Are small signal npn transistors really so different from one another?

[Paul Burridge]

Hi guys,

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain. Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV! I can't believe this substitution would
result in such a deterioration of the amp's efficiency. I'm
particularly pissed off as I don't have any 2N5769s lying around to
use for the actual circuit; I'd been planning on using some of my
extensive stock of '3904s.
Can anyone verify that such a huge disparity in gain is indeed
possilbe? And the quickest way around the problem to get my 60mV back?

thanks.

p.

 

[John Popelish]

Hi guys,

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain. Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV! I can't believe this substitution would
result in such a deterioration of the amp's efficiency. I'm
particularly pissed off as I don't have any 2N5769s lying around to
use for the actual circuit; I'd been planning on using some of my
extensive stock of '3904s.
Can anyone verify that such a huge disparity in gain is indeed
possilbe? And the quickest way around the problem to get my 60mV back?

If you are saying that one type produces an output of 60 mv (net gain
of 1200, which is pretty great at 40 MHz) and the other produces an
output of 4 uv (net gain of .08), then I agree that this is very
unexpected. They are not that different (one has a higher collector
voltage rating).

Unfortunately, there is almost no high frequency spec, on the data
sheets that are done under the same conditions, to allow an easy
comparison between them.

http://www.fairchildsemi.com/ds/2N/2N5769.pdf
http://www.fairchildsemi.com/ds/2N/2N3904.pdf

The data sheet for the 5769 is very terse, but this one from the same
process family has more details:
http://www.fairchildsemi.com/ds/PN/PN2369A.pdf

I think you made some inadvertent change in the schematic when you
swapped transistors.

 

[Thomas C. Sefranek]

If you are saying that one type produces an output of 60 mv (net gain
of 1200, which is pretty great at 40 MHz) and the other produces an
output of 4 uv (net gain of .08), then I agree that this is very
unexpected. They are not that different (one has a higher collector
voltage rating).

Unfortunately, there is almost no high frequency spec, on the data
sheets that are done under the same conditions, to allow an easy
comparison between them.

http://www.fairchildsemi.com/ds/2N/2N5769.pdf
http://www.fairchildsemi.com/ds/2N/2N3904.pdf

The data sheet for the 5769 is very terse, but this one from the same
process family has more details:
http://www.fairchildsemi.com/ds/PN/PN2369A.pdf

I think you made some inadvertent change in the schematic when you
swapped transistors.

Look at the switching times...
It tells the story CLEARLY.


--
*
| __O Thomas C. Sefranek [email protected]
|_-\<,_ Amateur Radio Operator: WA1RHP
(*)/ (*) Bicycle mobile on 145.41, 448.625 MHz

http://hamradio.cmcorp.com/inventory/Inventory.html
http://www.harvardrepeater.org

 

[Gary Lecomte]

While small signal transistors are usually simular, they don't make
Thousands of different ones, Just for the fun of it. They do have
different currents, Frequency ranges, voltages, HFE's, gains and other
parameters.

But if you need to incease the gain of the 2N3904, connecting two
together in a "Darlington configuration" should do it for you.

However also remember different transistors require different base
bias currents.

Take care.....Gary
*********************************************************************

 

[Winfield Hill]

Paul Burridge wrote...

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain. Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV! I can't believe this substitution would
result in such a deterioration of the amp's efficiency. I'm
particularly pissed off as I don't have any 2N5769s lying around to
use for the actual circuit; I'd been planning on using some of my
extensive stock of '3904s.
Can anyone verify that such a huge disparity in gain is indeed
possilbe? And the quickest way around the problem to get my 60mV back?

Paul, good circuit designs should not be dependent upon highly-specific
properties of individual components, unless those properties come from
basic device physics, e.g. Ebers-Moll, or other reliable considerations.

For example, the dc operating bias points of amplifier circuits should
not be dependent upon a transistor's beta, or the exact value of Vbe.

Unfortunately optimizing a design in Spice can lead exactly to such a
bad result, wherein the design will only work with said Spice device,
at a specific temperature and supply voltage. I wonder if you have
fallen into such a trap, so that one of your stages is saturated, or
otherwise rendered non-operational by changing transistors. Please
note that in real life such a design would also fail when made with
the normal part-to-part variation of the same type. BTW, I note the
two Spice models you posted are actually quite similar transistors,
and a good design should operate similarly with either type. One has
more capacitance, etc, so the bandwidth or HF gain could suffer some.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Fred Bloggs]

Hi guys,

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain.
Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV!

Chances are that your optimized biasing is putting something in cut-off
or saturation. You might run a DC Operating Point analysis and see
what's up.

 

[R.Legg]

Hi guys,

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain. Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV! I can't believe this substitution would
result in such a deterioration of the amp's efficiency.

It won't.

Troubleshoot and copare the models the same way you would a 'real'
circuit, section by section.

Don't forget DC static bias.

RL

 

[Paul Burridge]

Look at the switching times...
It tells the story CLEARLY.

What has switching times to do with it? I'm not using these devices in
a switching application; rather an RF amplifier.

 

[Paul Burridge]

Troubleshoot and copare the models the same way you would a 'real'
circuit, section by section.

Don't forget DC static bias.

Thanks all,

I've run two operating point checks; one for each circuit where in
each case all the transistors are either one type or the other. Got
some very interesting results...

Remember, 3 stages in c/e mode and the voltages around each transistor
were as follows:

Circuit using the 2N5769s:

Q1: e=4.39; b=5.04; c=4.79
Q2: e=4.46; b=5.12; c=7.60
Q3: e=4.22; b=4.89; c=5.06


Circuit using the 2N3904s:

Q1: e=4.50; b=5.19; c=4.59
Q2: e=5.07: b=5.76; c=6.95
Q3: e=4.49; b=5.20; c=4.59


As you can see, there are some very low Vce readings. I guess this
means I'm too close to saturation in some instances?
Comments/observations gentlemen, please.

 

[Leon Heller]

Hi guys,

I just designed (in Spice) this chain of three c/e untuned amplifying
stages to get from 50uV in to 60mV out (at 40Mhz) using what I thought
were 2N3904 transistors. Optimised the biasing arrangements over
several hours to get maximum Vgain. Now I've realised I was using the
model for the 2N5769 instead by mistake. Upon 'plugging back in' the
2n3904 models into the schematic however, a fresh simulation showed
the ouput was now only 4uV! I can't believe this substitution would
result in such a deterioration of the amp's efficiency. I'm
particularly pissed off as I don't have any 2N5769s lying around to
use for the actual circuit; I'd been planning on using some of my
extensive stock of '3904s.
Can anyone verify that such a huge disparity in gain is indeed
possilbe? And the quickest way around the problem to get my 60mV back?

I think your problem is due to 'designing' with SPICE. You'd be better
off using proper design techniques then using SPICE to check your design.

Leon

 

[Winfield Hill]

Paul Burridge wrote...

As you can see, there are some very low Vce readings. I guess
this means I'm too close to saturation in some instances?

Indeed. In general for high-frequency amplifiers you want
to use Vce = 2V or more to reduce Ccb. Furthermore the
last stage needs more room for output-voltage swing.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[John Popelish]

Thanks all,

I've run two operating point checks; one for each circuit where in
each case all the transistors are either one type or the other. Got
some very interesting results...

Remember, 3 stages in c/e mode and the voltages around each transistor
were as follows:

Circuit using the 2N5769s:

Q1: e=4.39; b=5.04; c=4.79
Q2: e=4.46; b=5.12; c=7.60
Q3: e=4.22; b=4.89; c=5.06

Circuit using the 2N3904s:

Q1: e=4.50; b=5.19; c=4.59
Q2: e=5.07: b=5.76; c=6.95
Q3: e=4.49; b=5.20; c=4.59

As you can see, there are some very low Vce readings. I guess this
means I'm too close to saturation in some instances?
Comments/observations gentlemen, please.

I am amazed that you got much gain out of either of them Do these two
versions still show the 2000:1 gain variation you spoke of, earlier?

Regardless, I am almost certain that you are not anywhere optimum
biasing in either case.

Would you post your schematic on a.b.s.e?

 

[Jim Thompson]

I think your problem is due to 'designing' with SPICE. You'd be better
off using proper design techniques then using SPICE to check your design.

Leon

Amen!

There are four BASIC factors that differentiate bipolar transistors...

(1) Physical area... defines maximum current AND power
(2) Doping levels... maximum voltage
(3) Finger structures... current handling AND fT
(4) Packaging... power handling and high frequency capability

Beta, which most everyone has a fetish over, is a function of the
first three.

...Jim Thompson

 

[Jim Thompson]

I am amazed that you got much gain out of either of them Do these two
versions still show the 2000:1 gain variation you spoke of, earlier?

Regardless, I am almost certain that you are not anywhere optimum
biasing in either case.

Would you post your schematic on a.b.s.e?

I'd like to see the schematic also.

...Jim Thompson

 

[Winfield Hill]

John Popelish wrote...

I am amazed that you got much gain out of either of them ...

It's only a gain of about 10 per stage on average, and Paul
didn't specify the output load. Maybe it's not 50 ohms.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Paul Burridge]

Paul, good circuit designs should not be dependent upon highly-specific
properties of individual components, unless those properties come from
basic device physics, e.g. Ebers-Moll, or other reliable considerations.

You're absolutely right as ever. I was forgetting this fundamental
principle. :-(

For example, the dc operating bias points of amplifier circuits should
not be dependent upon a transistor's beta, or the exact value of Vbe.

Unfortunately optimizing a design in Spice can lead exactly to such a
bad result, wherein the design will only work with said Spice device,
at a specific temperature and supply voltage. I wonder if you have
fallen into such a trap, so that one of your stages is saturated, or
otherwise rendered non-operational by changing transistors.

Yup. I've been such a dumb-ass as you say in America. Maybe
transistors give best gain when driven close to saturation, I don't
know, but it would certainly account for how my optimising the biasing
has rendered a model change a disaster.
I've been *very* slapdash since I discovered Spice. Instead of sitting
down and working out the bias conditions properly on paper first, I
typically draw up a schematic on the computer and pull values straight
out of the air as they 'look about right under the circumstances.' I
then wonder why the thing doesn't work very well (or at all). Only on
going back to doing the sums on paper can I see where I've gone wrong
- and it's often by an order of magnitude with currents and voltages
all over the place. It's laziness, I guess, and it doesn't work in the
long run! The temptation is it's just *so* easy to try a dozen
different combination of values in minutes just to eek out a bit more
gain here and there. But I now know that's not a very good idea!

 

[John Popelish]

John Popelish wrote...

It's only a gain of about 10 per stage on average, and Paul
didn't specify the output load. Maybe it's not 50 ohms.

For a broadband amplifier at 40 MHz built by trial and error, I am
still amazed at an average stage voltage gain of 10, unless he got it
by finishing with an output stage that drives no external load.

 

[Paul Burridge]

John Popelish wrote...

It's only a gain of about 10 per stage on average, and Paul
didn't specify the output load. Maybe it's not 50 ohms.

It's 380 ohms in this particular case, chaps.

 

[Paul Burridge]

I am amazed that you got much gain out of either of them Do these two
versions still show the 2000:1 gain variation you spoke of, earlier?

Regardless, I am almost certain that you are not anywhere optimum
biasing in either case.

Would you post your schematic on a.b.s.e?

I'll try to do so later, but am off up the pub in a minute. Does
anyone know if it's possible to post an LTSpice schematic to a
newsgroup in a way that's legible to everyone?

 

[Paul Burridge]

I think your problem is due to 'designing' with SPICE.

Guilty. See my other reply to Win.