the ultimate PNP transistor

[[email protected]]

Hi,

I am looking for a readily available PNP transistor in a through-hole
package with as small a collector-base ("Miller") capacitance as
possible. It is to be used for moderately fast saturated switching
(t-on
and t-off of a few 10ns) at low power (Ic about 1mA). Maximum collector
voltage (9V battery) is no problem.

I have tried the BF324 and BF450 (both Philips) with reasonable
results.
The limiting factor here indeed appears to be the collector-base
capacitance: measured zero-bias values are ~3pF and ~2pF, respectively.
Fast PNP switching transistors like the 700MHz 2N5771 are not
automatically better suited.

So, what is the ultimate PNP transistor for this type of application?
Is
something like the 5GHz BFT52 (Ccb ~ 1pF according to the Philips Spice
model) still available in a through-hole package? Operation with E and
C
interchanged may also be considered if the reverse beta exceeds five.

TIA,

Martin.

 

[Mark]

why not change the design of the base drive so it is not saturated,,,,
it will switch off much faster... or apply a positve voltage (current
actually) to turn it off faster...


Mark

 

[Robert Baer]

why not change the design of the base drive so it is not saturated,,,,
it will switch off much faster... or apply a positve voltage (current
actually) to turn it off faster...


Mark

Yet another way to pull charges out of the base is to use a resistor
from input drive to base of PNP, the drive to the base of an inverted
NPN: collector to ground and emitter to base of PNP.
No additinal supplies needed.

 

[Frithiof Andreas Jensen]

Hi,

I am looking for a readily available PNP transistor in a through-hole
package with as small a collector-base ("Miller") capacitance as
possible.

Use anti-saturation diodes or - if that does not work - two transistors in a
Cascode connection.

 

[[email protected]]

Hi,

I am looking for a readily available PNP transistor in a through-hole
package with as small a collector-base ("Miller") capacitance as
possible. It is to be used for moderately fast saturated switching
(t-on and t-off of a few 10ns) at low power (Ic about 1mA). Maximum
collector voltage (9V battery) is no problem.

I have tried the BF324 and BF450 (both Philips) with reasonable
results.
The limiting factor here indeed appears to be the collector-base
capacitance: measured zero-bias values are ~3pF and ~2pF, respectively.
Fast PNP switching transistors like the 700MHz 2N5771 are not
automatically better suited.

So, what is the ultimate PNP transistor for this type of application?
Is something like the 5GHz BFT52 (Ccb ~ 1pF according to the Philips
Spice model) still available in a through-hole package? Operation with
E and C interchanged may also be considered if the reverse beta exceeds
five.

The designation should have read BFT92. An obsolete TO-72 version was
called BFQ52.

 

[Robert Baer]

Use anti-saturation diodes or - if that does not work - two transistors in a
Cascode connection.

Anti-saturation diodes would *add* capacitance..
But cascode operation would definitely kill the miller effect (on
both transistors).

 

[Frithiof Andreas Jensen]

Anti-saturation diodes would *add* capacitance..

The "miller capacitance" the O.P. have problems with is (roughly): the
charge stored in the base-collector region + charge in the physical
capacitance of the base junction and then the whole thing is multiplied by
the voltage swing of the collector.

So adding a little more physical capacitance might bring an overall
improvement in performance by reducing the base-collector charge which is
penalised extra by the collector voltage swing.

But cascode operation would definitely kill the miller effect (on
both transistors).

By the elimination of the collector voltage change in the driver
transistor - the top transistor still gets the full whack BUT the base is
fed from a low impedance source so one does not see it. The stored charge
comes out through the base when the top transistor switches off - for fast
switching it is neccessary to decouple the base of the top transistor well.

 

[Geir Klemetsen]

Hi,

I am looking for a readily available PNP transistor in a through-hole
package with as small a collector-base ("Miller") capacitance as
possible. It is to be used for moderately fast saturated switching
(t-on
and t-off of a few 10ns) at low power (Ic about 1mA). Maximum collector
voltage (9V battery) is no problem.

I have tried the BF324 and BF450 (both Philips) with reasonable
results.
The limiting factor here indeed appears to be the collector-base
capacitance: measured zero-bias values are ~3pF and ~2pF, respectively.
Fast PNP switching transistors like the 700MHz 2N5771 are not
automatically better suited.

So, what is the ultimate PNP transistor for this type of application?
Is
something like the 5GHz BFT52 (Ccb ~ 1pF according to the Philips Spice
model) still available in a through-hole package? Operation with E and
C
interchanged may also be considered if the reverse beta exceeds five.

What about this little design-trick:


o
|| |
.----||----. |
| || | |
| | |
| ___ | |/
o-------o--|___|---o----|
| |>
| |
| |
- |
^ |
| |
| |
=== ===
GND GND
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

 

[[email protected]]

As I've mentioned somewhere else already, for BFT52 read BFT92.

What about this little design-trick:


o
|| |
.----||----. |
| || | |
| | |
| ___ | |/
o-------o--|___|---o----|
| |>
| |
| |
- |
^ |
| |
| |
=== ===
GND GND
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

My drive is a current source with little head room. So I'm just feeding

the current into the base of the PNP transistor acting as a "digitizing

current mirror". The mark-space ratio is determined by a resistor from
base to collector. With a BF324 (zero-bias Cbc ~ 3pF, beta ~ 50) at 1mA

collector current, this works only up to a few MHz. Easy to simulate
with
Spice - but be sure to get the CJC right!

VCC o-------------o------o
9V | |
|¯| |
| | |
Rb |_| |
| |<
current in >------o----| BF324, BF450, BFQ52, etc
|\
|
o------> voltage out
| CMOS load ~ 5pF
|¯|
| |
|_| Rc
|
0V o---------------------o


Martin.

 

[Fred Bartoli]

<[email protected]> a écrit dans le message de


As I've mentioned somewhere else already, for BFT52 read BFT92.

What about this little design-trick:


o
|| |
.----||----. |
| || | |
| | |
| ___ | |/
o-------o--|___|---o----|
| |>
| |
| |
- |
^ |
| |
| |
=== ===
GND GND
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

My drive is a current source with little head room. So I'm just feeding

the current into the base of the PNP transistor acting as a "digitizing

current mirror". The mark-space ratio is determined by a resistor from
base to collector. With a BF324 (zero-bias Cbc ~ 3pF, beta ~ 50) at 1mA

collector current, this works only up to a few MHz. Easy to simulate
with
Spice - but be sure to get the CJC right!

VCC o-------------o------o
9V | |
|¯| |
| | |
Rb |_| |
| |<
current in >------o----| BF324, BF450, BFQ52, etc
|\
|
o------> voltage out
| CMOS load ~ 5pF
|¯|
| |
|_| Rc
|
0V o---------------------o


Martin.


Martin,
if your source is a current source then Cbc is indeed the limiting factor in
this circuit.

What are your parameters? From your 9V/1mA conditions I expect Rc to be 9K
(probably 10K or 8.2K). This is already 50ns with your 5pF load parasitics,
and doesn't include any Miller effect at all.

With a current source base driving you'll have to go the cascode route.

Before going any further, what are your circuit parameters (base current
source, power supply allowed current, expected output 10-90% or whatever
rise time, expected output swing)?

 

[[email protected]]

<[email protected]> a écrit dans le message de



As I've mentioned somewhere else already, for BFT52 read BFT92.


My drive is a current source with little head room. So I'm just feeding
the current into the base of the PNP transistor acting as a "digitizing
current mirror". The mark-space ratio is determined by a resistor from
base to collector. With a BF324 (zero-bias Cbc ~ 3pF, beta ~ 50) at 1mA

^^^^^^^^^ emitter, of course!

collector current, this works only up to a few MHz. Easy to simulate
with
Spice - but be sure to get the CJC right!

VCC o-------------o------o
9V | |
|¯| |
| | |
Rb |_| |
| |<
current in >------o----| BF324, BF450, BFQ52, etc
|\
|
o------> voltage out
| CMOS load ~ 5pF
|¯|
| |
|_| Rc
|
0V o---------------------o


Martin.


Martin,
if your source is a current source then Cbc is indeed the limiting factorin
this circuit.

What are your parameters? From your 9V/1mA conditions I expect Rc to be 9K
(probably 10K or 8.2K). This is already 50ns with your 5pF load parasitics,
and doesn't include any Miller effect at all.

With a current source base driving you'll have to go the cascode route.

Before going any further, what are your circuit parameters (base current
source, power supply allowed current, expected output 10-90% or whatever
rise time, expected output swing)?

My current working circuitry corresponds to this Spice code:

*
* PNP digitizing current mirror
*
v0 1 0 9 ! 9V battery
*
ii 2 0 sin(0.6m 0.4m 4Meg) ! 0.2 to 1.0mA sinusoidal modulation
*
qp 3 2 1 BFT92 ! CJC as small as possible (~1p)
rb 2 1 1.1k ! adjust value if necessary
rc 3 0 4.7k ! 1mA average
co 3 0 5p ! CMOS gate input (ST: 5p typ, 7.5p max)
*
* Philips model, Version: 2.0, Date: July 1992
*
..model BFT92 PNP(IS=4.37563E-16 BF=33.5815 NF=1.00972 VAF=23.3946
+ IKF=9.95381E-2 ISE=8.70539E-14 NE=1.94395 BR=4.94721 NR=1.00254
+ VAR=3.90385 IKR=5.28157E-3 ISC=3.58864E-14 NC=1.39333 RB=5 IRB=1E-6
+ RBM=5 RE=1 RC=10 EG=1.11 XTI=3 CJE=7.46659E-13 VJE=0.6 MJE=0.356829
+ TF=1.74921E-11 XTF=1.35455 VTF=0.155654 ITF=1E-3 PTF=45
+ CJC=9.37103E-13 VJC=0.396455 MJC=0.199949)
*

Although I have included the model of the 5GHz BFT92 (SOT23 package),
my
present circuit uses the 450MHz BF324 (TO-92 package). At 4MHz, the
signal rise (as observed with a probe capacitance <1pF) takes about
40ns
(10% to 90%), and the falling curve shows a T1/2 of about 35ns. This is
somewhat better than my simulation for the Philips BF324 model, and
somewhat worse than for the BFT92 model above.

The purpose of my post was this: That people would name PNP transistors
in a through-hole package with a Miller capacitance as low as that of
the BFT92/BFQ52: below 1pF at zero bias. Surely there must be some?
ON-Semi perhaps?

But I don't want to leave your questions unanswered.
- base current source: see model above.
- power-supply allowed current: average as low as pos (9V NiMH block),
supporting caps up to a few cm^3 no problem.
- expected output 10-90% or whatever rise time: sufficient to reliably
drive CMOS when operating frequency is raised to 8MHz. The value of Rc
may be lowered if necessary.
- expected output swing: adequate for reliable CMOS drive. Each MOSFET
should be turned off for part of a cycle.

Martin.

 

[Winfield Hill]

[email protected] wrote...

... I have tried the BF324 and BF450 (both Philips) with
reasonable results.
The limiting factor here indeed appears to be the collector-base
capacitance: measured zero-bias values are ~3pF and ~2pF, respectively.
Fast PNP switching transistors like the 700MHz 2N5771 are not
automatically better suited.

That's no surprise, with the 2n5771's relatively high capacitances,
3pF, compared a much lower 0.3 and 0.55pF at 5-10V reverse bias.

So, what is the ultimate PNP transistor for this type of application?
Is something like the 5GHz BFT52 (Ccb ~ 1pF according to the Philips
Spice model) still available in a through-hole package?

About 0.7pF at sensible reverse bias according to the datasheets,
just a bit higher than you've been using. But there are plenty of
lower-capacitance smaller-die-size choices if that's important.

Operation with E and C interchanged may also be considered if the
reverse beta exceeds five.

Sheesh!

 

[Jim Thompson]

[email protected] wrote...

That's no surprise, with the 2n5771's relatively high capacitances,
3pF, compared a much lower 0.3 and 0.55pF at 5-10V reverse bias.


About 0.7pF at sensible reverse bias according to the datasheets,
just a bit higher than you've been using. But there are plenty of
lower-capacitance smaller-die-size choices if that's important.


Sheesh!

DOUBLE-SHEEEEESH! Such amateurs ;-)

...Jim Thompson

 

[Winfield Hill]

Jim Thompson wrote...

DOUBLE-SHEEEEESH! Such amateurs ;-)

What a minute. First I invent Sheesh!, then you go and invent
DOUBLE-SHEEEEESH!? Sheesh!

 

[Jim Thompson]

Jim Thompson wrote...

What a minute. First I invent Sheesh!, then you go and invent
DOUBLE-SHEEEEESH!? Sheesh!

Win, Did you really invent "she..esh"?

I've had it in my spelling checker... many different lengths of
e-strings, for a very very long time ;-)

Same with "Bwahahaha..haha", no matter the number of rat-a-tats, it
always passes spelling check ;-)

...Jim Thompson

 

[Winfield Hill]

Jim Thompson wrote...

Win, Did you really invent "she..esh"?

Of course not. IIRC, my dad used to use the term, so it must
go way back. I imagine it's a polite way of saying shit.

 

[Jim Thompson]

Jim Thompson wrote...

Of course not. IIRC, my dad used to use the term, so it must
go way back. I imagine it's a polite way of saying shit.

Come to think of it, I _do_ remember my mother saying "sheeesh" ;-)

...Jim Thompson

 

[Spehro Pefhany]

Jim Thompson wrote...

Of course not. IIRC, my dad used to use the term, so it must
go way back. I imagine it's a polite way of saying shit.

Cripes, no. The American Heritage dictionary says it's an alteration
of "Jesus".



Best regards,
Spehro Pefhany

 

[Winfield Hill]

Spehro Pefhany wrote...

Cripes, no. The American Heritage dictionary says it's an
alteration of "Jesus".

What, a religious swear word, taking the Lord's name in vain?
Shit!

 

[Frank Bemelman]

Jim Thompson wrote...

Of course not. IIRC, my dad used to use the term, so it must
go way back. I imagine it's a polite way of saying shit.

I actually thought it was another form of "Jesus". Like "Holy Mackarel"
or something.