, providing no difficulty at all merely means
repetition of what they already know and they grow bored from
that, too. Finding the sweet spot where a student is faced
with interesting problems that are not already known, but
perhaps within reach of grasping at with some effort, is the
key. Then it can be fun, educational, and motivate.

That's what you just did for me.

>I assume the input impedance of that example
>> is basically the parallel resistance of R1 and R2, but if we

>Yes.

Okay.

>> use split supplies I'd imagine replacing the two of them with
>> a single resistor to the center-ground point. �
>Yes, but you should probably think of a whole passive network to
>filter out low and high frequency - (think what happens if you amp is
>operated near a source of RF)

Well, every trace picks up like little antennae. All kinds
of trace voltages appearing here and there. Not good.

So. Can you make an audio amplifier that can withstand a
microwave oven environment and deliver good performance while
irradiated with 1kW banging around in there? ;)

>> There's no
>> miller cap on Q3,
>
>Depending on transistors layout etc it might not be needed, but more
>often it is the size that is the question.

I was thinking it helped locally linearize the VAS section
and that such would be "good" most anywhere. But I am just
taking things without having worked through them on my own.
So...

>> I'd probably replace the two diodes with
>> one of those BJT and a few resistor constructions I can't
>> remember the name of (which allows me to adjust the drop.)
>
>Vbe multiplier...

Okay. Thanks.

>> The feedback ... well, I need to think about that a little
>> more. �There's no degen resistors in the emitters of Q4 and
>> Q5.
>
>Why would/should you use them?

I'm still thinking about that. In general, I was thinking
about them because of the "little re" that is kT/q based in
each BJT, and varies on Ie. Since Ie is varying around, I
was thinking about something fixed there to overwhelm it and
"make it knowable" for the design, I suppose. Maybe that's
all wet, given your query. I'll toss the idea off the side,
for now.

>> Um.. okay, I need to sit down and think. �Mind is spinning,
>> but I've not set a finger to paper yet and there is lots to
>> think about in that one. �I could be way, way off base.
>
>Not at all.

Thanks for that. I'm just glad to be able to talk to someone
about any of this, at all. So please accept my thanks for
the moments you are offering.

>Is there a way you could post a schematic of where your thinking is
>and what you would like to discuss - there is no need for a complete
>circuit.

Yes. I can use ASCII here, for example. But before I go off
into the wild blue with this, do you want to focus on the
power supply first? Or just jump in on the amplifier?

Jon
From: Jon Kirwan on 29 Jan 2010 18:22
On Fri, 29 Jan 2010 15:06:40 -0800, I wrote:

>A concern I care not the least about. My _real_ preference,
>were I to impose it on the design, would be to use ONLY
>PN2222A BJTs for all the active devices. One part. That's
>it. Why? Because I've got thousands of them. ;)
>
>Literally. Something like 22,000 of the bastards. I give
>them away like popcorn to students at schools. Got them
>_very cheaply_. So if I were pushing something, I'd be
>pushing a 10W PN2222A design, use signal splitting approach
>probably (because it's the only way I think think of, right
>now), and distribute the dissipation across lots and lots of
>the things.
>
>What to go there? :)

I meant "Want to go there?"

One part of the brain says "type out word for concept X on
the keyboard," and that gets passed down to a low-level
manager function which maps concept to English word, gets the
wrong hash bucket ID for a "nearby word" and the passes on
motor instructions for mirror neurons driving the hand which
then types the wrong word. Meanwhile, eye reads "want" text,
this gets translated into a concept which is promptly ignored
because the concept already resides in local cache storage
and doesn't need replacement. So the brain checks that what
I wrote is what I intended, glancing quickly over it and gets
the cached version and matches everything up nicely and moves
on.

hehe. This is either a hardware problem or a software
problem, depending upon point of view... ;)

Jon
From: Jon Kirwan on 29 Jan 2010 18:23
On Fri, 29 Jan 2010 15:22:42 -0800, Jon Kirwan
<jonk(a)infinitefactors.org> wrote:

>eye reads "want" text

hehe. Or "What" text....

Nesting all the way down...

Jon
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Prev: Watch replacemnt (button) batteries?
Next: Low Power Satellite Based Laser / Imaging System Could Easily Track Activity Disturbing Cheap Reflecting Fibers