Maker Pro
Maker Pro

transistors for push pull medium power ?

P

Pinchy

Jan 1, 1970
0
I am building a waveformgenerator. To increase the output current I
intend to use a push pull class amplifier (NPN - PNP) in the feedback
loop of the last opamp. Hence it has to go up to 100 kHz I am looking
for the most recommended transistorset to build this stage. It should
be able to provide up to 15 V / 50 ohm (output resistance) = 300 mA in
order to get an automatic shortcut protection circuit (50 ohm acting
as a load)

The combination BC457 - BC557 (100 mA) is not suitable. The other
BD139 - BD140 (5A) is giving trouble in high frequency range as the
junctions of these are getting pretty big. The last option is a BC141
- BC161 (1A) but those are getting pretty old. Has anyone a better
idea (European parts) ? Is it usefull to bypass the collector - base
junction with Schottky diodes (as in digital 74LS, ALS, AS, ...) to
inprove the discharge speed of the collector junction ?

Thanks

Ing. Geert
 
K

Kevin Aylward

Jan 1, 1970
0
Pinchy said:
I am building a waveformgenerator. To increase the output current I
intend to use a push pull class amplifier (NPN - PNP) in the feedback
loop of the last opamp. Hence it has to go up to 100 kHz I am looking
for the most recommended transistorset to build this stage. It should
be able to provide up to 15 V / 50 ohm (output resistance) = 300 mA in
order to get an automatic shortcut protection circuit (50 ohm acting
as a load)

The combination BC457 - BC557 (100 mA) is not suitable. The other
BD139 - BD140 (5A) is giving trouble in high frequency range as the
junctions of these are getting pretty big. The last option is a BC141
- BC161 (1A) but those are getting pretty old. Has anyone a better
idea (European parts) ?

Have a look at mosfets.

Is it usefull to bypass the collector - base
junction with Schottky diodes (as in digital 74LS, ALS, AS, ...) to
inprove the discharge speed of the collector junction ?

Not in a linear amplifier.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
T

Tilmann Reh

Jan 1, 1970
0
Pinchy said:
I am building a waveformgenerator. To increase the output current I
intend to use a push pull class amplifier (NPN - PNP) in the feedback
loop of the last opamp. Hence it has to go up to 100 kHz I am looking
for the most recommended transistorset to build this stage. It should
be able to provide up to 15 V / 50 ohm (output resistance) = 300 mA in
order to get an automatic shortcut protection circuit (50 ohm acting
as a load)

The combination BC547 - BC557 (100 mA) is not suitable.

Correct. Too small.
The other
BD139 - BD140 (5A) is giving trouble in high frequency range as the
junctions of these are getting pretty big.
Exactly.

The last option is a BC141
- BC161 (1A) but those are getting pretty old. Has anyone a better
idea (European parts)?

There are much more modern parts than these really old fashioned types.
In this application, you'd also have to look at the power dissipation,
since at heavy loads, one transistor may have to dissipate 15V * 300mA
= 4.5W. Too much for anything without heat sink.

Look at BD 237/238 and BD 437/438. They are pretty standard and rather
fast for their power ratings. AFAIR, especially the 437/438, which also
have good gains (which is handy if you want to drive them from the
OPamp directly).

I would *not* suggest MOSFETs due to their large G/S capacity, and since
you have to drive their gates far beyond the supply voltage if you use
them in source follower circuit.
Is it usefull to bypass the collector - base
junction with Schottky diodes (as in digital 74LS, ALS, AS, ...) to
inprove the discharge speed of the collector junction ?

No, this is only to prevent BJTs from going into saturation. In your
emitter-follower stages, this will never happen anyway.
 
J

john jardine

Jan 1, 1970
0
Pinchy said:
I am building a waveformgenerator. To increase the output current I
intend to use a push pull class amplifier (NPN - PNP) in the feedback
loop of the last opamp. Hence it has to go up to 100 kHz I am looking
for the most recommended transistorset to build this stage. It should
be able to provide up to 15 V / 50 ohm (output resistance) = 300 mA in
order to get an automatic shortcut protection circuit (50 ohm acting
as a load)

The combination BC457 - BC557 (100 mA) is not suitable. The other
BD139 - BD140 (5A) is giving trouble in high frequency range as the
junctions of these are getting pretty big. The last option is a BC141
- BC161 (1A) but those are getting pretty old. Has anyone a better
idea (European parts) ? Is it usefull to bypass the collector - base
junction with Schottky diodes (as in digital 74LS, ALS, AS, ...) to
inprove the discharge speed of the collector junction ?

Thanks

Ing. Geert

Had a similar problem recently. Needed 40Vpp into 50ohms over a DC-10MHz
bandwidth. Trouble in finding suitable transistors so went full circle and
sorted it by feeding 6 off BC546-BC556 pairs in parallel. As the drive amp
also used BC546-BC556's the total cost ended up very low :).
There's nothing in there that needs 'speeding up' or saturation avoidance,
just focus on swinging sufficient drive current to allow the maximum slew
rate to be met consistent with the accumulated Collector-Base capacitances.
regards
john
 
K

Kevin Aylward

Jan 1, 1970
0
Tilmann said:
Correct. Too small.


There are much more modern parts than these really old fashioned
types. In this application, you'd also have to look at the power
dissipation, since at heavy loads, one transistor may have to
dissipate 15V * 300mA = 4.5W. Too much for anything without heat sink.

Look at BD 237/238 and BD 437/438. They are pretty standard and rather
fast for their power ratings. AFAIR, especially the 437/438, which
also have good gains (which is handy if you want to drive them from
the OPamp directly).

I would *not* suggest MOSFETs due to their large G/S capacity,

This is misleading. A bipolar transister with an ft of say, 200Mhz, at
300ma has a Cin of around 10nf, to wit.

Cin=40.IC/2.pi.ft

A small mosfet will be much, much lower than this.
and since you have to drive their gates far beyond the supply voltage
if
you use them in source follower circuit.

Yes, but you can use a capacitive bootstrap on both sides of the driver.


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

Pooh Bear

Jan 1, 1970
0
Pinchy said:
I am building a waveformgenerator. To increase the output current I
intend to use a push pull class amplifier (NPN - PNP) in the feedback
loop of the last opamp. Hence it has to go up to 100 kHz I am looking
for the most recommended transistorset to build this stage. It should
be able to provide up to 15 V / 50 ohm (output resistance) = 300 mA in
order to get an automatic shortcut protection circuit (50 ohm acting
as a load)

The combination BC457 - BC557 (100 mA) is not suitable. The other
BD139 - BD140 (5A) is giving trouble in high frequency range as the
junctions of these are getting pretty big. The last option is a BC141
- BC161 (1A) but those are getting pretty old. Has anyone a better
idea (European parts) ? Is it usefull to bypass the collector - base
junction with Schottky diodes (as in digital 74LS, ALS, AS, ...) to
inprove the discharge speed of the collector junction ?

What's the special requirement for 'pro-electron' designation type devices
exclusively ? Nothing really good that I can think of.

Scared to use anything that starts 2N ? MJ even !


Graham
 
R

Rich Grise

Jan 1, 1970
0
What's the special requirement for 'pro-electron' designation type devices
exclusively ? Nothing really good that I can think of.

Scared to use anything that starts 2N ? MJ even !

Probably availability of Right- or Left-pondian parts.
:)

Cheers!
Rich
 
P

Pooh Bear

Jan 1, 1970
0
Rich said:
Probably availability of Right- or Left-pondian parts.
:)

I suspect from the grin that you know that's a non-issue. I'm in the UK and only
use a couple of BC parts regularly in my designs. They happen to be widely used
worldwide.

The OP may be less well versed in component availability I suspect.

Graham
 
T

Tilmann Reh

Jan 1, 1970
0
Kevin said:
This is misleading. A bipolar transister with an ft of say, 200Mhz, at
300ma has a Cin of around 10nf, to wit.

Not in emitter follower circuit.
And, BTW, the 300 mA will flow only in case of overload.
A small mosfet will be much, much lower than this.

A small MOSFET won't be able to handle the power dissipation.
A larger MOSFET will have large capacities, which must be driven
by the OPamp or separate driver stages.
Yes, but you can use a capacitive bootstrap on both sides of the driver.

Driven by the OPamp directly?

Using rather modern BJTs sounds like the more effective answer
to the OPs question.
 
J

john jardine

Jan 1, 1970
0
Pooh Bear said:
I suspect from the grin that you know that's a non-issue. I'm in the UK and only
use a couple of BC parts regularly in my designs. They happen to be widely used
worldwide.

The OP may be less well versed in component availability I suspect.

Graham

In UK, for prototyping and development purposes, most people will select
from the RS or Farnell catalogues. If you look at (say) the Farnell "GP amps
and switches" transistor pages it's 95% BC... and ZTX... numbers. So for
simplicity and stock availability, these'll be bought in the first instance.
There's no pro-electron bias, it's just a function of whichever supplier is
giving Farnell/RS the best purchasing deals.
For volume production of course it's only the price that matters and each
device can be looked at for possible re-spec using worldwide sources.
regards
john
 
K

Kevin Aylward

Jan 1, 1970
0
Tilmann said:
Not in emitter follower circuit.

Ho hummm...

Sure, one can *sometimes* take advantage that Can't = Cin/(1+gm.Zl), but
again, this is misleading when switching as *fast* as possible. The
reality is that the load capacitance is large, causing essentially, a
s/c for the initial transient. That is ZL is, essentially, zero.
Whatever cin is directly, is what you need to drive if you trying for
the ultimate in speed. Of course, I don't know what the posters
requirements are. This is only a suggestion.
And, BTW, the 300 mA will flow only in case of overload.

I have no idea what the circuit is. I am only making suggestions in
general.
A small MOSFET won't be able to handle the power dissipation.
A larger MOSFET will have large capacities, which must be driven
by the OPamp or separate driver stages.

Ho hummm...A "large" mosfet able to take the power will usually still
have much lower capacitances than a bipolar. It why the standard device
in SMPS are mosfets, not bipolar. You can switch them an order of
magnitude faster. End of story.
Driven by the OPamp directly?

No idea what you mean here. One has output capacitors feeding the drive
supply rails in a suitable manner.
Using rather modern BJTs sounds like the more effective answer
to the OPs question.

Maybe, maybe not.

Well, I am currently playing around with simulations of an audio amp
using state of the art bipolar (ft=50Mhz, 15A, 230V, 150W) and its a no
contest. Mosfets switch much faster. I can get 1000V/us (500W @8ohms)
with mosfets, only about 100V/us with bipolar.

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

Klaus Vestergaard Kragelund

Jan 1, 1970
0
Kevin Aylward said:
Ho hummm...
Snip

Ho hummm...A "large" mosfet able to take the power will usually still
have much lower capacitances than a bipolar. It why the standard device
in SMPS are mosfets, not bipolar. You can switch them an order of
magnitude faster. End of story.

Not according to my oppinion. The reason MOSFETs are used in switch-mode
supplies are more that they are easy to drive (no gate current). Ofcourse
the low capacitance is a second order effect

Cheers

Klaus
 
K

Kevin Aylward

Jan 1, 1970
0
Klaus said:
Not according to my oppinion.

Which apparently, is as useful as your spelling.
The reason MOSFETs are used in
switch-mode supplies are more that they are easy to drive

One factor, but not the fundamental reason for their use.
(no gate
current). Ofcourse the low capacitance is a second order effect

You obviously don't understand what you have just wrote. They have low
gate current *because* they have low capacitances. Yeah, sure, the lack
of a static current is useful, but not much if you have to drive them
with 10 times the static current in pulses to switch them on an off.

The fundamental reason mosfets are used are that they are much faster.
Its that simple. Your trying to teach your grandpa to suck eggs dude:)

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

Pinchy

Jan 1, 1970
0
Pooh Bear said:
What's the special requirement for 'pro-electron' designation type devices
exclusively ? Nothing really good that I can think of.

Scared to use anything that starts 2N ? MJ even !


Graham

OK everyone, thanks a lot for help. I'll work out a couple of things
and will see what is obtaining in the best result.

Geert
 
B

Ban

Jan 1, 1970
0
Kevin said:
Well, I am currently playing around with simulations of an audio amp
using state of the art bipolar (ft=50Mhz, 15A, 230V, 150W) and its a
no contest. Mosfets switch much faster. I can get 1000V/us (500W
@8ohms) with mosfets, only about 100V/us with bipolar.
Well, with 500W into 8R you will have a +/-90V peak swing, which corresponds
to a slew rate of 11V/us at 20kHz.
max. slew = dU/dt = 2pi*f*Upeak, so what is the problem with these bipolars?
I like the Sanken 2SA1216/2SC2922.
 
K

Kevin Aylward

Jan 1, 1970
0
Ban said:
Well, with 500W into 8R you will have a +/-90V peak swing, which
corresponds to a slew rate of 11V/us at 20kHz.
max. slew = dU/dt = 2pi*f*Upeak, so what is the problem with these
bipolar?

Well, first its difficult to get those magic specs that those audio
goldern ears dudes like. Secondly, its a marketing game, not a reality
game.

Thirdly, this slew rate limit is a bit misleading. You need say, at
least twice what this number says in practice. When an amp is slewing
everything collapses. Class A stages don't want to be run to their full
swings. If an amp is collapsing at 20khz, its recovery time might well
be significant.

I can tell you this, if an amp slews at say, only 25khz, its 20khz/19khz
imd will be quite large. It all goes hand in hand, a low distortion amp
at hf, requires a larger slew rate than what might be implied by simple
overload calculations. That is, an amp that has 0.01% at 20khz, implies
indirectly an associated faster slew rate. The fast slew rate is a
*side* effect of obtaining good hf performance. This seems to be missed
by everyone since Baxendale made his observation of records only
requiring 0.5V/us.
I like the Sanken 2SA1216/2SC2922.

I am aware of these, but I don't have their spice models. I have looked
at the 2SC4001 and 2SA1546 as well, of which I also don't have models
for. I am playing with the Motorola MJL4281/MJL4302. I did have to
adjust the models a bit. Their hfe was a bit optimistic.

My spice designs so far with the mos are quite good. One with a simple
design is doing < 0.01% at 20khz, the second one, with with an extra
special custom feedback loop, is < 0.0004% at 20Khz, acording to spice.
At which point I will say, this. Its hard to accurately analyse x-over
in spice, so I dont know what this will do to the figures yet.
Simulation time for this would take way too long. I use the ac small
signal distortion feature of SS, a feature not available in LTSpice:).
This feature I think is indispensable. It allows one to at check that
the class A driver stuff is not limiting performance, and I can assure
you that D.Self's "Blameless" amplifier
(http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm), isn't. Its just
can't cut it with this level of performance. Indeed below 1khz 1'm
getting, -140db distortion!

Once I'm finished playing, I will post some of the schematics. My ouput
stage is rather novel.

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

Pooh Bear

Jan 1, 1970
0
john said:
In UK, for prototyping and development purposes, most people will select
from the RS or Farnell catalogues. If you look at (say) the Farnell "GP amps
and switches" transistor pages it's 95% BC... and ZTX... numbers. So for
simplicity and stock availability, these'll be bought in the first instance.
There's no pro-electron bias, it's just a function of whichever supplier is
giving Farnell/RS the best purchasing deals.

For low power devices - you're possibly mainly right. We used to buy certain
MPSAs from Farnell though.

Look at high power and there's hardly a pro-electron designation device to be
seen. Mainly 'cos there aren't any of note. Indeed I'm barely aware of any
newish pro-electron designation types at all.

Many power devices of note are manufacturer part no type - like the MJs.

For volume production of course it's only the price that matters and each
device can be looked at for possible re-spec using worldwide sources.

Also, there are quite specific devices that are application specific. I would be
lost without certain 2 SAs for example.


Graham
 
P

Pooh Bear

Jan 1, 1970
0
Kevin said:
Well, I am currently playing around with simulations of an audio amp
using state of the art bipolar (ft=50Mhz, 15A, 230V, 150W) and its a no
contest. Mosfets switch much faster. I can get 1000V/us (500W @8ohms)
with mosfets, only about 100V/us with bipolar.

And such slew rates are relevant to audio in *which* way ?


Graham
 
P

Pooh Bear

Jan 1, 1970
0
Klaus said:
The reason MOSFETs are used in switch-mode
supplies are more that they are easy to drive (no gate current).

My arse !

Of course the low capacitance is a second order effect

You *have* to be kidding !

The gate power required to drive MOSFETs at modern switching frequencies is a
MAJOR consideration.

Why do you think certain semi makers make specialist gate drivers ! ?


Graham
 
P

Pooh Bear

Jan 1, 1970
0
Ban said:
Well, with 500W into 8R you will have a +/-90V peak swing, which corresponds
to a slew rate of 11V/us at 20kHz.
max. slew = dU/dt = 2pi*f*Upeak, so what is the problem with these bipolars?
I like the Sanken 2SA1216/2SC2922.

Sanken have always made excellent devices IMHO. A certain person I know ( who
should know ) reckons they under-rate Pt considerably compared to most semi
makers.


Graham
 
Top