The bi-polar transistor at RF
 

Hi all,

Firstly, does anyone bother designing with Y-parameters *at all* these
days?

Then... (talking of the common-emitter configuration in this case)
The only variable according to the Ebers-Moll transistor model apart
from the device-specific "Is" which has any effect on Ic is the
potental difference applied across the B/E junction. The signal
voltage thus applied is loaded by the resistance of this diode. At
DC., the loading is at a maximum and the entire PD appears across it.
Right so far? As the applied signal voltage increases in frequency,
the feedback capacitance (B-C) and the B-E junction capacitance form
an AC bypass path across the B/E resistance above-mentioned. The two
capacitances acting in concert shunt more and more of the applied
signal voltage to ground, bypassing the emitter diode resistance,
lowering the device input impedance and resulting in less and less
applied Vbe across this diode and consequently less and less Ic output
swing?

What I'm getting at is that Ebers-Moll is still good at RF, *provided*
one allows for the bypassing of the emitter diode's resistance by the
combination of Cb and Ce. Correct?
And CJC and CJE are the relevant Spice model parameters?

Thanks,

p.
--

"What is now proved was once only imagin'd." - William Blake, 1793.

You didn't even mention base spreading resistance, a major
figure of merit for RF transistors. Newer designers have
considerably lower r-sub-bb-prime than legacy transistors.
Look up the hybrid-pi transistor model. SiGe technology's
claim to fame is low base spreading resistance.

Rick N6RK

"Paul Burridge" wrote in message
...
Hi all,

Firstly, does anyone bother designing with Y-parameters *at all* these
days?

Then... (talking of the common-emitter configuration in this case)
The only variable according to the Ebers-Moll transistor model apart
from the device-specific "Is" which has any effect on Ic is the
potental difference applied across the B/E junction. The signal
voltage thus applied is loaded by the resistance of this diode. At
DC., the loading is at a maximum and the entire PD appears across it.
Right so far? As the applied signal voltage increases in frequency,
the feedback capacitance (B-C) and the B-E junction capacitance form
an AC bypass path across the B/E resistance above-mentioned. The two
capacitances acting in concert shunt more and more of the applied
signal voltage to ground, bypassing the emitter diode resistance,
lowering the device input impedance and resulting in less and less
applied Vbe across this diode and consequently less and less Ic output
swing?

What I'm getting at is that Ebers-Moll is still good at RF, *provided*
one allows for the bypassing of the emitter diode's resistance by the
combination of Cb and Ce. Correct?
And CJC and CJE are the relevant Spice model parameters?

Thanks,

p.
--

"What is now proved was once only imagin'd." - William Blake, 1793.

In article 4hFLc.141721$JR4.5322@attbi_s54,
"Rick Karlquist N6RK" writes:
You didn't even mention base spreading resistance, a major
figure of merit for RF transistors. Newer designers have
considerably lower r-sub-bb-prime than legacy transistors.
Look up the hybrid-pi transistor model. SiGe technology's
claim to fame is low base spreading resistance.

What you say about rbb is true, but SiGe has other 'cool'
characteristics, some derived from low rbb, and some due
to other effects of the 'strained' silicon?... Ignoring
the rbb itself, SiGe tends also to have very low 1/f noise
and good LF noise in general. Also, it tends to have high
Beta (at LF.) So, where an RF transistor might tend to have
a Beta of 20-50, an SiGE part might be 100-300 or higher.

With the combo of the low rbb (and low 1/f noise), along with
the high Beta, the total amount of input current noise and
input voltage noise is damned low.

SiGe would make good oscillators (for less PM noise) and
of course, good preamps. This is one case where GaAs FETS
that are very fast, might be undesirable because of their
worse 1/F noise characteristics.

One big disadvantage of the typical SiGe transistors is
that their breakdown voltage is low. However, the tradeoff
of breakdown voltage is BETTER for a given frequency response
and Beta than a normal Si transistor.

The SiGe transistors are also not very expensive. A part that
works well with reasonably low distortion and reasonably low
noise figure at 600MHz would be significantly less than $1.00.
Unless the transistor is too fast for a given layout, SiGe
can be used at low frequencies (e.g. VHF) while still avoiding
the low frequency noise problems that are common from GaAs FETS
and even other fast BJTs.

John

Hi John,

Interesting. As a tubehead myself, I have not heard of these. Who mfrs
them? What's a typical part #?

Thanks :-)

--
Gregg t3h g33k
"Ratings are for transistors....tubes have guidelines"
http://geek.scorpiorising.ca

Gregg schrieb:
Hi John,

Interesting. As a tubehead myself, I have not heard of these. Who mfrs
them? What's a typical part #?

SiGe Transistors?

One example:

http://www.infineon.com/cmc_upload/d...669/bfp620.pdf

--
Cheers
Stefan

Paul Burridge wrote:

Hi all,

Firstly, does anyone bother designing with Y-parameters *at all* these
days?

Then... (talking of the common-emitter configuration in this case)
The only variable according to the Ebers-Moll transistor model apart
from the device-specific "Is" which has any effect on Ic is the
potental difference applied across the B/E junction. The signal
voltage thus applied is loaded by the resistance of this diode. At
DC., the loading is at a maximum and the entire PD appears across it.
Right so far? As the applied signal voltage increases in frequency,
the feedback capacitance (B-C) and the B-E junction capacitance form
an AC bypass path across the B/E resistance above-mentioned. The two
capacitances acting in concert shunt more and more of the applied
signal voltage to ground, bypassing the emitter diode resistance,
lowering the device input impedance and resulting in less and less
applied Vbe across this diode and consequently less and less Ic output
swing?

What I'm getting at is that Ebers-Moll is still good at RF, *provided*
one allows for the bypassing of the emitter diode's resistance by the
combination of Cb and Ce. Correct?
And CJC and CJE are the relevant Spice model parameters?

Thanks,

p.
--

"What is now proved was once only imagin'd." - William Blake, 1793.

At RF, the base spreading resistance can be large when compared with
the calculated emitter resistance; this makes a serious contribution to
input noise and the NF of the stage.
So the particular version of the model one uses can be rather poor in
determining real-life NF.
BTW, noise measurements at audio frequencies using different collector
currents can be used to determine the transistor's base spreading
resistance.
Once that is known, and the collector current used in the RF amplifier
(for determining Re), one can then calculate noise (or NF) and be rather
close to measured values!

Rick Karlquist N6RK wrote:

You didn't even mention base spreading resistance, a major
figure of merit for RF transistors. Newer designers have
considerably lower r-sub-bb-prime than legacy transistors.
Look up the hybrid-pi transistor model. SiGe technology's
claim to fame is low base spreading resistance.

Rick N6RK

"Paul Burridge" wrote in message
...
Hi all,

Firstly, does anyone bother designing with Y-parameters *at all* these
days?

Then... (talking of the common-emitter configuration in this case)
The only variable according to the Ebers-Moll transistor model apart
from the device-specific "Is" which has any effect on Ic is the
potental difference applied across the B/E junction. The signal
voltage thus applied is loaded by the resistance of this diode. At
DC., the loading is at a maximum and the entire PD appears across it.
Right so far? As the applied signal voltage increases in frequency,
the feedback capacitance (B-C) and the B-E junction capacitance form
an AC bypass path across the B/E resistance above-mentioned. The two
capacitances acting in concert shunt more and more of the applied
signal voltage to ground, bypassing the emitter diode resistance,
lowering the device input impedance and resulting in less and less
applied Vbe across this diode and consequently less and less Ic output
swing?

What I'm getting at is that Ebers-Moll is still good at RF, *provided*
one allows for the bypassing of the emitter diode's resistance by the
combination of Cb and Ce. Correct?
And CJC and CJE are the relevant Spice model parameters?

Thanks,

p.
--

"What is now proved was once only imagin'd." - William Blake, 1793.

Beat me to it; see my comment.

John S. Dyson wrote...

The SiGe transistors are also not very expensive. A part that
works well with reasonably low distortion and reasonably low
noise figure at 600MHz would be significantly less than $1.00.

Examples of low-cost high-performance (30GHz at 10mA) SiGe
transistors would be Infineon's BFP620 (82 cents at DigiKey)
http://www.infineon.com/cgi/ecrm.dll....jsp?oid=26182
and Philips' BFU510 and BFU540
http://www.semiconductors.philips.com/pip/BFU510.html
http://www.semiconductors.philips.com/pip/BFU540.html

The Philips transistors look good, but I don't know where to
get them. Mouser stocks a set of CEL's nice SiGe transistors,
http://www.mouser.com/?handler=produ...riteria =SiGe

Thanks,
- Win

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

On Thu, 22 Jul 2004 10:32:55 GMT, Robert Baer
wrote:

At RF, the base spreading resistance can be large when compared with
the calculated emitter resistance; this makes a serious contribution to
input noise and the NF of the stage.

Sorry guys, I did *mean* to include BSR in series with the B/E
junction resistance, so any reference I made to this junction
resistance should be taken to mean the total of the two together.

So the particular version of the model one uses can be rather poor in
determining real-life NF.

NF isn't a consideration in this instance; please ignore it.
And I am well aware of the pi-model. I just want to know if I have it
right that Ebers-Moll will work accurately into UHF provided one
allows for the feedback capacitance and emitter junction capacitance
shunting the input signal around BSR+EBR and thereby reducing the
signal voltage developed across them. Do I have this right?

Thanks,

Paul
--

"What is now proved was once only imagin'd." - William Blake, 1793.

On Thu, 22 Jul 2004 12:24:54 +0100, Paul Burridge
wrote:

NF isn't a consideration in this instance; please ignore it.
And I am well aware of the pi-model. I just want to know if I have it
right that Ebers-Moll will work accurately into UHF provided one
allows for the feedback capacitance and emitter junction capacitance
shunting the input signal around [EBR alone] and thereby reducing the
signal voltage developed across it. Do I have this right?

Sorry! Corrected above. IOW: whilst the emitter diode resistance is
bypassed at RF by these two capacitances, the base spreading
resistance *isn't* - apart from that, the rest of the post is now
correct, yes? IOW, as the signal frequency increases, the BSR becomes
the dominant component of the device's input impedance... Phew!
Unless of course, someone knows otherwise...
--

"What is now proved was once only imagin'd." - William Blake, 1793.

I read in sci.electronics.design that Paul Burridge
wrote (in 8afvf05guvnvmjgqtafgau2d2li3ckn
) about 'The bi-polar transistor at RF', on Thu, 22 Jul
2004:
IOW, as the signal frequency increases, the BSR becomes the dominant
component of the device's input impedance... Phew! Unless of course,
someone knows otherwise...
Emitter lead inductance?
--
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

On Thu, 22 Jul 2004 14:41:45 +0100, John Woodgate
wrote:

I read in sci.electronics.design that Paul Burridge
wrote (in 8afvf05guvnvmjgqtafgau2d2li3ckn
) about 'The bi-polar transistor at RF', on Thu, 22 Jul
2004:
IOW, as the signal frequency increases, the BSR becomes the dominant
component of the device's input impedance... Phew! Unless of course,
someone knows otherwise...
Emitter lead inductance?

Er, yes, but I'm only interested in the *internal* characteristics of
the device here, so even the bonding wires' inductance isn't an issue.
Thanks for giving me the chance to clarify, though.
--

"What is now proved was once only imagin'd." - William Blake, 1793.

I read in sci.electronics.design that Paul Burridge
wrote (in 61kvf05660gk62pnm7o3bthepkujnep
) about 'The bi-polar transistor at RF', on Thu, 22 Jul
2004:
Er, yes, but I'm only interested in the *internal* characteristics of
the device here, so even the bonding wires' inductance isn't an issue.
Thanks for giving me the chance to clarify, though.

A bit of the emitter lead is inside the encapsulation, and a bit is on
the die.
--
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

On Thu, 22 Jul 2004 03:09:10 +0000 (UTC), (John S.
Dyson) wrote:

One big disadvantage of the typical SiGe transistors is
that their breakdown voltage is low.

Unless the transistor is too fast for a given layout, SiGe
can be used at low frequencies (e.g. VHF) while still avoiding
the low frequency noise problems that are common from GaAs FETS
and even other fast BJTs.

While the IP3 and compression figures may good to comparable devices
working in the GHz bands, the huge gain with the low Vce (typically
less than 2.3 V) will damage the input IP3 values quite quickly at
VHF. This can be a problem in the VHF and lower UHF bands, in which
signal levels can be quite high and multiple strong signals may pass
the front end selectivity.

For VHF applications, a high current SiGe device operating at
impedance levels well below 50 ohms would give good IP3 figures, but
apparently the noise figure increases quite rapidly with high
collector currents.

Paul OH3LWR

On Thu, 22 Jul 2004 15:46:25 +0100, John Woodgate
wrote:

I read in sci.electronics.design that Paul Burridge
wrote (in 61kvf05660gk62pnm7o3bthepkujnep
) about 'The bi-polar transistor at RF', on Thu, 22 Jul
2004:
Er, yes, but I'm only interested in the *internal* characteristics of
the device here, so even the bonding wires' inductance isn't an issue.
Thanks for giving me the chance to clarify, though.

A bit of the emitter lead is inside the encapsulation, and a bit is on
the die.

Thank you, John, that gives me another chance to re-state the question
more succinctly. I'm not concerned with inductances here at all,
though.
Does the Ebers-Moll equation hold good at UHF+, provided the value one
inserts for Vbe is adjusted to account for the loss of signal voltage
the B/E junction will suffer as much of it (the applied signal
voltage) is shunted around it via the device's internal capacitances?

There! I think I've nailed it this time!

--

"What is now proved was once only imagin'd." - William Blake, 1793.

In article ,
Paul Keinanen writes:
On Thu, 22 Jul 2004 03:09:10 +0000 (UTC), (John S.
Dyson) wrote:

One big disadvantage of the typical SiGe transistors is
that their breakdown voltage is low.

Unless the transistor is too fast for a given layout, SiGe
can be used at low frequencies (e.g. VHF) while still avoiding
the low frequency noise problems that are common from GaAs FETS
and even other fast BJTs.

While the IP3 and compression figures may good to comparable devices
working in the GHz bands, the huge gain with the low Vce (typically
less than 2.3 V) will damage the input IP3 values quite quickly at
VHF. This can be a problem in the VHF and lower UHF bands, in which
signal levels can be quite high and multiple strong signals may pass
the front end selectivity.

"Noiseless feedback' is very helpful to mitigate the excess amounts
of gain, while pushing the return loss match (the impedance match)
closer to the noise match (the input impedance where the noise is lowest.)
A simple emitter (source) inductor and a little bit of parallel, noisy
feedback can be used to tame some of the interesting UHF+ components.
(The emitter (source) inductor is a case where a small amount of
inductance is much better than too much, because instability can ensue
with too much series feedback (too large an emitter (source) inductor).)

I do agree with your implication that a high current device can be helpful
at low frequencies, but some SiGe components do seem to maintain
reasonable noise performance at high currents.

In any case, the SiGe components do give the GaAs type components a
run for their money. In some cases, the SiGe components are actually
better. The PHEMTs (HP 54143) are also an interesting variation on
the 'fet' theme, which helps to mitigate some of the problems WRT
GaAs. For example, a PHEMT can provide a good noise match at 50ohms,
a transconductance of almost 1MHO at 60ma, and reasonable IP3 (38dBm isn't
impossible.) Feedback can be especially helpful with the high
transconductance FETs.

John

On Thu, 22 Jul 2004 16:01:02 +0000 (UTC), (John S.
Dyson) wrote:

"Noiseless feedback' is very helpful to mitigate the excess amounts
of gain, while pushing the return loss match (the impedance match)
closer to the noise match (the input impedance where the noise is lowest.)
A simple emitter (source) inductor and a little bit of parallel, noisy
feedback can be used to tame some of the interesting UHF+ components.
(The emitter (source) inductor is a case where a small amount of
inductance is much better than too much, because instability can ensue
with too much series feedback (too large an emitter (source) inductor).)

Has anybody actually used noiseless feedback with these devices at
VHF? After all, the fT of these SiGe transistors are in the 30 GHz+
range, so I would guess that the parasitics would mess the situation
quite badly, especially when using feedback components.

How critical is the layout compared to for instance MAR-x series MMICs
(that are essentially darlingtons) ? Can these SiGe transistors used
with dual sided boards and microstrips or do they require multilayer
boards and full striplines in order to use feedback ?

Paul OH3LWR