Where's my degree

[Fred Bartoli]

Ok, this is probably a fraction of a degree.

Hunting for a supposed thermal offset in a circuit I set up this little
contraption to get/verify some figures but I fail to see anything...


+40V--+-------+----.
| | |
.-. |/ .-.
adjust | |<---| R3| |2K
clamp | | |> | |
level '-' | '-' 100n
| Q3 | | ||
GND '----+-------||---+---->
| || | 100uV/div
|/ .-. scope input
+25.7V >--| Q2 | |
|> | |1G
| '-'
| |
.-. GND
Rc| |10K
| |or short
'-'
|
|
|/ Q1 = TUT
+5.7V >--|
|> 20K
| ___
+---|___|-.
| R2 |
R1.-. |
2K| | +-||
| | ->||
'-' +-||---< pulse
| |
GND GND

Q1 is the TUT (transistor under test), Q3 is there to clamp the signal
amplitude and avoid thermal distorsion in the scope stages.

With Rc=short Q1 dissipates 5mW or 5.5mW depending on the pulse input state.

Tested Q1 are some TO92 small signal devices, like BC548-9, 2N3904,
2222, ...

Pulse has been varied to check for thermal effects.
Say Tp= 50ms, duty ratio = 20%.
Looking at some moto datasheet for the BC558 we see the transient
thermal resistance to be 0.25*200K/W = 50K/W
500uW*50K/W*2mV/K = 50uV.

I should see it, but nada. Down to the uV level.
Strange...

 

[john jardine]

"Fred Bartoli"

Ok, this is probably a fraction of a degree.

Hunting for a supposed thermal offset in a circuit I set up this little
contraption to get/verify some figures but I fail to see anything...


+40V--+-------+----.
| | |
.-. |/ .-.
adjust | |<---| R3| |2K
clamp | | |> | |
level '-' | '-' 100n
| Q3 | | ||
GND '----+-------||---+---->
| || | 100uV/div
|/ .-. scope input
+25.7V >--| Q2 | |
|> | |1G
| '-'
| |
.-. GND
Rc| |10K
| |or short
'-'
|
|
|/ Q1 = TUT
+5.7V >--|
|> 20K
| ___
+---|___|-.
| R2 |
R1.-. |
2K| | +-||
| | ->||
'-' +-||---< pulse
| |
GND GND

Q1 is the TUT (transistor under test), Q3 is there to clamp the signal
amplitude and avoid thermal distorsion in the scope stages.

With Rc=short Q1 dissipates 5mW or 5.5mW depending on the pulse input state.

Tested Q1 are some TO92 small signal devices, like BC548-9, 2N3904,
2222, ...

Pulse has been varied to check for thermal effects.
Say Tp= 50ms, duty ratio = 20%.
Looking at some moto datasheet for the BC558 we see the transient
thermal resistance to be 0.25*200K/W = 50K/W
500uW*50K/W*2mV/K = 50uV.

I should see it, but nada. Down to the uV level.
Strange...

How's that clamp transistor work?. I can see the need for one due the
big -0.5V step but can't get my head round how it's actually clamping.

 

[Winfield]

Ok, this is probably a fraction of a degree.

Hunting for a supposed thermal offset in a circuit I set up this little
contraption to get/verify some figures but I fail to see anything...

+40V--+-------+----.
| | |
.-. |/ .-.
adjust | |<---| R3| |2K
clamp | | |> | |
level '-' | '-' 100n
| Q3 | | ||
GND '----+-------||---+---->
| || | 100uV/div
|/ .-. scope input
+25.7V >--| Q2 | |
|> | |1G
| '-'
| |
.-. GND
Rc| |10K
| |or short
'-'
|
|
|/ Q1 = TUT
+5.7V >--|
|> 20K
| ___
+---|___|-.
| R2 |
R1.-. |
2K| | +-||
| | ->||
'-' +-||---< pulse
| |
GND GND

Q1 is the TUT (transistor under test), Q3 is there to clamp the
signal amplitude and avoid thermal distorsion in the scope stages.

With Rc=short Q1 dissipates 5mW or 5.5mW depending on the pulse input
state. Tested Q1 are some TO92 small signal devices, like BC548-9,
2N3904, 2222, ...

Pulse has been varied to check for thermal effects.
Say Tp= 50ms, duty ratio = 20%. Looking at some moto datasheet for
the BC558 we see the transient thermal resistance to be 0.25*200K/W
= 50K/W
500uW*50K/W*2mV/K = 50uV.

I should see it, but nada. Down to the uV level. Strange...

Nice drawing. Can you run some numbers for the lazy reader? We
see a very small change in power dissipation, from 5.0 to 5.5mW,
which we imagine won't cause much temperature rise in a device
that can handle 500mW or more (is it 0.1 or 0.2C?). And we see
a current sink with 5V across the CS-setting resistor(s), and a
by-comparison microscopic change in Vbe from the aforesaid small
change in temperature (about 0.1m/5V = 25ppm?). Just how much
current change do you expect (about 0.1mV across the load)?

There's gotta be a better way!

Thanks for the diversion, hav'ta get back to the taxes!

 

[Tony Williams]

Ok, this is probably a fraction of a degree.

Hunting for a supposed thermal offset in a circuit I set up this
little contraption to get/verify some figures but I fail to see
anything...

Snip diagram.

With Rc=short Q1 dissipates 5mW or 5.5mW depending on the pulse
input state.

20V and 2.5mA is 50mW?

Pulse has been varied to check for thermal effects.
Say Tp= 50ms, duty ratio = 20%.
Looking at some moto datasheet for the BC558 we see the transient
thermal resistance to be 0.25*200K/W = 50K/W

I could not find a data sheet for the BC458 (npn) with
a transient thermal resistance graph, so is this why
you used the BC558 (matching pnp) data sheet?

Looking at the Semtech data sheet for the BC558,
with a tp of 50mS and a tp/T of 50/250mS, rth-A looks
to be about 180 K/W.

500uW*50K/W*2mV/K = 50uV.

So it looks more like a 5mW step and 180 K/W.

Would it be better to step the dissipation by
stepping the 25.7V on Q2's base (by 2V or so)?

 

[Fred Bartoli]

Tony Williams a écrit :

Snip diagram.


20V and 2.5mA is 50mW?

Arghh. You're right, and 250uA.20V is 5mW, so it's even worse than I
thought.

I could not find a data sheet for the BC458 (npn) with
a transient thermal resistance graph, so is this why
you used the BC558 (matching pnp) data sheet?

Yep.

Looking at the Semtech data sheet for the BC558,
with a tp of 50mS and a tp/T of 50/250mS, rth-A looks
to be about 180 K/W.

I used:
http://www.onsemi.com/pub/Collateral/BC556B-D.PDF

The way they phrase it isn't very clear, but I think the abscisse time
is the pulse duration, not the repetition rate (which wouldn't make
sense for single pulse BTW).

Also, here we are not interested in peak temperature, rather in
temperature excursion, so we have to use the single pulse curve, not the
ones with duty ratio as parameter. This temperature excursion will ride
on an elevated average temperature.

From this, I read the 'excursion Rth' to be Rth_ex = 0.25*Rth = 50K/W

So it looks more like a 5mW step and 180 K/W.

I make it 5mW and 50K/W, so it's now 500uV that are missing.

Would it be better to step the dissipation by
stepping the 25.7V on Q2's base (by 2V or so)?

Don't know why it should be better, but it'd be less simple.

 

[Tony Williams]

I used:
http://www.onsemi.com/pub/Collateral/BC556B-D.PDF

Thanks.... might as well sing off the same songsheet.

[snip]

Don't know why it should be better, but it'd be less simple.

Because you are currently looking for a small slope
after a relatively large current step (10%), plus
the change in Q1 current must produce a small change
in Q1's Vbe,

Stepping the 25.7V can produce the same change in
dissipation without having any Q1 current step, maybe
a small front end transient only.

It's not that difficult. Put a 2222 ohm resistor
in series with the base of Q2, and then move the top
of the 20k to the base. That should step Vb(Q2)
from 25.7 to 23.13, or by -10%.

 

[James Arthur]

I used:
http://www.onsemi.com/pub/Collateral/BC556B-D.PDF

Thanks.... might as well sing off the same songsheet.

[snip]

Don't know why it should be better, but it'd be less simple.

Because you are currently looking for a small slope
after a relatively large current step (10%), plus
the change in Q1 current must produce a small change
in Q1's Vbe,

But only during the pulse. The measurement, AIUI, is made
afterwards. Stepping Vce would be cleaner then, since it doesn't
directly affect Vbe, but the error from briefly stepping i(e) isn't
great.

Stepping the 25.7V can produce the same change in
dissipation without having any Q1 current step, maybe
a small front end transient only.

It's not that difficult. Put a 2222 ohm resistor
in series with the base of Q2, and then move the top
of the 20k to the base. That should step Vb(Q2)
from 25.7 to 23.13, or by -10%.

Since he's trying to measure Vbe(Q1), why not take the 'scope signal
directly from Q1's emitter? That avoids the need for clamping (not
knowing the clamp level, Q3 bugs me).

And, if he then steps Vce as you suggest, Q1's emitter gives the
error signal directly.

Best regards,
James Arthur

 

[Tony Williams]

Since he's trying to measure Vbe(Q1), why not take the 'scope
signal directly from Q1's emitter? That avoids the need for
clamping (not knowing the clamp level, Q3 bugs me).

Buggerit. You've just trashed my next post.

And, if he then steps Vce as you suggest, Q1's emitter gives the
error signal directly.

Let's connect Q1's base to +0.7V (using a Vbe at 2.5mA)
and the bottom end of the 2K to -5V.

 

[John Larkin]

I used:
http://www.onsemi.com/pub/Collateral/BC556B-D.PDF

Thanks.... might as well sing off the same songsheet.

[snip]

Would it be better to step the dissipation by
stepping the 25.7V on Q2's base (by 2V or so)?
Don't know why it should be better, but it'd be less simple.

Because you are currently looking for a small slope
after a relatively large current step (10%), plus
the change in Q1 current must produce a small change
in Q1's Vbe,

But only during the pulse. The measurement, AIUI, is made
afterwards. Stepping Vce would be cleaner then, since it doesn't
directly affect Vbe, but the error from briefly stepping i(e) isn't
great.

Stepping the 25.7V can produce the same change in
dissipation without having any Q1 current step, maybe
a small front end transient only.

It's not that difficult. Put a 2222 ohm resistor
in series with the base of Q2, and then move the top
of the 20k to the base. That should step Vb(Q2)
from 25.7 to 23.13, or by -10%.

Since he's trying to measure Vbe(Q1), why not take the 'scope signal
directly from Q1's emitter? That avoids the need for clamping (not
knowing the clamp level, Q3 bugs me).

And, if he then steps Vce as you suggest, Q1's emitter gives the
error signal directly.

Best regards,
James Arthur

Yeah, that's clean:

Vc
|
|
|
c
+5---r1---b
e
|
|
|
+------- scope
|
r2
|
gnd


Step Vc from, say, +10 to +20. That will change power dissipation and
the delta-Vbe waveform should say a lot about the junction thermal
time constants. R1 is maybe 33 ohms, just to keep it from oscillating.

John

 

[James Arthur]

I used:
http://www.onsemi.com/pub/Collateral/BC556B-D.PDF
Thanks.... might as well sing off the same songsheet.
[snip]
Would it be better to step the dissipation by
stepping the 25.7V on Q2's base (by 2V or so)?
Don't know why it should be better, but it'd be less simple.
Because you are currently looking for a small slope
after a relatively large current step (10%), plus
the change in Q1 current must produce a small change
in Q1's Vbe,

But only during the pulse. The measurement, AIUI, is made
afterwards. Stepping Vce would be cleaner then, since it doesn't
directly affect Vbe, but the error from briefly stepping i(e) isn't
great.

Since he's trying to measure Vbe(Q1), why not take the 'scope signal
directly from Q1's emitter? That avoids the need for clamping (not
knowing the clamp level, Q3 bugs me).

And, if he then steps Vce as you suggest, Q1's emitter gives the
error signal directly.

Best regards,
James Arthur

Yeah, that's clean:

Vc
|
|
|
c
+5---r1---b
e
|
|
|
+------- scope
|
r2
|
gnd

Step Vc from, say, +10 to +20. That will change power dissipation and
the delta-Vbe waveform should say a lot about the junction thermal
time constants. R1 is maybe 33 ohms, just to keep it from oscillating.

John

Right. The setup allows continuous real-time measurement of the
thermal impulse-response, too.

James

 

[James Arthur]

Buggerit. You've just trashed my next post.


Let's connect Q1's base to +0.7V (using a Vbe at 2.5mA)
and the bottom end of the 2K to -5V.

Sounds good. Add a small potentiometer adjustment to that base
voltage and he can d.c.-couple the 'scope, eliminating more potential
gremlins (from 'scope bias currents, rectified RFI, blah blah blah).

Okay Fred, this one's whipped...next please!

James Arthur