Are these 2N3055s borderline spec?

[Terry Pinnell]

This is a side issue arising from my ongoing attempts to get my
push-pull audio power amp working properly, as discussed currently in
separate thread here and in alt.binaries.schematics.electronic.

I dismantled the stripboard and set up the circuit again on
breadboard, allowing me easier access for testing and experiment. That
was successful. Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

That good result was using two 2N3055s on the breadboard itself. So I
then used short croc clips to wire in the existing case-mounted
2N3055s instead. That gave me distortion again. I tried a variety of
different 2N3055s from my small and ancient supply. As a result, and
to cut to the chase, I'm coming to think that a major part of the
problems I've been having (earlier and now) may be due to these power
NPNs being on the borderline of usability.

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are other
parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

As just one example of several inconsistencies, if I swapped certain
pairs of 2N3055s between the Q5 and Q6 positions, I got wildly
different results. For instance, with Q5 = #7 and Q6 = #4, I got a
reasonable output, albeit with a little crossover distortion. But if I
swapped those, so that Q5 = #4 and Q6 = #7, the result was badly
distorted. Here are the respective waveforms:

http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardPuzzle.gif

The power section of the circuit looks pretty symmetrical to me. So
why should this non-symmetrical behaviour occur please?

More important, are my old 2N3055s still OK for this application? Or
showing signs of their age and their likely origin in a 'surplus' bulk
buy?

[Also posted with attachments in alt.binaries.schematics.electronic.]

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Winfield Hill]

Terry Pinnell wrote...

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are
other parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

At what current did you measure the gain? And with what Vce?

The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

Perhaps someone has mentioned this, but Q4 is vunerable to damage
in this circuit. That's because if a negative signal swing fails
to deliver as much current to the load as the feedback circuit asks,
it can create excessively high currents in Q4, attempting to get Q6
to respond. Generally this will happen only under high negative Vout
conditions, which means the voltage across Q4 will be low, but it's
an issue nonetheless. If during prototype debugging the amplifier
fails to perform as expected, it'd be wise to check Q4's condition.

Thanks,
- Win

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

 

[Terry Pinnell]

Winfield Hill <Winfield_member@newsguy.com> wrote:

Terry Pinnell wrote...

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are
other parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

At what current did you measure the gain? And with what Vce?

Not sure, but I recognise it was probably untypically low. I took
three measurements, two with separate DMMs (which gave very close
results), and the third with my own tester, which was also designed
for 'general purpose' low-level transistors.

The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

BTW, I'd not drawn in C4 (3.3pF at present; will change to 68pF), but
that's now added to the schematic.

Perhaps someone has mentioned this, but Q4 is vunerable to damage
in this circuit. That's because if a negative signal swing fails
to deliver as much current to the load as the feedback circuit asks,
it can create excessively high currents in Q4, attempting to get Q6
to respond. Generally this will happen only under high negative Vout
conditions, which means the voltage across Q4 will be low, but it's
an issue nonetheless. If during prototype debugging the amplifier
fails to perform as expected, it'd be wise to check Q4's condition.

Q4 seems OK here. Every time I get a poor result I go through the
chore of changing all four low level BJTs.

So, is there any reason you can suggest to explain why merely swapping
Q5/Q6 should cause failure please?

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[John Larkin]

On Thu, 20 May 2004 14:22:43 +0100, Terry Pinnell
<terrypinDELETE@THISdial.pipex.com> wrote:

This is a side issue arising from my ongoing attempts to get my
push-pull audio power amp working properly, as discussed currently in
separate thread here and in alt.binaries.schematics.electronic.

I dismantled the stripboard and set up the circuit again on
breadboard, allowing me easier access for testing and experiment. That
was successful. Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

That good result was using two 2N3055s on the breadboard itself. So I
then used short croc clips to wire in the existing case-mounted
2N3055s instead. That gave me distortion again. I tried a variety of
different 2N3055s from my small and ancient supply. As a result, and
to cut to the chase, I'm coming to think that a major part of the
problems I've been having (earlier and now) may be due to these power
NPNs being on the borderline of usability.

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are other
parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

As just one example of several inconsistencies, if I swapped certain
pairs of 2N3055s between the Q5 and Q6 positions, I got wildly
different results. For instance, with Q5 = #7 and Q6 = #4, I got a
reasonable output, albeit with a little crossover distortion. But if I
swapped those, so that Q5 = #4 and Q6 = #7, the result was badly
distorted. Here are the respective waveforms:

http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardPuzzle.gif

The power section of the circuit looks pretty symmetrical to me. So
why should this non-symmetrical behaviour occur please?

More important, are my old 2N3055s still OK for this application? Or
showing signs of their age and their likely origin in a 'surplus' bulk
buy?

[Also posted with attachments in alt.binaries.schematics.electronic.]

3055's have a beta peak at fairly high current, so a regular DVM-type
"beta meter" would show pretty small numbers.

But a proper circuit design should be indifferent to beta over a huge
range.

There's nothing really wrong with "old" silicon transistors. In fact,
many of the older, big-die diffused parts are a lot more rugged than
small-chip epitxial parts with the same 2N number.

John

 

[Stefan Heinzmann]

Terry Pinnell wrote:

This is a side issue arising from my ongoing attempts to get my
push-pull audio power amp working properly, as discussed currently in
separate thread here and in alt.binaries.schematics.electronic.

I dismantled the stripboard and set up the circuit again on
breadboard, allowing me easier access for testing and experiment. That
was successful. Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

That good result was using two 2N3055s on the breadboard itself. So I
then used short croc clips to wire in the existing case-mounted
2N3055s instead. That gave me distortion again. I tried a variety of
different 2N3055s from my small and ancient supply. As a result, and
to cut to the chase, I'm coming to think that a major part of the
problems I've been having (earlier and now) may be due to these power
NPNs being on the borderline of usability.

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are other
parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

Vceo is not the reverse voltage. It is the maximum forward voltage
between collector and emitter with the base left open. So could you
please provide some more detail about your measurement setup?

The gain is only meaningful if you provide information about the
measurement conditions. At the very least, you should provide the
collector current and the collector-emitter voltage. HFe also depends on
the temperature, but that is harder to measure.

Furthermore, you can determine for yourself whether your transistors are
within specification by looking at the data sheet. The 2N3055 is
available from a number of sources, a good start would be On
Semiconductor. Just have a look at their website www.onsemi.com

The 2N3055 is a very old transistor. The manufacturing processes which
are being used nowadays are quite different from those when the
transistor was new. You can expect some differences in behaviour when
comparing old models with newer vintages, although both should fulfill
the same specs. For example, it is possible that newer devices exhibit
appreciable gain at frequencies where older models would have given up.
So if the stability of your circuit depends on the limitations of
earlier transistor devices, you may find that newer devices of nominally
the same type may make the circuit unstable.

As just one example of several inconsistencies, if I swapped certain
pairs of 2N3055s between the Q5 and Q6 positions, I got wildly
different results. For instance, with Q5 = #7 and Q6 = #4, I got a
reasonable output, albeit with a little crossover distortion. But if I
swapped those, so that Q5 = #4 and Q6 = #7, the result was badly
distorted. Here are the respective waveforms:

http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardPuzzle.gif

The power section of the circuit looks pretty symmetrical to me. So
why should this non-symmetrical behaviour occur please?

More important, are my old 2N3055s still OK for this application? Or
showing signs of their age and their likely origin in a 'surplus' bulk
buy?

Given the vintage of the circuit I would guess that old devices may work
better than new ones. Do your devices have a manufacturer logo and/or a
manufacturing date?

--
Cheers
Stefan

 

[Terry Pinnell]

Winfield Hill <Winfield_member@newsguy.com> wrote:

At what current did you measure the gain? And with what Vce?

Forgot to answer the Vce point. Not quite sure I used the right term,
or whether I've understood your question. I measured the reverse
breakdown voltage with my homebrew tester. During press of push
button, that applies around 340 V DC across the CE junction (in this
case of NPN), limited by 3 x 100k resistors, while I read voltage on
my DMM.

I was surprised at the wide range.

Does that measurement give a reasonable approximation to what I see
called 'VCE Max' or 'Vceo' in my ref books? For 2N3055 both my sources
have that as 60 V. My sample of 8 ranged from a miserable 17 to an
extraordinary 297!

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Kevin Carney]

2N3055 came from many vendors. Polypack used to sell them by the barrell
full. I was always suspect of a 2n3055 unless it was in a quality piece of
gear. Tektronix and HP used it with thier own number assigned. When we
needed 2N3055's I would order them from Tek or HP (using thier number)
because they were at least tested or met thier spec.

--
change .combo to .com for correct email

***************************************************
"We ought always to know precisely why a given job
is done in a particular way, and why it is done at
all, and why it can't be done more efficiently,
if it must be done at all."-- T.J.Watson

***************************************************

"Terry Pinnell" <terrypinDELETE@THISdial.pipex.com> wrote in message
news:88jpa0lr6qkbssnj9ggij8bfifstqd62oj@4ax.com...

Winfield Hill <Winfield_member@newsguy.com> wrote:

At what current did you measure the gain? And with what Vce?

Forgot to answer the Vce point. Not quite sure I used the right term,
or whether I've understood your question. I measured the reverse
breakdown voltage with my homebrew tester. During press of push
button, that applies around 340 V DC across the CE junction (in this
case of NPN), limited by 3 x 100k resistors, while I read voltage on
my DMM.

I was surprised at the wide range.

Does that measurement give a reasonable approximation to what I see
called 'VCE Max' or 'Vceo' in my ref books? For 2N3055 both my sources
have that as 60 V. My sample of 8 ranged from a miserable 17 to an
extraordinary 297!

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[John Popelish]

Terry Pinnell wrote:
(snip)

Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

Off on a tangent:
I have been reading about your amplifier with great interest, since I
have been simmering several amplifier design ideas for many years.
(Since I was a teen, I always wanted to design the perfect
amplifier.)

I want to talk about the right position for R11. All the discussion I
saw about it maintained its effect or lack of effect directly on Q6.
But I think this has missed the point of having it. Q4 and Q6 form a
unity gain amplifier that is actually an amplifier with lots of
voltage gain but 100% feedback to reduce that to 1. So from the point
of view of the base voltage on Q4, Q6 just adds current gain to Q4's
emitter follower, without doing anything to the emitter voltage. The
important function of R11 should be to provide an output current
feedback to the operation of the emitter follower, Q4, to hold its DC
bias point more stable. The high feedback around Q6 will make it do
what ever is necessary to cooperate in this.

Thinking this way, R11 does not belong in either the emitter or
collector of Q6, but should be carrying all the collector current of
Q6 and the emitter current of Q4 to the load and feedback. In other
words, separate the emitter of Q4 from the junction of the feedback
path, the output, R8 and R10 and connect it directly to the emitter of
Q6. Then move R11 to connect this node (the emitter of Q4 and the
collector of Q6) to the output, feedback, R8 and R10. Then all the
output current from both Q4 and its current booster Q6 will drop
voltage across R11 in a way that will reduce the voltage across the
base emitter junctions of of Q3 and Q4. So if the output bias current
increases, this drop in effective base voltage will tend to lower it.

Ideally, R8 should also be connected directly across the base emitter
junction of Q5, so that the drop across R10 represents all the current
to the output from both Q3 and Q5.

Since the collector of Q4 is a current source feeding base current to
Q6, R9 should not have anything to do with the signal earth, but just
be a detour directly around the base emitter junction of Q6.

These changes should improve the thermal bias stabilization of the
output section and improve your oscillation and distortion problems.

--
John Popelish

 

[Rich Grise]

"Terry Pinnell" <terrypinDELETE@THISdial.pipex.com> wrote in message

More important, are my old 2N3055s still OK for this application? Or
showing signs of their age and their likely origin in a 'surplus' bulk
buy?

FWIW, for a little while back in the late 1970's I worked at a Rat
Shack in the service department. We'd routinely use 2N3055s to replace
output trannies. I'm not good enough to comment on whether yours are
getting enough drive, but that was my first gut feeling.

Just FYI, the complement to the 2N2219A is the 2N2905A.

Cheers!
Rich

 

[Stefan Heinzmann]

John Popelish schrieb:

Terry Pinnell wrote:
(snip)


Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif


Off on a tangent:
I have been reading about your amplifier with great interest, since I
have been simmering several amplifier design ideas for many years.
(Since I was a teen, I always wanted to design the perfect
amplifier.)

I want to talk about the right position for R11. All the discussion I
saw about it maintained its effect or lack of effect directly on Q6.
But I think this has missed the point of having it. Q4 and Q6 form a
unity gain amplifier that is actually an amplifier with lots of
voltage gain but 100% feedback to reduce that to 1. So from the point
of view of the base voltage on Q4, Q6 just adds current gain to Q4's
emitter follower, without doing anything to the emitter voltage. The
important function of R11 should be to provide an output current
feedback to the operation of the emitter follower, Q4, to hold its DC
bias point more stable. The high feedback around Q6 will make it do
what ever is necessary to cooperate in this.

Thinking this way, R11 does not belong in either the emitter or
collector of Q6, but should be carrying all the collector current of
Q6 and the emitter current of Q4 to the load and feedback. In other
words, separate the emitter of Q4 from the junction of the feedback
path, the output, R8 and R10 and connect it directly to the emitter of
Q6. Then move R11 to connect this node (the emitter of Q4 and the
collector of Q6) to the output, feedback, R8 and R10. Then all the
output current from both Q4 and its current booster Q6 will drop
voltage across R11 in a way that will reduce the voltage across the
base emitter junctions of of Q3 and Q4. So if the output bias current
increases, this drop in effective base voltage will tend to lower it.

Maybe my message was too brief, but this is precisely what I tried to
suggest several days ago. Another way of thinking about this is that Q4
and Q6 together form a sort of high-gain PNP transistor, with its
collector being the emitter of Q6, and the emitter being the emitter of
Q4, which is connected to the collector of Q6. If you combine Q3/Q5 into
a similar NPN transistor (it is a darlington pair anyhow) you have a
simple picture with a conventional push-pull output. From there it is
quite clear where R11 needs to go.

Ideally, R8 should also be connected directly across the base emitter
junction of Q5, so that the drop across R10 represents all the current
to the output from both Q3 and Q5.

Given the small contribution of the current through R8 do you think this
would matter?

Since the collector of Q4 is a current source feeding base current to
Q6, R9 should not have anything to do with the signal earth, but just
be a detour directly around the base emitter junction of Q6.

Agreed.

These changes should improve the thermal bias stabilization of the
output section and improve your oscillation and distortion problems.

Talking about thermal stability: Has anyone mentioned that it would be a
good idea to thermally couple D1 and D2 to the output transistors to
reduce bias voltage at higher temperatures?

--
Cheers
Stefan

 

[Rich Grise]

"Terry Pinnell" <terrypinDELETE@THISdial.pipex.com> wrote in message

So, is there any reason you can suggest to explain why merely swapping
Q5/Q6 should cause failure please?

Well, I wish I knew enough theory here, but I just have this nagging
feeling that gain is an issue. The 2N3055 data sheet shows a beta min.
20 at 4.0A, and 5 at 10A; and your table shows a lot of variation.

Win Hill beat me to asking what current you test your 3055s at, but
I wonder if a changing beta through a range of Ics would cause the
kind of distortion you're seeing? And is the loop gain low enough
that their beta would come into play so dramatically? (when you
swap pairs)

What would happen if you interposed another 2N2219 as an emitter-
follower at the collector of Q4? And at the emitter of Q3, for
that matter, but then they're Darlingtons, and that's a whole
nother discussion, I'd think.

Good Luck!
Rich

 

[legg]

was successful. Using a regulated 24V supply, I got a clean 6.8 V rms
output into an 8 ohm resistor. The present circuit is this:
http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardNow.gif

Here's some hard data. For each of my 2N3055s, I measured two
parameters: the gain, and the reverse voltage. (Maybe there are other
parameters important in this application?)

2N3055 Hfe Vceo
------ --- ----
#1 29 27 Original Q6
#2 24 62 Original Q5
#3 15 297
#4 30 126
#5 22 17
#6 13 151
#7 13 89
#8 23 21

Don't know how or what you are measuring here, but your schematic
provides differing base current drive capacity for the two drive
transistors.

Positive going output signals are current limited by the fixed voltage
present across R7 - the voltage across R10 cannot exceed this value.
Veb and temperature will be signifigant.

Output current is also beta-limited, under worst-case conditions, to
that current provided by the amplified source combination of C2/R5, or
approximately 3mA x hfeQ3 x hfeQ5.

Niether of these limitations apply to the Q4-Q6 complementary section,
which has no emitter resistor back-biasing Q4, nor resistively
limited base current from the collector of Q2.

RL

 

[John Popelish]

Stefan Heinzmann wrote:

John Popelish schrieb:

I want to talk about the right position for R11. All the discussion I
saw about it maintained its effect or lack of effect directly on Q6.
But I think this has missed the point of having it. Q4 and Q6 form a
unity gain amplifier that is actually an amplifier with lots of
voltage gain but 100% feedback to reduce that to 1. So from the point
of view of the base voltage on Q4, Q6 just adds current gain to Q4's
emitter follower, without doing anything to the emitter voltage. The
important function of R11 should be to provide an output current
feedback to the operation of the emitter follower, Q4, to hold its DC
bias point more stable. The high feedback around Q6 will make it do
what ever is necessary to cooperate in this.

Thinking this way, R11 does not belong in either the emitter or
collector of Q6, but should be carrying all the collector current of
Q6 and the emitter current of Q4 to the load and feedback. In other
words, separate the emitter of Q4 from the junction of the feedback
path, the output, R8 and R10 and connect it directly to the emitter of
Q6. Then move R11 to connect this node (the emitter of Q4 and the
collector of Q6) to the output, feedback, R8 and R10. Then all the
output current from both Q4 and its current booster Q6 will drop
voltage across R11 in a way that will reduce the voltage across the
base emitter junctions of of Q3 and Q4. So if the output bias current
increases, this drop in effective base voltage will tend to lower it.

Maybe my message was too brief, but this is precisely what I tried to
suggest several days ago.

I must have missed your contribution, entirely.

Another way of thinking about this is that Q4
and Q6 together form a sort of high-gain PNP transistor, with its
collector being the emitter of Q6, and the emitter being the emitter of
Q4, which is connected to the collector of Q6. If you combine Q3/Q5 into
a similar NPN transistor (it is a darlington pair anyhow) you have a
simple picture with a conventional push-pull output. From there it is
quite clear where R11 needs to go.

Ideally, R8 should also be connected directly across the base emitter
junction of Q5, so that the drop across R10 represents all the current
to the output from both Q3 and Q5.

Given the small contribution of the current through R8 do you think this
would matter?

Not much, except when the total current from the compound transistor
is made up almost entirely of the current through R9, which is close
to the case when quiescent.

Since the collector of Q4 is a current source feeding base current to
Q6, R9 should not have anything to do with the signal earth, but just
be a detour directly around the base emitter junction of Q6.

Agreed.

These changes should improve the thermal bias stabilization of the
output section and improve your oscillation and distortion problems.

Talking about thermal stability: Has anyone mentioned that it would be a
good idea to thermally couple D1 and D2 to the output transistors to
reduce bias voltage at higher temperatures?

In this configuration, the temperature of Q6 is not directly involved
in the total of 3 junction drop between the base of Q3 and the base of
Q4. But D1 and D2 should be D1, D2 and D3. I would mount one diodes
near Q5, one against Q3 and one against Q4, since all 3 of these base
emitter junction voltage drops are involved in the bias current
setting. However, since none of these locations actually see the full
temperature rise of any of those junctions, it might be better to
mount them all on the case of Q5 (which I would expect to have more
temperature rise than either Q3 or Q4) and risk slight over
compensation at high temperature.

--
John Popelish

 

[Terry Pinnell]

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:

Vceo is not the reverse voltage. It is the maximum forward voltage
between collector and emitter with the base left open. So could you
please provide some more detail about your measurement setup?

See my earlier reply to Win's similar query.

The gain is only meaningful if you provide information about the
measurement conditions. At the very least, you should provide the
collector current and the collector-emitter voltage. HFe also depends on
the temperature, but that is harder to measure.

Furthermore, you can determine for yourself whether your transistors are
within specification by looking at the data sheet. The 2N3055 is
available from a number of sources, a good start would be On
Semiconductor. Just have a look at their website www.onsemi.com

I have the spec, as I mentioned, from a couple of reference books. But
don't you think it's a tad unrealistic to suggest I can determine
whether a particular transistor fully meets that spec? That would need
an exhaustive set of measurements beyond my capabilities, even if the
effort was worthwhile. For the 2N3055, Towers International Transistor
Selector gives his data:
Vcb max = 100V, Vce max = 60V, Veb max = 7V, Ic max = 15A, Tj max =
200C, Ptot = 115WC, Ft min = 200k, Cob max not specified, Hfe = 20-70,
Hfe bias = 4A.
A.M Ball's Semiconductor data differs somewhat:
fT min = 700k, Vcbo max = 160V, Vebo max = 4V, Ic = 10A, Typical Hfe =
45 at 3A.

What I *can* do is determine if a transistor *fails* to meet a key
parameter. That's what I was trying to do with the reverse voltage
breakdown test. But if, as you say, that doesn't give an estimate of
Vceo (to compare with the above 60V), then how do you measure it?

The 2N3055 is a very old transistor. The manufacturing processes which
are being used nowadays are quite different from those when the
transistor was new. You can expect some differences in behaviour when
comparing old models with newer vintages, although both should fulfill
the same specs. For example, it is possible that newer devices exhibit
appreciable gain at frequencies where older models would have given up.
So if the stability of your circuit depends on the limitations of
earlier transistor devices, you may find that newer devices of nominally
the same type may make the circuit unstable.

As just one example of several inconsistencies, if I swapped certain
pairs of 2N3055s between the Q5 and Q6 positions, I got wildly
different results. For instance, with Q5 = #7 and Q6 = #4, I got a
reasonable output, albeit with a little crossover distortion. But if I
swapped those, so that Q5 = #4 and Q6 = #7, the result was badly
distorted. Here are the respective waveforms:

http://www.terrypin.dial.pipex.com/Images/PushPullBreadboardPuzzle.gif

The power section of the circuit looks pretty symmetrical to me. So
why should this non-symmetrical behaviour occur please?

More important, are my old 2N3055s still OK for this application? Or
showing signs of their age and their likely origin in a 'surplus' bulk
buy?

Given the vintage of the circuit I would guess that old devices may work
better than new ones. Do your devices have a manufacturer logo and/or a
manufacturing date?

Most are worn so faint that they are impossible to read reliably.

Thanks for the help. See also my separate post reporting good
progress, using original circuit.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Terry Pinnell]

legg <legg@nospam.magma.ca> wrote:

Don't know how or what you are measuring here, but your schematic
provides differing base current drive capacity for the two drive
transistors.

Positive going output signals are current limited by the fixed voltage
present across R7 - the voltage across R10 cannot exceed this value.
Veb and temperature will be signifigant.

Output current is also beta-limited, under worst-case conditions, to
that current provided by the amplified source combination of C2/R5, or
approximately 3mA x hfeQ3 x hfeQ5.

Niether of these limitations apply to the Q4-Q6 complementary section,
which has no emitter resistor back-biasing Q4, nor resistively
limited base current from the collector of Q2.

Thanks for all the replies so far. I will experiment with as many of
those design changes as time allows, once I have a robust performance
from the present circuit.

To summarise the current status (as also reported in the original
thread): I seem to have made progress this evening. It looks as if Q4
and one of the 2N3055s were *both* 'borderline'. Although Q4 was
testing OK on my transistor tester, and showing a reasonable gain, in
exasperation I swapped it for an alternative and immediately started
getting better results.

I now have a combination that delivers a clean-looking sine of about
6.9V rms into 8 ohms. That's on breadboard (with 6 croc-clip
connections to the case-mounted 2N3055s), and with a regulated 24.0 V
bench supply. It remains to be seen if I can retain that good
performance when I rebuild the stripboard circuit and try alternative
supplies.

I intend to replace the two presets for setting the output DC level
and quiescent current respectively. The first is obviously easy to set
(half supply voltage), but how do I determine optimum setting for
quiescent current please?

I may well also experiment with other improvements and design changes
later. But I want to get this working robustly first. Its author, R
Torrens, did so with a variety of supply voltages from 12 to 36 V. His
tests showed distortion was "...around 0.5% or better, which is
approaching hi-fi levels." But I wonder if he had to almost
'hand-pick' his transistors as I seem to be doing for Q4, Q5 and Q6?


--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Tam/WB2TT]

Terry,

One thing that sticks out is that the 2N2219 is too wimpy to drive a 2N3055.
Also, I believe you said the distortion changes as you interchanged the two
output transistors; did you readjust the current through D1 and D2? You
might want to re measure the breakdown voltages on the transistors with the
base and emitter tied together. Throw out any that don't meet spec.

Tam
"Terry Pinnell" <terrypinDELETE@THISdial.pipex.com> wrote in message
news:r2cqa0les4qv5i8ndp2rm9qvsecfiub1eh@4ax.com...

legg <legg@nospam.magma.ca> wrote:

Don't know how or what you are measuring here, but your schematic
provides differing base current drive capacity for the two drive
transistors.

Positive going output signals are current limited by the fixed voltage
present across R7 - the voltage across R10 cannot exceed this value.
Veb and temperature will be signifigant.

Output current is also beta-limited, under worst-case conditions, to
that current provided by the amplified source combination of C2/R5, or
approximately 3mA x hfeQ3 x hfeQ5.

Niether of these limitations apply to the Q4-Q6 complementary section,
which has no emitter resistor back-biasing Q4, nor resistively
limited base current from the collector of Q2.

Thanks for all the replies so far. I will experiment with as many of
those design changes as time allows, once I have a robust performance
from the present circuit.

To summarise the current status (as also reported in the original
thread): I seem to have made progress this evening. It looks as if Q4
and one of the 2N3055s were *both* 'borderline'. Although Q4 was
testing OK on my transistor tester, and showing a reasonable gain, in
exasperation I swapped it for an alternative and immediately started
getting better results.

I now have a combination that delivers a clean-looking sine of about
6.9V rms into 8 ohms. That's on breadboard (with 6 croc-clip
connections to the case-mounted 2N3055s), and with a regulated 24.0 V
bench supply. It remains to be seen if I can retain that good
performance when I rebuild the stripboard circuit and try alternative
supplies.

I intend to replace the two presets for setting the output DC level
and quiescent current respectively. The first is obviously easy to set
(half supply voltage), but how do I determine optimum setting for
quiescent current please?

I may well also experiment with other improvements and design changes
later. But I want to get this working robustly first. Its author, R
Torrens, did so with a variety of supply voltages from 12 to 36 V. His
tests showed distortion was "...around 0.5% or better, which is
approaching hi-fi levels." But I wonder if he had to almost
'hand-pick' his transistors as I seem to be doing for Q4, Q5 and Q6?


--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Stefan Heinzmann]

Terry Pinnell wrote:

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:

Vceo is not the reverse voltage. It is the maximum forward voltage
between collector and emitter with the base left open. So could you
please provide some more detail about your measurement setup?


See my earlier reply to Win's similar query.

Yes, I've seen that, but it didn't make me much wiser. You seem to have
measured the Hfe at a collector current in the milliamp range, with an
unknown voltage from collector to emitter. I'm also still unsure about
your Vceo measurement. You keep talking about reverse breakdown, which
makes me suspect that you applied the test voltage the wrong way. For
Vceo on a NPN transistor you have to apply the test voltage with the
positive end on the collector. The base is left open. Again, it has
nothing to do with reverse breakdown.

If you did that with a 2N3055 and it shows a Vceo below 60V then the
device is damaged. It is vital, however, to ensure that there's no
conduction path from the collector into the base (i.e. through
contamination).

You can also short the base to the emitter, in which case you measure
Vces and not Vceo. Vces is usually higher than Vceo. You should get at
least 100V with a 2N3055.

The gain is only meaningful if you provide information about the
measurement conditions. At the very least, you should provide the
collector current and the collector-emitter voltage. HFe also depends on
the temperature, but that is harder to measure.

Furthermore, you can determine for yourself whether your transistors are
within specification by looking at the data sheet. The 2N3055 is
available from a number of sources, a good start would be On
Semiconductor. Just have a look at their website www.onsemi.com


I have the spec, as I mentioned, from a couple of reference books. But
don't you think it's a tad unrealistic to suggest I can determine
whether a particular transistor fully meets that spec? That would need
an exhaustive set of measurements beyond my capabilities, even if the
effort was worthwhile. For the 2N3055, Towers International Transistor
Selector gives his data:
Vcb max = 100V, Vce max = 60V, Veb max = 7V, Ic max = 15A, Tj max =
200C, Ptot = 115WC, Ft min = 200k, Cob max not specified, Hfe = 20-70,
Hfe bias = 4A.
A.M Ball's Semiconductor data differs somewhat:
fT min = 700k, Vcbo max = 160V, Vebo max = 4V, Ic = 10A, Typical Hfe =
45 at 3A.

I didn't want to suggest that you measure every parameter. That would
also be beyond my own capabilities. But you've measured Hfe and Vceo (at
least you said so, I'm not sure you did the second one right). You can
compare that data with the values from the data sheet yourself. For a
meaningful comparison, however, it is necessary to know the measurement
conditions. They are stated in the data sheet for a good reason. Maybe
the user guides of your testers specify the measurement conditions, or
you can measure them somehow. You could for example use a multimeter to
measure the collector current of the 2N3055 in your transistor tester.

Note, BTW, one thing in the data sheet values above: fT varies a lot
between devices. OnSemi even specifies 2.5MHz. This supports what I
wrote earlier about amplifier stability and bandwidth. If you have a
transistor with Ft = 200kHz you needn't worry much about high-frequency
oscillations. With Ft = 2.5Mhz the situation is quite different. And
those values are minimum values. Actual devices will exceed them.

What I *can* do is determine if a transistor *fails* to meet a key
parameter. That's what I was trying to do with the reverse voltage
breakdown test. But if, as you say, that doesn't give an estimate of
Vceo (to compare with the above 60V), then how do you measure it?

See above.

--
Cheers
Stefan

 

[Terry Pinnell]

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:

Terry Pinnell wrote:
Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:

Vceo is not the reverse voltage. It is the maximum forward voltage
between collector and emitter with the base left open. So could you
please provide some more detail about your measurement setup?


See my earlier reply to Win's similar query.

Yes, I've seen that, but it didn't make me much wiser. You seem to have
measured the Hfe at a collector current in the milliamp range, with an
unknown voltage from collector to emitter.

Thanks for the comprehensive follow up.

On Hfe, you're right. And it was a conscious choice. I could have
rigged up a temporary test circuit to apply higher Ib and Ic and use a
couple of DMMs and plot a chart to get a 'proper' measure. But hey,
life's too short! Neither of my DMM user guides/leaflets are to hand,
but I expect they would confirm low test levels.

I'm also still unsure about
your Vceo measurement. You keep talking about reverse breakdown, which
makes me suspect that you applied the test voltage the wrong way. For
Vceo on a NPN transistor you have to apply the test voltage with the
positive end on the collector. The base is left open. Again, it has
nothing to do with reverse breakdown.

That's down to my careless terminology. It's clear now that I did
measure Vceo, as I intended. (With correct polarity; that would have
been a reasonably easy error to spot <g>.) In fact I gave the unit I
built a decade or two ago the label 'Reverse Breakdown Tester', as its
anticipated primary purpose was to test zeners. It does, of course,
serve well as a 'Vceo Tester'.

If you did that with a 2N3055 and it shows a Vceo below 60V then the
device is damaged.

Specimens 1, 5 and 8 as you saw gave 33, 17 and 21, so they get
disqualified.

It is vital, however, to ensure that there's no
conduction path from the collector into the base (i.e. through
contamination).

That's a possibility. I wasn't too painstaking. And the original
couple I was using still had a little silicon 'heatsink compound' on
surface after unbolting from case.

You can also short the base to the emitter, in which case you measure
Vces and not Vceo. Vces is usually higher than Vceo. You should get at
least 100V with a 2N3055.

I'll try that too.

The gain is only meaningful if you provide information about the
measurement conditions. At the very least, you should provide the
collector current and the collector-emitter voltage. HFe also depends on
the temperature, but that is harder to measure.

OK, but I still think a quick check of the sort I did can be useful.
It provides a helpful *comparative* measure of gain. Across my eight,
it ranged from 13 to 30. All things being equal, including passing the
Vceo>60V requirement, I'd choose specimens from the higher end.

Furthermore, you can determine for yourself whether your transistors are
within specification by looking at the data sheet. The 2N3055 is
available from a number of sources, a good start would be On
Semiconductor. Just have a look at their website www.onsemi.com


I have the spec, as I mentioned, from a couple of reference books. But
don't you think it's a tad unrealistic to suggest I can determine
whether a particular transistor fully meets that spec? That would need
an exhaustive set of measurements beyond my capabilities, even if the
effort was worthwhile. For the 2N3055, Towers International Transistor
Selector gives his data:
Vcb max = 100V, Vce max = 60V, Veb max = 7V, Ic max = 15A, Tj max =
200C, Ptot = 115WC, Ft min = 200k, Cob max not specified, Hfe = 20-70,
Hfe bias = 4A.
A.M Ball's Semiconductor data differs somewhat:
fT min = 700k, Vcbo max = 160V, Vebo max = 4V, Ic = 10A, Typical Hfe =
45 at 3A.

I didn't want to suggest that you measure every parameter. That would
also be beyond my own capabilities.

OK, understod.

But you've measured Hfe and Vceo (at
least you said so, I'm not sure you did the second one right). You can
compare that data with the values from the data sheet yourself. For a
meaningful comparison, however, it is necessary to know the measurement
conditions. They are stated in the data sheet for a good reason. Maybe
the user guides of your testers specify the measurement conditions, or
you can measure them somehow. You could for example use a multimeter to
measure the collector current of the 2N3055 in your transistor tester.

Agreed, but see above.

Note, BTW, one thing in the data sheet values above: fT varies a lot
between devices. OnSemi even specifies 2.5MHz. This supports what I
wrote earlier about amplifier stability and bandwidth. If you have a
transistor with Ft = 200kHz you needn't worry much about high-frequency
oscillations. With Ft = 2.5Mhz the situation is quite different. And
those values are minimum values. Actual devices will exceed them.

What I *can* do is determine if a transistor *fails* to meet a key
parameter. That's what I was trying to do with the reverse voltage
breakdown test. But if, as you say, that doesn't give an estimate of
Vceo (to compare with the above 60V), then how do you measure it?

See above.

Appreciate your on-going help.

See also latest post, attaching current, well-performing circuit.

--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Terry Pinnell]

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

Terry,

One thing that sticks out is that the 2N2219 is too wimpy to drive a 2N3055.
Also, I believe you said the distortion changes as you interchanged the two
output transistors; did you readjust the current through D1 and D2? You
might want to re measure the breakdown voltages on the transistors with the
base and emitter tied together. Throw out any that don't meet spec.

Tam

Thanks. My BJT stocks are a bit low, but I rather liked the size and
feel of those chunky 2N2219As, and their spec looked adequate. Not
sure why you say they're wimpy? My book says Vcb =75V, Vce = 50V, Veb
= 5V, Ic = 800mA, Tj = 175C, Ptot = 800mW (air), and Hfe > 100 at
150mA.

That's an interesting suggestion re Vce. It doesn't seem to feature in
either of my refernce books. What exactly is its merit?


--
Terry Pinnell
Hobbyist, West Sussex, UK

 

[Stefan Heinzmann]

Terry Pinnell wrote:

Stefan Heinzmann <stefan_heinzmann@yahoo.com> wrote:
[...]
I'm also still unsure about
your Vceo measurement. You keep talking about reverse breakdown, which
makes me suspect that you applied the test voltage the wrong way. For
Vceo on a NPN transistor you have to apply the test voltage with the
positive end on the collector. The base is left open. Again, it has
nothing to do with reverse breakdown.


That's down to my careless terminology. It's clear now that I did
measure Vceo, as I intended. (With correct polarity; that would have
been a reasonably easy error to spot <g>.) In fact I gave the unit I
built a decade or two ago the label 'Reverse Breakdown Tester', as its
anticipated primary purpose was to test zeners. It does, of course,
serve well as a 'Vceo Tester'.

Ok, that sorts this one. Thanks for the clarification.

If you did that with a 2N3055 and it shows a Vceo below 60V then the
device is damaged.


Specimens 1, 5 and 8 as you saw gave 33, 17 and 21, so they get
disqualified.

I realized in the meantime that your test current was probably too low
for this test. Somewhere you said that your tester applies 340V through
a 300k series resistance. That limits the current to something around
1mA. For a power transistor like the 3055 this is perilously close to
the leakage (or cut-off) current. In other words the values you measured
may not indicate the breakdown voltage, but be an indication of a high
leakage current. The leakage current tends to rise with temperature, so
if you heat up the transistor you may find the "apparent" breakdown
voltage falls markedly.

Some 2N3055 datasheets specify limits for the cut-off current. For
example, the Fairchild MJE3055T (same chip in different package)
specifies a maximum of 1mA at Vce=70V and the base voltage at 1.5V
/below/ the emitter. This rises to 5mA at 150°C.

Consequently, your measurement of Vceo needs to be done with a collector
current that is significantly above the leakage current. The said
datasheet specifies 200mA collector current as the mesurement condition
for Vceo. That leads to significant power dissipation in the transistor,
so you need to either ensure sufficient cooling or use pulsed
measurement techniques. The latter is probably beyond the means of a
hobbyist. So in your case I would probably try a current of about
10-20mA. The question is whether your breakdown tester can supply this
current level. If so, you may try replacing the series resistor of 300k
with a 15k resistor (or better yet, make it switchable). Watch the power
dissipation!

Before you throw away the suspect transistors, you may want to try the
test with a higher current just to make sure they're not just exemplars
with a higher cut-off current, and there's nothing wrong with the Vceo.

I found a bunch of MJE3055T in my drawer and took the opportunity to put
them into a curve tracer. They are from Wingshing, apparently made in
2000. My five samples showed a Vceo between 250V and 400V, despite their
specification of 60V minimum.

[...]

You can also short the base to the emitter, in which case you measure
Vces and not Vceo. Vces is usually higher than Vceo. You should get at
least 100V with a 2N3055.


I'll try that too.

Actually, i found some manufacturers specify only 70V minimum for Vces.
Whichever, it will be higher than Vceo.

The gain is only meaningful if you provide information about the
measurement conditions. At the very least, you should provide the
collector current and the collector-emitter voltage. HFe also depends on
the temperature, but that is harder to measure.


OK, but I still think a quick check of the sort I did can be useful.
It provides a helpful *comparative* measure of gain. Across my eight,
it ranged from 13 to 30. All things being equal, including passing the
Vceo>60V requirement, I'd choose specimens from the higher end.

That's a reasonable idea. The Fairchild datasheet contains an
interesting graph that shows the dependency between collector current
and gain at Vce=2V. At the left end of the graph the gain for Ic=10mA
shows as ~55. The maximum gain occurs at Ic=200mA, where it surpasses
100. Those are presumably typical values, not minimum ones.

This shows that the gain falls for lower currents. Your tools may well
use currents even below 10mA, as would be appropriate for small signal
transistors. I checked this with one of my MJE3055T devices, and it
shows a gain of about 30 at 150ľA collector current. The gain rises to
about 80 at 150mA collector current.

You see that the results of a measurement depend crucially on the
measurement conditions. If you don't know how your instruments measure a
certain value, you will have difficulties coming to a valid conclusion.
There's a saying in German that looses its rhyme when translating it to
English: "Wer mißt mißt Mist" (Who measures, measures rubbish).

--
Cheers
Stefan