Low noise at low Vce

[Thomas Philips]

I've heard it said that bipolar transistors generate less noise when
operated with Vce very close to 0. I assume the noise in question is
flicker noise or current noise, as the voltage noise depends primarily
on rbb and gm, both of which are essentially independent of Vce.
Questions:

1. Does current noise (which is really shot noise in Ib) decline as
Vce is reduced with Ic fixed? Does beta vary significantly with Vce?

2. Does flicker noise (or 1/f) decline as Vce is reduced while Ic is
fixed? Is there a relationship between the 1/f corner frequency, Ic
and Vce?

3. What is the mechanism by which noise declines with Vce?

4. What is the impact of low Vce on fT?

Thanks in advance

Thomas Philips

 

[Kevin Aylward]

Thomas Philips wrote:

I've heard it said that bipolar transistors generate less noise when
operated with Vce very close to 0.

Actually its Vbc ~ 0. A Vce near 0V is in saturation.

I assume the noise in question is
flicker noise

Yes.

or current noise, as the voltage noise depends primarily
on rbb and gm, both of which are essentially independent of Vce.
Questions:

Shot noise is also, essentially, indepandant of Vcb.

1. Does current noise (which is really shot noise in Ib) decline as
Vce is reduced with Ic fixed? Does beta vary significantly with Vce?

Again, by current noise, I assume you mean normal white shot noise.
This, is pretty much independent of vbc.

2. Does flicker noise (or 1/f) decline as Vce is reduced while Ic is
fixed?

Yes.

Is there a relationship between the 1/f corner frequency, Ic
and Vce?

There is some dependence on the 1/f corner and current, but not a lot.

3. What is the mechanism by which noise declines with Vce?

Some 1/f noise is associated with parasitic/contamination leakages
across junctions. With OV across the base emitter the leakage is
obviously going to go down.

There is no much wrote about this effect. I discovered it by accident
about 25 years ago. I was building guitar pre-amps. When I started to
investigate it, I didn't find anything of substance on the effect. I
haven't looked recently.

The basic effect that as soon as Vbc drops below 0V, the noise falls of
dramatically. This means that Vce can still be say, 0.5V, and therefore
not be in saturation, which would kill the response completely.

4. What is the impact of low Vce on fT?

It reduces it

Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[Winfield Hill]

Kevin Aylward wrote...

Thomas Philips wrote:
I've heard it said that bipolar transistors generate less
noise when operated with Vce very close to 0.

Actually its Vbc ~ 0. A Vce near 0V is in saturation.

I assume the noise in question is flicker noise

Yes.

2. Does flicker noise (or 1/f) decline as Vce is reduced
while Ic is fixed?

Yes.

3. What is the mechanism by which noise declines with Vce?

Some 1/f noise is associated with parasitic/contamination
leakages across junctions. With OV across the base emitter the
leakage is obviously going to go down.

There is no much written about this effect. I discovered it by
accident about 25 years ago. I was building guitar pre-amps.
When I started to investigate it, I didn't find anything of
substance on the effect. I haven't looked recently.

The basic effect that as soon as Vbc drops below 0V, the noise
falls of dramatically. This means that Vce can still be say,
0.5V, and therefore not be in saturation, which would kill the
response completely.

You're saying the flicker (1/f) noise is reduced as Vcb, being
positive, is reduced towards zero, and the flicker noise is
low at Vcb = 0, and completely gone for Vcb slightly negative,
or negative by a few hundred millivolts?

BTW, I assume the 1/f noise you were observing was input current
noise density (i_n) rather than input voltage noise (v_n)?

Do you remember what type of transistor you were working with?
2n5088, 2sd786, 2n5962, 2n2484, LM394, etc.?

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Kevin Aylward]

Winfield Hill wrote:

I know this was a private email, but my emails were bounced back. There
doesn't seem to be anything here that I would say is compromising, so I
have replied here. I sniped the bit about, the issues your girlfriend
had with your other girlfriend and your wife.

Kevin Aylward wrote...

Thomas Philips wrote:
I've heard it said that bipolar transistors generate less
noise when operated with Vce very close to 0.

Actually its Vbc ~ 0. A Vce near 0V is in saturation.

I assume the noise in question is flicker noise

Yes.

2. Does flicker noise (or 1/f) decline as Vce is reduced
while Ic is fixed?

Yes.

3. What is the mechanism by which noise declines with Vce?

Some 1/f noise is associated with parasitic/contamination
leakages across junctions. With OV across the base emitter the
leakage is obviously going to go down.

There is no much written about this effect. I discovered it by
accident about 25 years ago. I was building guitar pre-amps.
When I started to investigate it, I didn't find anything of
substance on the effect. I haven't looked recently.

The basic effect that as soon as Vbc drops below 0V, the noise
falls of dramatically. This means that Vce can still be say,
0.5V, and therefore not be in saturation, which would kill the
response completely.

You're saying the flicker (1/f) noise is reduced as Vcb, being
positive, is reduced towards zero, and the flicker noise is
low at Vcb = 0, and completely gone for Vcb slightly negative,
or negative by a few hundred millivolts?

No

I don't know how much it goes down exactly. It doesn't disappear,
probably, just
gets lower.

I never got around to making any real measurements. It was all done by
ear. I
just noticed the effect, making sure that the signal didn't go away as
well. It
was a very obvious reduction in noise.

BTW, I assume the 1/f noise you were observing was input current
noise density (i_n) rather than input voltage noise (v_n)?

Again, I don't recall my exact set-up. When I later looked for some sort
of
research on it I drew a blank. It was only much later that I noticed
some vague
references to it. For example, I think the LM394 quote measurements at
vbc=0,
but they don't say why they did that.

Do you remember what type of transistor you were working with?
2n5088, 2sd786, 2n5962, 2n2484, LM394, etc.?

Not really, but probably a BC109, as it was some time ago. How about you
popping
into your lab and doing a few measurements. The easiest way to do it is
to use a
cascode so that you can vary Vcb without messing with the other bias
conditions. Obviously, the already low 1/f devices will probably not
show much.
The 2n4403 often have highish 1/f.

Best Regards,

Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[Frank Bemelman]

"Kevin Aylward" <kevindotaylwardEXTRACT@anasoft.co.uk> schreef in bericht
news:wlX6c.70$Kt4.57@newsfep3-gui.server.ntli.net...

I know this was a private email, but my emails were bounced back. There
doesn't seem to be anything here that I would say is compromising, so I
have replied here. I sniped the bit about, the issues your girlfriend
had with your other girlfriend and your wife.

That's very thoughtful

--
Thanks, Frank.
(remove 'x' and 'invalid' when replying by email)

 

[Winfield Hill]

Kevin Aylward wrote...

I sniped the bit about, the issues your girlfriend had with
your other girlfriend and your wife.

You mean where I left my wife with the impression I was with
one of my girlfriends, and I told her I was with my wife, so
I could go to the lab, undisturbed?

Do you remember what type of transistor you were working with?
2n5088, 2sd786, 2n5962, 2n2484, LM394, etc.?

Not really, but probably a BC109, as it was some time ago. How
about you popping into your lab and doing a few measurements.
The easiest way to do it is to use a cascode so you can vary Vcb
without messing with the other bias conditions. Obviously, the
already low 1/f devices will probably not show much. The 2n4403
often have highish 1/f.

OK, good idea.

p.s. I'll fix my picovolt-dot-com email one of these days. :&gt

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Jim Thompson]

On 20 Mar 2004 14:32:18 -0800, Winfield Hill
<Winfield_member@newsguy.com> wrote:

Winfield Hill wrote...

Kevin Aylward wrote...

I sniped the bit about, the issues your girlfriend had with
your other girlfriend and your wife.

You mean where I left my wife with the impression I was with
one of my girlfriends, and I told her I was with my wife, so
I could go to the lab, undisturbed?

Do you remember what type of transistor you were working with?
2n5088, 2sd786, 2n5962, 2n2484, LM394, etc.?

Not really, but probably a BC109, as it was some time ago. How
about you popping into your lab and doing a few measurements.
The easiest way to do it is to use a cascode so you can vary Vcb
without messing with the other bias conditions. Obviously, the
already low 1/f devices will probably not show much. The 2n4403
often have highish 1/f.

OK, good idea.

OK, I fired up my trusty hp4470A Transistor Noise Analyzer, and
tried a 2sd786, which pegged in at about 0.6nV root Hz at 10mA.
I could see the effects of collector current and frequency, but
nothing showing for Vce, from 0.5V to 10V, either in e_n or i_n.
I'll try more noisy transistors sometime soon. Any suggestions?

Thanks,
- Win

whill_at_picovolt-dot-com

Win, What do you get for a 2N2222? I seem to recall seeing curves
that showed a voltage sensitivity.

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice480)460-2350 | |
| E-mail Address at Website Fax480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

John "Peace for our Time" Kerry, Neville Chamberlain of this Century

 

[Kevin Aylward]

Winfield Hill wrote:

Winfield Hill wrote...

Kevin Aylward wrote...

I sniped the bit about, the issues your girlfriend had with
your other girlfriend and your wife.

You mean where I left my wife with the impression I was with
one of my girlfriends, and I told her I was with my wife, so
I could go to the lab, undisturbed?

Do you remember what type of transistor you were working with?
2n5088, 2sd786, 2n5962, 2n2484, LM394, etc.?

Not really, but probably a BC109, as it was some time ago. How
about you popping into your lab and doing a few measurements.
The easiest way to do it is to use a cascode so you can vary Vcb
without messing with the other bias conditions. Obviously, the
already low 1/f devices will probably not show much. The 2n4403
often have highish 1/f.

OK, good idea.

OK, I fired up my trusty hp4470A Transistor Noise Analyzer, and
tried a 2sd786, which pegged in at about 0.6nV root Hz at 10mA.

Well, obviously, this is a low noise device, so I doubt if this would
show much of an effect.

I could see the effects of collector current and frequency, but
nothing showing for Vce, from 0.5V to 10V, either in e_n or i_n.
I'll try more noisy transistors sometime soon. Any suggestions?

One guess is that the noise that gets reduced is the excess noise that
bad devices make due to process imperfections. For the 2n4403 case, its
a hit and miss. Some devices just happen to be quite noisy. Its possible
later processes are much better so the effect is not so apparent.
However, if it is a general effect that can always work on non
guaranteed low 1/f devices, it would be a useful production enhancer by
making sourcing componets easier.

Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[Winfield Hill]

Kevin Aylward wrote...

OK, I fired up my trusty hp4470A Transistor Noise Analyzer, and
tried a 2sd786, which pegged in at about 0.6nV root Hz at 10mA.

Well, obviously, this is a low noise device, so I doubt if this
would show much of an effect.

Yes, I remember you asserting this, but I had to start with a
low-noise transistor to test the 4470A, to be sure it was still
working after years of neglect. Then I ran out of time. :&gt

I could see the effects of collector current and frequency, but
nothing showing for Vce, from 0.5V to 10V, either in e_n or i_n.
I'll try more noisy transistors sometime soon. Any suggestions?

One guess is that the noise that gets reduced is the excess noise
that bad devices make due to process imperfections. For the 2n4403
case, its a hit and miss. Some devices just happen to be quite noisy.
Its possible later processes are much better so the effect is not
so apparent. However, if it is a general effect that can always
work on non guaranteed low 1/f devices, it would be a useful
production enhancer by making sourcing componets easier.

OK, the next time I have the chance, I'll look for a really noisy
part first, then make careful measurements. Should I concentrate
on PNP or NPN parts?

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Kevin Aylward]

Winfield Hill wrote:

Kevin Aylward wrote...

OK, I fired up my trusty hp4470A Transistor Noise Analyzer, and
tried a 2sd786, which pegged in at about 0.6nV root Hz at 10mA.

Well, obviously, this is a low noise device, so I doubt if this
would show much of an effect.

Yes, I remember you asserting this, but I had to start with a
low-noise transistor to test the 4470A, to be sure it was still
working after years of neglect. Then I ran out of time. :&gt

I could see the effects of collector current and frequency, but
nothing showing for Vce, from 0.5V to 10V, either in e_n or i_n.
I'll try more noisy transistors sometime soon. Any suggestions?

One guess is that the noise that gets reduced is the excess noise
that bad devices make due to process imperfections. For the 2n4403
case, its a hit and miss. Some devices just happen to be quite noisy.
Its possible later processes are much better so the effect is not
so apparent. However, if it is a general effect that can always
work on non guaranteed low 1/f devices, it would be a useful
production enhancer by making sourcing componets easier.

OK, the next time I have the chance, I'll look for a really noisy
part first, then make careful measurements. Should I concentrate
on PNP or NPN parts?

Beats me. I noticed the effect (in NPNs), but that's about it. Its a
*real* research job. The standard theory on shot and thermal noise
doesn't explain it. As I noted, the guess is that it is random leakages
(feedback) across the collector base, injecting noise into the base
emitter. I don't know enough about processes to know if npns or pnp are
better or worse.

Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[John Woodgate]

I read in sci.electronics.design that Kevin Aylward <kevindotaylwardEXTR
ACT@anasoft.co.uk> wrote (in <sGm7c.1285$cw6.830@newsfep3-gui.server.ntl
i.net&gt about 'For Win Hill - Low noise at low Vce', on Sun, 21 Mar
2004:

Beats me. I noticed the effect (in NPNs), but that's about it. Its a
*real* research job. The standard theory on shot and thermal noise
doesn't explain it. As I noted, the guess is that it is random leakages
(feedback) across the collector base, injecting noise into the base
emitter. I don't know enough about processes to know if npns or pnp are
better or worse.

I have a collection of noisy BC109s, rejected by a screening process.
The chap who gave them to me thought that surface contamination was the
main cause, because some would 'clean up', more or less permanently, if
warmed up with a hot air gun.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Winfield Hill]

John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening process.
The chap who gave them to me thought that surface contamination was the
main cause, because some would 'clean up', more or less permanently, if
warmed up with a hot air gun.

Could you send me a handful? Or a few?

Winfield Hill hill-at-rowland-dot-org
Rowland Institute
100 Land Blvd
Cambridge, MA 02142 USA

Thanks,
- Win

whill_at_picovolt-dot-com

 

[John Woodgate]

I read in sci.electronics.design that Winfield Hill
<Winfield_member@newsguy.com> wrote (in <c3kv5701546@drn.newsguy.com&gt
about 'For Win Hill - Low noise at low Vce', on Sun, 21 Mar 2004:

John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening process.
The chap who gave them to me thought that surface contamination was the
main cause, because some would 'clean up', more or less permanently, if
warmed up with a hot air gun.

Could you send me a handful? Or a few?

Replied by e-mail. Air-snail mail tomorrow.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Winfield Hill]

John Woodgate wrote...

Winfield Hill wrote'
John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening
process. The chap who gave them to me thought that surface
contamination was the main cause, because some would 'clean up',
more or less permanently, if warmed up with a hot air gun.

Could you send me a handful? Or a few?

Replied by e-mail. Air-snail mail tomorrow.

Thanks, John, got the bag of BC109's earlier today; just
a quick look for now.

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V. That's interesting (!!),
so I'll take more measurements later when I find the time.

A quick look at a 2sd786 under the same conditions had an
e_n about 1.8nV at 5V, and showed no change at Vce = 1V.
Ditto for 10mA (where the noise is lower, under 1nV).

I didn't look at i_n, or at lower frequencies.

All of these noise measurements had lots of fluctuation,
which I need to examine later in more detail. There may
have been some RF interference in my lab (and the hp4470A
has only a half-shell shield).

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Jim Thompson]

On 25 Mar 2004 11:43:54 -0800, Winfield Hill
<Winfield_member@newsguy.com> wrote:

John Woodgate wrote...

Winfield Hill wrote'
John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening
process. The chap who gave them to me thought that surface
contamination was the main cause, because some would 'clean up',
more or less permanently, if warmed up with a hot air gun.

Could you send me a handful? Or a few?

Replied by e-mail. Air-snail mail tomorrow.

Thanks, John, got the bag of BC109's earlier today; just
a quick look for now.

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V.
[snip]

Hi Win,

That's been my experience in the lab with a variety of transistors;
although an attempt at showing that by simulation failed. (I don't
think most spice models characterize noise parameters very well... in
fact most everything I've seen uses the defaults.)

Please keep us posted on your lab results.

Back when I used to "roll-my-own" audio preamplifiers I always
designed the first stage for VCB=0.

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice480)460-2350 | |
| E-mail Address at Website Fax480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |

Throughout the history of this great country there have actually
been people of only two political persuasions: fighters and cowards.
WE MUST NOT LET THE LATTER PREVAIL IN THE NEXT ELECTION!

 

[John Woodgate]

I read in sci.electronics.design that Winfield Hill
<Winfield_member@newsguy.com> wrote (in <c3vcpq024u8@drn.newsguy.com&gt
about 'For Win Hill - Low noise at low Vce', on Thu, 25 Mar 2004:

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V. That's interesting (!!),
so I'll take more measurements later when I find the time.

Yes. It's a long time ago, but that's roughly what we found. We thought
that the problem was surface contamination, and the surface currents
were lower with low Vce. But it's only conjecture.

A quick look at a 2sd786 under the same conditions had an
e_n about 1.8nV at 5V, and showed no change at Vce = 1V.
Ditto for 10mA (where the noise is lower, under 1nV).

Yes, that's what the 'good' BC109s did. Pity: we could have used some
that went down to 0.8 nV/rt-Hz or so at Vce = 1 V.(;-)

I didn't look at i_n, or at lower frequencies.

All of these noise measurements had lots of fluctuation,
which I need to examine later in more detail.

We put that down to the surface currents, too. The ones we warmed up,
IIRC, showed somewhat less of this 'popcorn noise' even after several
days. The trouble is that it's difficult to get a meaningful measurement
of such a wild signal. We found that a peak-programme meter gave
reasonably repeatable results, but read far higher than an r.m.s. meter,
which didn't give repeatable results.
--
Regards, John Woodgate, OOO - Own Opinions Only.
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Winfield Hill]

John Woodgate wrote...

Winfield Hill wrote:

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V. That's interesting (!!),
so I'll take more measurements later when I find the time.

Yes. It's a long time ago, but that's roughly what we found. We
thought that the problem was surface contamination, and the surface
currents were lower with low Vce. But it's only conjecture.

A quick look at a 2sd786 under the same conditions had an
e_n about 1.8nV at 5V, and showed no change at Vce = 1V.
Ditto for 10mA (where the noise is lower, under 1nV).

Yes, that's what the 'good' BC109s did. Pity: we could have used
some that went down to 0.8 nV/rt-Hz or so at Vce = 1 V.(;-)

I started at 10mA, but they seemed pretty noisy, then I noticed
the BC109 noise specs and curves were all at 200uA, so I tested
at 300uA, the closest setting on my instrument (without using an
uncalibrated vernier). They were either more quiet down there,
or had by then become less noisy after a few minute's operation.

A reducing noise level with time is something I've often seen
with the hp4470A, but I've assumed this is the instument's pilot-
tone phase-locked-loop gain servo that's slowly settling down.
Slow 1-Hz bandwidth filters and all that, ya'know. Or something.

Clearly further exploration is called for. But the 5V vs 1V noise
effect seemed quite clear, as I went back and forth testing it.
It would have been nice to test Vce = Vbe or lower (Vcb negative),
but I'll have to add external instruments to the setup to do this.

I didn't look at i_n, or at lower frequencies.

All of these noise measurements had lots of fluctuation,
which I need to examine later in more detail.

We put that down to the surface currents, too. The ones we warmed
up, IIRC, showed somewhat less of this 'popcorn noise' even after
several days. The trouble is that it's difficult to get a meaningful
measurement of such a wild signal. We found that a peak-programme
meter gave reasonably repeatable results, but read far higher than
an r.m.s. meter, which didn't give repeatable results.

<sigh>

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Kevin Aylward]

Jim Thompson wrote:

On 25 Mar 2004 11:43:54 -0800, Winfield Hill
Winfield_member@newsguy.com> wrote:

John Woodgate wrote...

Winfield Hill wrote'
John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening
process. The chap who gave them to me thought that surface
contamination was the main cause, because some would 'clean up',
more or less permanently, if warmed up with a hot air gun.

Could you send me a handful? Or a few?

Replied by e-mail. Air-snail mail tomorrow.

Thanks, John, got the bag of BC109's earlier today; just
a quick look for now.

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V.
[snip]

Hi Win,

That's been my experience in the lab with a variety of transistors;
although an attempt at showing that by simulation failed. (I don't
think most spice models characterize noise parameters very well... in
fact most everything I've seen uses the defaults.)

Of course spice is not going to model any noise variation with Vce. This
is not a standard or understood effect. However, Spice models the
current and thermal noise no problem.

Best Regards,

Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[Kevin Aylward]

Winfield Hill wrote:

John Woodgate wrote...

Winfield Hill wrote'
John Woodgate wrote...

I have a collection of noisy BC109s, rejected by a screening
process. The chap who gave them to me thought that surface
contamination was the main cause, because some would 'clean up',
more or less permanently, if warmed up with a hot air gun.

Could you send me a handful? Or a few?

Replied by e-mail. Air-snail mail tomorrow.

Thanks, John, got the bag of BC109's earlier today; just
a quick look for now.

Two parts grabbed at random both have e_n about 2.7nV/rt-Hz
at 1kHz with 300uA bias at Vce = 5V, and they both improve
by about 15 to 20% for Vce = 1V. That's interesting (!!),
so I'll take more measurements later when I find the time.

A quick look at a 2sd786 under the same conditions had an
e_n about 1.8nV at 5V, and showed no change at Vce = 1V.
Ditto for 10mA (where the noise is lower, under 1nV).

I don't expect it to do much at Vce = 1V at all. I think you need to get
Vbc < 0. You want to eliminate the reverse bias on the bc junction.

Set it to a forward voltage of around 200mv, i.e. vce ~0.5V, but make
sure its not saturating. i.e. that the gain is still up.



Kevin Aylward
salesEXTRACT@anasoft.co.uk
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"quotes with no meaning, are meaningless" - Kevin Aylward.

 

[Roy McCammon]

Winfield Hill wrote:

Clearly further exploration is called for. But the 5V vs 1V noise
effect seemed quite clear, as I went back and forth testing it.
It would have been nice to test Vce = Vbe or lower (Vcb negative),
but I'll have to add external instruments to the setup to do this.

I don't know, but I conjecture that if it is shot noise, that
the "shots" are more energetic at the higher voltage.


--
local optimization seldom leads to global optimization

my e-mail address is: <my first name> <my last name> AT mmm DOT com