Are these 2N3055s borderline spec?

[Terry Pinnell]

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry,

I ran SWCAD on your amp, and got the following results:

1. Maximum output before clipping 10V peak, or 7 VRMS.
2. To get symmetrical clipping I had to change R2 to 51K. Since the bias
point depends on the VBE of Q1 and Q2, a little trimming makes sense.
3. Input signal for 10V peak output was around 0.1V peak. Since R3 and R1
set the closed loop voltage gain at 100, this makes sense.
4. I haven't looked at this yet, but I suspect you have very little negative
feedback. I will take a look at what your open loop gain is.
5. Your problem might have to do with leakage currents of Q5 and Q6 having
nowhere to go which might explain why it worked when you switched the
transistors.

6. With no signal, the DC voltage at the collector of Q6 should be about 12
V. If not, you got problems - see 5).

Excellent, Tam, thanks for going to that trouble. You're reporting a
virtually identical result to the (physical) one I posted yesterday
morning. Did you see that?
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay22-1.gif

That was in the main thread, at
Newsgroups: sci.electronics.design
Subject: Re: Distortion from audio power amp
Date: Sat, 22 May 2004 11:17:26 +0100
Message-ID: <2v9ua0h8fu8d5mh8bhjhnq4eooim544i0t@4ax.com>

Also, did you see my observations here
Newsgroups: sci.electronics.design
Subject: Re: Distortion from audio power amp
Date: Fri, 21 May 2004 12:35:02 +0100
Message-ID: <rfora0h95ts066ii4i72qe8jgnstesuvjt@4ax.com>

about trimming R1? My actual R1 is a measured 53R (not the nominal 47
I drew), yet I had to adjust it to about 72R to get a simulation from
CircuitMaker that matched the actual (i.e waveform unclipped amplitude
and DC level precisely 12.0 V). That fairly wide gap still puzzles me
a bit...

I have now tested the very latest circuit, incorporating a few small
changes:
http://www.terrypin.dial.pipex.com/Images/PushPull-May22-1.gif

and it gave what appeared to be identical visual results. I nudged up
the input and I convinced myself I was reading 7.00 V rms on my DMM,
with no visible clipping.

I have yet to follow up R Legg's observation (albeit sarcastically
made!) about the input showing more distortion than the output. It's
small, but plainly visible at the cross-over point.

Much appreciate your interest, and hope to hear more from you.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Fred Bloggs]

Winfield Hill wrote:

Fred Bloggs wrote...

Stefan Heinzmann wrote:

Fred Bloggs wrote:


Okay- I have located this amplifier in the OLD-OLD-OLD literature- we
are talking state of the art developmental work presented to the Audio
Engineering Society circa 1957!-LOL.

Good forensic work!

Have you got the AES preprint number?
Could be Preprint 6 or 11 maybe...

Author is Marvin B. Herscher, Radio Corporation of America, Camden, NJ
Presented to Audio Engineering Society Convention, New York, October 1957
My source is a collection of Electronics magazine articles from the era.


It'd be great if you could scan and post that presentation / article.
I'd like to compare it to H.C. Lin's similar design, c. 1956 - 1957.

Thanks,
- Win

(email: use hill_at_rowland-dot-org for now)

I will get a scanner and post a TIFF or something to abse. I don't have
the famous 1956 Lin article, but from what I find on the web, he used a
thermistor rather than diodes in the preamp collector circuit- otherwise
the circuit is identical to Herscher's.

 

[Fred Bloggs]

Spehro Pefhany wrote:

On Sat, 22 May 2004 16:55:28 GMT, the renowned Fred Bloggs
nospam@nospam.com> wrote:



Speaking of forensic work, there was CNN article yesterday about the US
NTSB re-opening the investigation into a Cessna Caravan cargo airplane
crash on the night of Oct 23, 2002 near Mobile, Alabama. To date the
exact cause is unknown and the wreckage was characterized as a mid-air
impact that sliced the engine in half, splattered elements of the
fuselage with a polyethylene based red colored material,


Nooooo, NOT polyethelene. Something else.

http://www.ntsb.gov/ntsb/brief2.asp?ev_id=20021029X05400&ntsbno=ATL03FA008&akey=1

"polyester materials based on tere- and iso- phthalates. The spectra
also suggested the possible presence of inorganic silicate compounds."

Maybe something like this boating type fiberglass fuel tank

http://continuouswave.com/whaler/reference/images/redPate531x340.jpeg

made by:

Pate Plastics Inc
(305) 754-0896 360 NW 71st St
Miami, FL

The sort of thing that would be familar to the kind of people you are
talking about.


and left an
embedded piece of mystery anodized aluminum. The only other aircraft in
the vicinity was a FedEx DC-10 traveling on a perpendicular trajectory
1000 feet above the victim aircraft at 4000 feet. It is obvious to me
that the ONLY credible explanation is a midair collision with an
"unscheduled" delivery being made by that damned FedEx DC 10- a small
RCS tightly bound volumetrically dense bundle that hit that Cessna like
a cannon ball- dropped approximately 10 seconds prior to impact so that
it retained a substantial amount of initial horizontal velocity. The
anodized aluminum is most likely the remnants of an RF package most
likely for GPS location of the delivery by the ground recovery
recipients- what little drug residue that did attach to the victim
aircraft was dissolved or washed away by the watery crash site. It is
also an incredible coincidence that this is a high activity drug
smuggling area- and the FAA allowed their ATC data recording capability
to be out of service for months. This scenario is too perfect- low
altitude drop requiring no suspicious flight path deviation- small RCS
delivery designed to free fall off the radar screen in minimum time-
altitude activated parachute at safe unobservable height- GPS location
of drop in remote inaccessible area. Who the hell do these people think
they are fooling?


But how do you get stuff to drop out of a DC-10? Why the
cloak-and-dagger stuff if the DC-10 was already customs-cleared, a
white van would suffice, no? But I don't see what the origin of the
FedEx flight was....

Best regards,
Spehro Pefhany

I am fairly certain that at 4000 feet and 150 knots or less- you can
just kick it out- or drop it from the septic tank- these have been known
to dump big loads. Why the drop? Who knows- the most likely explanation
would be a hole in security that makes it easier to put something on
than take it off- and the take it through drug sniffing dogs or some
other form of screening- and the less people they have to pay off , the
more profit. The liquid tight plastic enclosure would make sense for a
septic tank ride and/or water landing- it should have landed in tidal
marsh around Mobile Bay- the crash victim certainly did. The thing is
that NTSB is strictly a safety investigator by law- and what's more,
their findings cannot be used as evidence in a court of law. Any
reasonable suspicion of criminal activity is supposed to make the FBI
the lead investigator. It's beyond coincidence that immediately after
the victim pilot radio acknowledges the DC-10 is right overhead, the
plane takes a major altitude dive.

 

[Terry Pinnell]

Winfield Hill <Winfield_member@newsguy.com> wrote:

Fred Bloggs wrote...

Stefan Heinzmann wrote:
Fred Bloggs wrote:

Okay- I have located this amplifier in the OLD-OLD-OLD literature- we
are talking state of the art developmental work presented to the Audio
Engineering Society circa 1957!-LOL.

Good forensic work!

Have you got the AES preprint number?
Could be Preprint 6 or 11 maybe...

Author is Marvin B. Herscher, Radio Corporation of America, Camden, NJ
Presented to Audio Engineering Society Convention, New York, October 1957
My source is a collection of Electronics magazine articles from the era.

It'd be great if you could scan and post that presentation / article.
I'd like to compare it to H.C. Lin's similar design, c. 1956 - 1957.

I've just downloaded it. A hefty 10USD for me, as I'm not a member!

I printed it, of course, but it's plainly one of these 'secure' PDF
files and offers no File>Save. However, I located it on my HD in
temporary storage and duly saved. I figure that's the least I deserve
for ten dollars <g>. It's a 900KB file and I've emailed a copy to you
and Stefan. Not sure how I stand re copyright if I make it more widely
available, but reckon I'm entitled to pass copies for discussion to
others, yes? Can an AES member tell me the rules please?

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Terry Pinnell]

Fred Bloggs <nospam@nospam.com> wrote:

I will get a scanner and post a TIFF or something to abse. I don't have
the famous 1956 Lin article, but from what I find on the web, he used a
thermistor rather than diodes in the preamp collector circuit- otherwise
the circuit is identical to Herscher's.

No need - see my post a few minutes aferwards.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Stefan Heinzmann]

Terry Pinnell wrote:

I've just downloaded it. A hefty 10USD for me, as I'm not a member!

Excellent! So you downloaded two articles, the one in the journal and
the conference preprint, right?

I printed it, of course, but it's plainly one of these 'secure' PDF
files and offers no File>Save. However, I located it on my HD in
temporary storage and duly saved. I figure that's the least I deserve
for ten dollars <g>. It's a 900KB file and I've emailed a copy to you
and Stefan. Not sure how I stand re copyright if I make it more widely
available, but reckon I'm entitled to pass copies for discussion to
others, yes? Can an AES member tell me the rules please?

Very nice, thanks. I am a member, but don't know the details of the
rules either. I suppose it is the same as with other texts you buy, such
as books. I would guess that you may share it with us for the
discussion, but we're not allowed to keep it ;-)

I was amused to see that the author almost apologizes for wasting 1 watt
of signal power in the resistors R7/R8. This sounds a bit like the tube
age reservations against negative feedback ("we're throwing gain away").

I find it also curious to see from today's perspective that it was
deemed appropriate to omit a reasonable input stage to decouple the
amplifier from the impedance of the signal source and to allow a
symmetrical power supply that could have made the output capacitor
superfluous. This must have been because transistors were the most
expensive parts in the circuit by far.

I also was reminded that this were the times when even a decent NPN
transistor was difficult to make. Note that the one NPN transistor in
the circuit was an experimental device.

Regarding the bandwidth of the amp and Terry's problems with amplifier
stability, it is also noteworthy that the author explicitly states that
the high frequency response of his amplifier is limited by the
properties of the transistors. It shouldn't be a surprise, therefore, to
find that when more modern transistors are used the circuit can become
unstable, and the frequency response must be controlled explicitly.

It is interesting to look back almost 50 years, it gives you a
perspective on how the art has developed, how circuit economics have
changed, and also how much things have stayed the same.

--
Cheers
Stefan

 

[Tam/WB2TT]

Terry,

Some more observations. I was wrong about the leakage current. Simulated 5ma
of leakage in Q6, which had no meaningful effect on output. The only
difference in swapping Q5 and Q6 that I can see is that the crossover
distortion depends on the Vbe of Q3, Q5, and Q4, and is independent of the
Vbe of Q6. Also, see last point.

There is too little loop gain to run a closed loop gain of 100. It looks
better with a loop gain of 10. Since 470K is pushing things, I changed R3 to
47K and R2 to 5.2K. Output clips at 22 V p-p, which is 7.5W.

If you want 20 - 20000 Hz bandwidth, you need to increase the values of C1
and C3. Also, C4 starts to roll off the closed loop gain before 20 KHz. I
changed that to 1 PF.

The gain seems to depend on signal amplitude. The smaller I make the input
signal (like 1mV), the less the gain. This is more severe at the 100 x gain
setting. At 10 x gain, I do not see any crossover distortion at 1 mV input.

I have not figured out completely the action of bootstrap capacitor, C2, but
it does do its job.

My previous comment about using 2N2219s to drive the output transistors is
moot, because the Q5 and Q6 collector current is only a little over an amp.
In fact, you would get some output, even if Q5 had an open collector.
Because of R9, Q4 could not drive the speaker directly as well as Q3 can.
There might be something here.

I did print your schematic with the 5/22 date.

Tam
"Terry Pinnell" <terrypinDELETE@THESEdial.pipex.com> wrote in message
news:hod0b0thlr9bgmauq128alsegqb436l49m@4ax.com...

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry,

I ran SWCAD on your amp, and got the following results:

1. Maximum output before clipping 10V peak, or 7 VRMS.
2. To get symmetrical clipping I had to change R2 to 51K. Since the bias
point depends on the VBE of Q1 and Q2, a little trimming makes sense.
3. Input signal for 10V peak output was around 0.1V peak. Since R3 and R1
set the closed loop voltage gain at 100, this makes sense.
4. I haven't looked at this yet, but I suspect you have very little
negative
feedback. I will take a look at what your open loop gain is.
5. Your problem might have to do with leakage currents of Q5 and Q6
having
nowhere to go which might explain why it worked when you switched the
transistors.

6. With no signal, the DC voltage at the collector of Q6 should be about
12
V. If not, you got problems - see 5).

Excellent, Tam, thanks for going to that trouble. You're reporting a
virtually identical result to the (physical) one I posted yesterday
morning. Did you see that?
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay22-1.gif

That was in the main thread, at
Newsgroups: sci.electronics.design
Subject: Re: Distortion from audio power amp
Date: Sat, 22 May 2004 11:17:26 +0100
Message-ID: <2v9ua0h8fu8d5mh8bhjhnq4eooim544i0t@4ax.com

Also, did you see my observations here
Newsgroups: sci.electronics.design
Subject: Re: Distortion from audio power amp
Date: Fri, 21 May 2004 12:35:02 +0100
Message-ID: <rfora0h95ts066ii4i72qe8jgnstesuvjt@4ax.com

about trimming R1? My actual R1 is a measured 53R (not the nominal 47
I drew), yet I had to adjust it to about 72R to get a simulation from
CircuitMaker that matched the actual (i.e waveform unclipped amplitude
and DC level precisely 12.0 V). That fairly wide gap still puzzles me
a bit...

I have now tested the very latest circuit, incorporating a few small
changes:
http://www.terrypin.dial.pipex.com/Images/PushPull-May22-1.gif

and it gave what appeared to be identical visual results. I nudged up
the input and I convinced myself I was reading 7.00 V rms on my DMM,
with no visible clipping.

I have yet to follow up R Legg's observation (albeit sarcastically
made!) about the input showing more distortion than the output. It's
small, but plainly visible at the cross-over point.

Much appreciate your interest, and hope to hear more from you.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Terry Pinnell]

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry,

Some more observations. I was wrong about the leakage current. Simulated 5ma
of leakage in Q6, which had no meaningful effect on output. The only
difference in swapping Q5 and Q6 that I can see is that the crossover
distortion depends on the Vbe of Q3, Q5, and Q4, and is independent of the
Vbe of Q6. Also, see last point.

There is too little loop gain to run a closed loop gain of 100. It looks
better with a loop gain of 10. Since 470K is pushing things, I changed R3 to
47K and R2 to 5.2K. Output clips at 22 V p-p, which is 7.5W.

If you want 20 - 20000 Hz bandwidth, you need to increase the values of C1
and C3. Also, C4 starts to roll off the closed loop gain before 20 KHz. I
changed that to 1 PF.

The gain seems to depend on signal amplitude. The smaller I make the input
signal (like 1mV), the less the gain. This is more severe at the 100 x gain
setting. At 10 x gain, I do not see any crossover distortion at 1 mV input.

I have not figured out completely the action of bootstrap capacitor, C2, but
it does do its job.

My previous comment about using 2N2219s to drive the output transistors is
moot, because the Q5 and Q6 collector current is only a little over an amp.
In fact, you would get some output, even if Q5 had an open collector.
Because of R9, Q4 could not drive the speaker directly as well as Q3 can.
There might be something here.

I did print your schematic with the 5/22 date.
Thanks a lot- that's really helpful.

I've just finished getting the circuit back onto stripboard and it
looks fine. About 7.3 V rms into 8 ohms.
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay23.gif

Now I'll do some tweaking. In particular, so far I've not really seen
any effects of adjusting the quiescent current. I've replaced R7 with
a fixed 15R and a '100R' preset that measures 150R. So I have a range
of 15 to 115 ohms. The original Torrens circuit had R2 = 68R for a 24V
supply, but the variation was small for other supply voltages: 12V &
18V specified 82R, 30V & 36V were 47R.

Your notes will provide a very useful basis for my tweaks. Maybe I'll
go back to the breadboard to play with gain, bandwidth, clipping
level, etc.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Tam/WB2TT]

"Terry Pinnell" <terrypinDELETE@THESEdial.pipex.com> wrote in message
news:s3n1b0l65gae9oag2hgjfrgcc7tatfaek2@4ax.com...
...................................

Now I'll do some tweaking. In particular, so far I've not really seen
any effects of adjusting the quiescent current. I've replaced R7 with
a fixed 15R and a '100R' preset that measures 150R. So I have a range
of 15 to 115 ohms. The original Torrens circuit had R2 = 68R for a 24V
supply, but the variation was small for other supply voltages: 12V &
18V specified 82R, 30V & 36V were 47R.

Your notes will provide a very useful basis for my tweaks. Maybe I'll
go back to the breadboard to play with gain, bandwidth, clipping
level, etc.

--
Terry Pinnell
Hobbyist, West Sussex, UK

Terry, I did not notice that it hurts anything, but powering Q1 off the

output signal seems odd. At the negative peaks, there is very little Vce on
Q1. I tried connecting it to +24V, but saw no difference. Still, just before
clipping, the collector-base voltage is .5V.

If you don't have it, you might want to download the SWCAD program from the
Linear Technologies web site. It is free, and includes schematic entry and
generic parts in addition to the LT parts.

Good luck!

Tam

 

[Rich Grise]

"Spehro Pefhany" <speffSNIP@interlogDOTyou.knowwhat> wrote in message

But how do you get stuff to drop out of a DC-10?

Put it in a wheel well?

Cheers!
Rich

 

[Terry Pinnell]

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry, I did not notice that it hurts anything, but powering Q1 off the
output signal seems odd. At the negative peaks, there is very little Vce on
Q1. I tried connecting it to +24V, but saw no difference. Still, just before
clipping, the collector-base voltage is .5V.

If you don't have it, you might want to download the SWCAD program from the
Linear Technologies web site. It is free, and includes schematic entry and
generic parts in addition to the LT parts.

Good luck!

Tam

Interesting point. No one's mentioned it before. Yes, I get same
result (no visible difference) if I take Q1C direct to 24V.

I have SWCADIII and could run anything you send or post to compare
with my CM sims (Spice3f5/XSpice). But from experience so far, I would
expect it to give similar results. What BJTs have you been using for
your sims?

One thing you may be able to comment on please. The actual circuit
results I posted previously have no cross-over distortion, to my
untrained eye at least.
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay23.gif
(And waveforms look even cleaner on a regular, analog 'scope.)

Yet the sim does, even on the smaller scale of this shot:
http://www.terrypin.dial.pipex.com/Images/PushPullR2=72kWave.gif

Finally, what would be your practical approach to setting the
'optimum' value of R7, and how would you most conveniently measure the
quiescent current that resistor is meant to optimise?
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay22-1.gif
BTW, on the stripboard circuit, finished yesterday, I replaced that
fixed 47R by a nominal 100R preset plus fixed 15R.


--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Stefan Heinzmann]

Terry Pinnell wrote:

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:


Terry, I did not notice that it hurts anything, but powering Q1 off the
output signal seems odd. At the negative peaks, there is very little Vce on
Q1. I tried connecting it to +24V, but saw no difference. Still, just before
clipping, the collector-base voltage is .5V.

If you don't have it, you might want to download the SWCAD program from the
Linear Technologies web site. It is free, and includes schematic entry and
generic parts in addition to the LT parts.

Good luck!

Tam


Interesting point. No one's mentioned it before. Yes, I get same
result (no visible difference) if I take Q1C direct to 24V.

Any difference you might see would be associated with the effect of the
collector-base capacitance of Q1. The way it is wired at the moment,
this capacitance helps C4 in its task of frequency compensation.

[...]

One thing you may be able to comment on please. The actual circuit
results I posted previously have no cross-over distortion, to my
untrained eye at least.
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay23.gif
(And waveforms look even cleaner on a regular, analog 'scope.)

Yet the sim does, even on the smaller scale of this shot:
http://www.terrypin.dial.pipex.com/Images/PushPullR2=72kWave.gif

This has to do with proper biasing. It would help if you could determine
(in the actual circuit and/or the simulation) the quiescent current
through Q5/Q6.

Finally, what would be your practical approach to setting the
'optimum' value of R7, and how would you most conveniently measure the
quiescent current that resistor is meant to optimise?
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay22-1.gif
BTW, on the stripboard circuit, finished yesterday, I replaced that
fixed 47R by a nominal 100R preset plus fixed 15R.

Short the input to ground (or turn down R0). Measure the total current
consumption of the amp with R7 set to minimum resistance. You can also
measure the voltage drop across R10 and R11. Then increase R7 and watch
the current consumption rise. At some point the rise should become much
steeper, indicating that Q5 and Q6 open up simultaneously. This is where
you leave class B operation and enter the class AB mode. There is
nothing inherently wrong with this except that the quiescent power
dissipation increases, which reduces the efficiency of the amp. This
knee point where power consumption starts to rise significantly is where
you probably want operate your amp, and it marks the best setting for R7.

Another problem is how to stabilize this against variations of
temperature, load, supply voltage, etc.

--
Cheers
Stefan

 

[Tam/WB2TT]

Terry,
I will try to tack the SWCAD file to the end of this.

Tam
-----------------------------
Version 4

SHEET 1 940 680

WIRE 448 192 448 176

WIRE 448 272 448 288

WIRE 656 304 656 448

WIRE 656 528 656 576

WIRE 656 576 448 576

WIRE -208 576 -208 512

WIRE -48 512 -48 576

WIRE -48 576 -208 576

WIRE 272 512 272 576

WIRE 272 576 80 576

WIRE 448 560 448 576

WIRE 448 576 272 576

WIRE 448 480 448 464

WIRE 592 304 608 304

WIRE 528 304 448 304

WIRE 448 304 448 368

WIRE 448 80 448 -144

WIRE 448 -144 272 -144

WIRE 128 -128 128 -144

WIRE 128 -144 -496 -144

WIRE 128 -32 128 -48

WIRE 272 16 272 -144

WIRE 272 -144 128 -144

WIRE 272 112 272 128

WIRE 272 224 272 240

WIRE 272 384 272 416

WIRE 880 192 880 -144

WIRE 880 -144 448 -144

WIRE 880 272 880 576

WIRE 656 576 880 576

WIRE 880 576 880 608

WIRE 128 48 128 64

WIRE 128 64 208 64

WIRE 128 64 128 80

WIRE 128 144 128 176

WIRE 128 240 128 256

WIRE 128 256 80 256

WIRE 80 256 80 272

WIRE 80 352 80 368

WIRE 80 464 80 576

WIRE 80 576 -48 576

WIRE -80 320 -80 336

WIRE -80 336 -48 336

WIRE -48 336 -48 416

WIRE 16 416 -48 416

WIRE -48 416 -48 432

WIRE -80 224 -80 144

WIRE -160 144 -208 144

WIRE -208 432 -208 384

WIRE -208 144 -224 144

WIRE -208 272 -144 272

WIRE -208 272 -208 144

WIRE -288 144 -320 144

WIRE 16 16 16 -32

WIRE 16 -32 128 -32

WIRE 16 80 16 144

WIRE 16 144 -80 144

WIRE 272 256 352 256

WIRE 272 256 272 288

WIRE 352 256 352 288

WIRE 352 288 448 288

WIRE 448 288 448 304

WIRE 208 336 176 336

WIRE 176 336 176 368

WIRE 176 368 80 368

WIRE 272 240 192 240

WIRE 272 240 272 256

WIRE 192 240 192 160

WIRE 192 160 16 160

WIRE 16 160 16 144

WIRE 384 128 272 128

WIRE 272 128 272 144

WIRE -512 400 -512 576

WIRE -512 576 -208 576

WIRE -512 320 -512 144

WIRE -512 144 -400 144

WIRE 608 304 656 304

WIRE 384 416 272 416

WIRE 272 416 272 432

WIRE -160 384 -208 384

WIRE -208 384 -208 272

WIRE -96 384 0 384

WIRE 0 384 0 368

WIRE 0 368 80 368

FLAG 880 608 0

FLAG 608 304 OUT

SYMBOL npn -144 224 R0

SYMATTR InstName Q1

SYMATTR Value 2N2219A

SYMBOL npn 16 368 R0

SYMATTR InstName Q2

SYMATTR Value 2N2219A

SYMBOL npn 208 16 R0

SYMATTR InstName Q4

SYMATTR Value 2N2219A

SYMBOL npn 384 80 R0

SYMATTR InstName Q5

SYMATTR Value 2N3055

SYMBOL npn 384 368 R0

SYMATTR InstName Q6

SYMATTR Value 2N3055

SYMBOL pnp 208 384 M180

SYMATTR InstName Q3

SYMATTR Value 2N2905A

SYMBOL res -64 416 R0

SYMATTR InstName R1

SYMATTR Value 10k

SYMBOL res -224 416 R0

SYMATTR InstName R2

SYMATTR Value 5.3k

SYMBOL res 256 416 R0

SYMATTR InstName R3

SYMATTR Value 100

SYMBOL res 432 464 R0

SYMATTR InstName R4

SYMATTR Value .2

SYMBOL res 432 176 R0

SYMATTR InstName R5

SYMATTR Value .2

SYMBOL res 256 128 R0

SYMATTR InstName R6

SYMATTR Value 100

SYMBOL res 64 256 R0

SYMATTR InstName R7

SYMATTR Value 68

SYMBOL res 112 -48 R0

SYMATTR InstName R8

SYMATTR Value 1.5k

SYMBOL res 112 -144 R0

SYMATTR InstName R9

SYMATTR Value 1.5k

SYMBOL res -64 128 R90

WINDOW 0 0 56 VBottom 0

WINDOW 3 32 56 VTop 0

SYMATTR InstName R10

SYMATTR Value 47k

SYMBOL res -304 128 R90

WINDOW 0 0 56 VBottom 0

WINDOW 3 32 56 VTop 0

SYMATTR InstName R11

SYMATTR Value 4.7k

SYMBOL cap -224 128 R90

WINDOW 0 0 32 VBottom 0

WINDOW 3 32 32 VTop 0

SYMATTR InstName C1

SYMATTR Value 10e-6

SYMBOL res 640 432 R0

SYMATTR InstName R12

SYMATTR Value 8

SYMBOL polcap 528 288 M90

WINDOW 0 0 32 VBottom 0

WINDOW 3 32 32 VTop 0

SYMATTR InstName C2

SYMATTR Value 3300ľf

SYMATTR Description Capacitor

SYMATTR Type cap

SYMATTR SpiceLine V=35 Irms=1.98 Rser=0.029 MTBF=5000 Lser=0 ppPkg=1

SYMBOL diode 112 80 R0

WINDOW 3 -34 23 Left 0

SYMATTR InstName D1

SYMBOL diode 112 176 R0

WINDOW 0 -39 33 Left 0

SYMATTR InstName D2

SYMBOL voltage 880 176 R0

WINDOW 123 0 0 Left 0

WINDOW 39 0 0 Left 0

SYMATTR InstName V1

SYMATTR Value 24

SYMBOL polcap 0 16 R0

WINDOW 3 24 64 Left 0

SYMATTR Value 10ľ

SYMATTR InstName C3

SYMATTR Description Capacitor

SYMATTR Type cap

SYMATTR SpiceLine V=16 Irms=35m Rser=2.8 MTBF=1000 Lser=0 mfg="Nichicon"
pn="UPR1C100MAH" type="Al electrolytic" ppPkg=1

SYMBOL voltage -512 304 R0

WINDOW 123 24 132 Left 0

WINDOW 39 0 0 Left 0

SYMATTR InstName V2

SYMATTR Value SINE(0 1.1 1000 0 0 0 100)

SYMBOL cap -160 400 R270

WINDOW 0 32 32 VTop 0

WINDOW 3 0 32 VBottom 0

SYMATTR InstName C4

SYMATTR Value 1.0e-12

TEXT -546 632 Left 0 !.tran 0 .1 0 1e-6

 

[Terry Pinnell]

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:

Terry Pinnell wrote:

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:


Terry, I did not notice that it hurts anything, but powering Q1 off the
output signal seems odd. At the negative peaks, there is very little Vce on
Q1. I tried connecting it to +24V, but saw no difference. Still, just before
clipping, the collector-base voltage is .5V.

If you don't have it, you might want to download the SWCAD program from the
Linear Technologies web site. It is free, and includes schematic entry and
generic parts in addition to the LT parts.

Good luck!

Tam


Interesting point. No one's mentioned it before. Yes, I get same
result (no visible difference) if I take Q1C direct to 24V.

Any difference you might see would be associated with the effect of the
collector-base capacitance of Q1. The way it is wired at the moment,
this capacitance helps C4 in its task of frequency compensation.

Thanks. Maybe that would have come to light if/when I got around to
testing with a range of frequencies.

One thing you may be able to comment on please. The actual circuit
results I posted previously have no cross-over distortion, to my
untrained eye at least.
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay23.gif
(And waveforms look even cleaner on a regular, analog 'scope.)

Yet the sim does, even on the smaller scale of this shot:
http://www.terrypin.dial.pipex.com/Images/PushPullR2=72kWave.gif

This has to do with proper biasing. It would help if you could determine
(in the actual circuit and/or the simulation) the quiescent current
through Q5/Q6.

Oh - maybe this is a source of my confusion. I'd been looking at
quiescent current elsewhere, not through the 2N3055s. It was hard to
read from the rather poor scans of the original I posted on 14th May
http://www.terrypin.dial.pipex.com/Images/AudioPowerAmpOriginal1.jpg
http://www.terrypin.dial.pipex.com/Images/AudioPowerAmpOriginalText.gif
but here's what Torrens says:
"For correct minimum distortion, a small current should flow through
Tr3, R8 and Tr4, R9. This is why D1, D2, and R7 are included: to set
up a small voltage difference between Tr3 and Tr4 bases. If this
voltage difference is too small Tr3 and Tr4 will both be off and
crossover distortion occurs. If too high the current through R8 and R9
develops voltage high enough to turn the output transistor on: they
will both conduct and quickly overheat."

Note that he doesn't venture any figures...

Those labels all match my circuit
http://www.terrypin.dial.pipex.com/Images/PushPull-May22-1.gif
except that I use 'Q' not 'Tr' for transistors. (And of course I've
implemented a few strongly recommended small changes.)

So, could you clarify whether there is an issue about *where* to
measure this quiescent current please?

BUT - your post prompted me to return to my simulation of the
quiescent circuit. And I'm pleased to say I found and corrected a
silly mistake that now eliminates the cross-over distortion from the
simulation! I hadn't bothered to specify a particular type of diode
for D1 and D2. CircuitMaker had therefore chosen the first on its
alphabetic list (not the DEFAULT diode, which I'd assumed, and often
use when not fussy). And that entry happened to be a 100A power diode!
('100HF100PV 1000V 100000mA 3us Si Diode pkgIODE0.7 A,K')

I specified a 1N4148, which I believe would have been the type I used
back in 1981. That removed the cross-over distortion.

I then proceeded to try and answer your question, and tabulated some
values:

R7: 33 47 82 100 120 150 180 200
Ic (Q5/Q6): 0.1uA 0.5uA 29uA 230uA 2.2mA 35mA 140mA 230mA

Note that these values of R7 go way above those apparently used by
Torrens. He specified 82 for a 12V supply, ranging up to 47 for 30 or
36V. (With 68R for my 24V case; I'm using 47.)

I also thought I'd see what effect the D1/D2 type had, leaving aside
power diodes.
R7----------------
Type 47 120 150
------ ----- ----- ------
DEFAULT 5nA 36nA 850nA
1N4148 500nA 2.2mA 35mA
1N914 5nA 13nA 220nA
1N4001 5nA 7nA 54nA

I'm very surprised that what I'd always thought of as a 'typical'
small signal Si diode, the 1N914, gives virtually no quiescent current
at all, even at the 'high' values I selected.

BTW, Torrens doesn't specify a type. Just 'DUS', i.e. Diode Universal
Silicon (the style I recall used in all Elektor magazine circuits in
that era.)

So ... a few more puzzles there <g>.

Finally, what would be your practical approach to setting the
'optimum' value of R7, and how would you most conveniently measure the
quiescent current that resistor is meant to optimise?
http://www.terrypin.dial.pipex.com/Images/PushPullOutMay22-1.gif
BTW, on the stripboard circuit, finished yesterday, I replaced that
fixed 47R by a nominal 100R preset plus fixed 15R.

Short the input to ground (or turn down R0). Measure the total current
consumption of the amp with R7 set to minimum resistance. You can also
measure the voltage drop across R10 and R11. Then increase R7 and watch
the current consumption rise. At some point the rise should become much
steeper, indicating that Q5 and Q6 open up simultaneously. This is where
you leave class B operation and enter the class AB mode. There is
nothing inherently wrong with this except that the quiescent power
dissipation increases, which reduces the efficiency of the amp. This
knee point where power consumption starts to rise significantly is where
you probably want operate your amp, and it marks the best setting for R7.

Thanks a lot, that will be very helpful. But we need to resolve this
apparent issue of *where* to measure first.

Another problem is how to stabilize this against variations of
temperature, load, supply voltage, etc.

Yes - I've postponed thinking much about those for now <g>.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Terry Pinnell]

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry,
I will try to tack the SWCAD file to the end of this.
Thanks Tam. Simulates fine here. Haven't done more than load and run

it once. (I got 100mS worth by default - was that intended?)

BTW, when I first loaded the ASC file (which I'd made by pasting your
text), I got this odd error message:
pn="UPR1C100MAH" type="Al electrolytic" ppPkg=1

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Stefan Heinzmann]

Terry Pinnell wrote:

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:
[...]
This has to do with proper biasing. It would help if you could determine
(in the actual circuit and/or the simulation) the quiescent current
through Q5/Q6.


Oh - maybe this is a source of my confusion. I'd been looking at
quiescent current elsewhere, not through the 2N3055s. It was hard to
read from the rather poor scans of the original I posted on 14th May
http://www.terrypin.dial.pipex.com/Images/AudioPowerAmpOriginal1.jpg
http://www.terrypin.dial.pipex.com/Images/AudioPowerAmpOriginalText.gif
but here's what Torrens says:
"For correct minimum distortion, a small current should flow through
Tr3, R8 and Tr4, R9. This is why D1, D2, and R7 are included: to set
up a small voltage difference between Tr3 and Tr4 bases. If this
voltage difference is too small Tr3 and Tr4 will both be off and
crossover distortion occurs. If too high the current through R8 and R9
develops voltage high enough to turn the output transistor on: they
will both conduct and quickly overheat."

Note that he doesn't venture any figures...

Those labels all match my circuit
http://www.terrypin.dial.pipex.com/Images/PushPull-May22-1.gif
except that I use 'Q' not 'Tr' for transistors. (And of course I've
implemented a few strongly recommended small changes.)

So, could you clarify whether there is an issue about *where* to
measure this quiescent current please?

It is not that critical *where* you measure it. It rather depends on
what is convenient. In a simulation, you can of couse measure the
collector current of the power transistors directly. In a real circuit
it may be more convenient to measure the voltage drop across R10 and/or
R11. You may even measure the current consumption of the entire
amplifier, because the knee in the curve will also show.

BUT - your post prompted me to return to my simulation of the
quiescent circuit. And I'm pleased to say I found and corrected a
silly mistake that now eliminates the cross-over distortion from the
simulation! I hadn't bothered to specify a particular type of diode
for D1 and D2. CircuitMaker had therefore chosen the first on its
alphabetic list (not the DEFAULT diode, which I'd assumed, and often
use when not fussy). And that entry happened to be a 100A power diode!
('100HF100PV 1000V 100000mA 3us Si Diode pkgIODE0.7 A,K')

I specified a 1N4148, which I believe would have been the type I used
back in 1981. That removed the cross-over distortion.

I then proceeded to try and answer your question, and tabulated some
values:

R7: 33 47 82 100 120 150 180 200
Ic (Q5/Q6): 0.1uA 0.5uA 29uA 230uA 2.2mA 35mA 140mA 230mA

If you sketch that as a graph on a piece of paper, you'll see what knee
I'm talking about. A good bias setting would be somewhere around 120 Ohm.

Note that these values of R7 go way above those apparently used by
Torrens. He specified 82 for a 12V supply, ranging up to 47 for 30 or
36V. (With 68R for my 24V case; I'm using 47.)

Various factors have an influence on this, such as the transistor type,
the diode type, the temperature, etc.

Note also the the addition of R10 and R11 makes the bias setting more
stable. That is one of the more important reasons why you want them. If
the quiescent current is too high for some reason, it will develop a
voltage across those resistors that effectively reduce the bias voltage
available for the output transistors, thereby counteracting the current
increase.

I also thought I'd see what effect the D1/D2 type had, leaving aside
power diodes.
R7----------------
Type 47 120 150
------ ----- ----- ------
DEFAULT 5nA 36nA 850nA
1N4148 500nA 2.2mA 35mA
1N914 5nA 13nA 220nA
1N4001 5nA 7nA 54nA

I'm very surprised that what I'd always thought of as a 'typical'
small signal Si diode, the 1N914, gives virtually no quiescent current
at all, even at the 'high' values I selected.

BTW, Torrens doesn't specify a type. Just 'DUS', i.e. Diode Universal
Silicon (the style I recall used in all Elektor magazine circuits in
that era.)

So ... a few more puzzles there <g>.

Not that surprising really. Not all diodes and transistors are created
equal. The differences may add up to disturb the biasing enough for
creating noticeable distortion, as you've experienced.

--
Cheers
Stefan

 

[Winfield Hill]

Stefan Heinzmann wrote...

I also was reminded that this were the times when even a decent NPN
transistor was difficult to make. Note that the one NPN transistor
in the circuit was an experimental device.

That's because they all were Germanium transistors. They could
make NPN types, but they weren't very powerful, hence the need
for the complementary Darlington (Sziklai) connection.

Thanks,
- Win

(email: use hill_at_rowland-dot-org for now)

 

[John Popelish]

Terry Pinnell wrote:
(snip)

I also thought I'd see what effect the D1/D2 type had, leaving aside
power diodes.
R7----------------
Type 47 120 150
------ ----- ----- ------
DEFAULT 5nA 36nA 850nA
1N4148 500nA 2.2mA 35mA
1N914 5nA 13nA 220nA
1N4001 5nA 7nA 54nA

I'm very surprised that what I'd always thought of as a 'typical'
small signal Si diode, the 1N914, gives virtually no quiescent current
at all, even at the 'high' values I selected.
(snip)

I suspect your model for the 1N914 is faulty.

By the way, your bias adjust circuit has a suicide failure mode. If
the sliding contact becomes the least bit intermittent, the bias
current pops up destructively. The solution to this problem is to use
a transistor as a diode multiplier, with a fixed resistor collector to
base and an adjustable resistor base to emitter, so that the
adjustable element is what holds the bias voltage up. That way, if
the resistor opens, the bias voltage collapses to a 1 diode drop.

I have seen lots of Vbe multipliers with the sliding contact connected
to the base, and the ends connected to the collector and emitter (with
fixed resistors in series) but this configuration has the same fatal
failure mode as your diode resistor circuit does.
--
John Popelish

 

[JeffM]

Nooooo, NOT polyethelene. Something else.
"polyester materials based on tere- and iso- phthalates. The spectra
also suggested the possible presence of inorganic silicate
compounds."
Spehro Pefhany

Speff,
Were you a chemist in a previous life?
http://www.univarcanada.com/Personnel/Display?readform&group=Compounding&industry=Chemical+Compounding

 

[Tam/WB2TT]

Terry, I think I solved your original question. Run SWCAD with the collector
of Q5 tied to the base, instead of to +24. Now put that back, but open up
the emitter of Q6, and tie that to the base. One will work, the other will
not. I bet the 2N2905 was running hot!

Tam
"Terry Pinnell" <terrypinDELETE@THESEdial.pipex.com> wrote in message
news:7ek4b0ld01uihd6953nbva1215i3pvuvhq@4ax.com...

"Tam/WB2TT" <t-tammaru@c0mca$t.net> wrote:

Terry,
I will try to tack the SWCAD file to the end of this.
Thanks Tam. Simulates fine here. Haven't done more than load and run
it once. (I got 100mS worth by default - was that intended?)

BTW, when I first loaded the ASC file (which I'd made by pasting your
text), I got this odd error message:
pn="UPR1C100MAH" type="Al electrolytic" ppPkg=1

--
Terry Pinnell
Hobbyist, West Sussex, UK