Op-amp Gain-Bandwidth Product (GBP) spread.

[[email protected]]

Has anyone ever looked into the piece to piece variation in the
GBP of an opamp. I'm talking about "garden variety" voltage feedback
opamps with GBP less than 50 MHz. I've read somewhere that
manufactures work to keep this fairly constant, but I've never seen
any tolerances, min and max values or a nice histogram of values. I'd
be happy to hear that the variation is less than 10%, which is what my
limited experience would tend to suggest.

Thanks,
George Herold

 

[Telstar Electronics]

     Has anyone ever looked into the piece to piece variation in the
GBP of an opamp.  I'm talking about "garden variety" voltage feedback
opamps with GBP less than 50 MHz.  I've read somewhere that
manufactures work to keep this fairly constant, but I've never seen
any tolerances, min and max values or a nice histogram of values.  I'd
be happy to hear that the variation is less than 10%, which is what my
limited experience would tend to suggest.

Thanks,
George Herold

George, sorry to say that I don't know the answer... but I'm
interested in why you are asking.
Are you planning on operating so close to the GBP?

 

[Eeyore]

Has anyone ever looked into the piece to piece variation in the
GBP of an opamp. I'm talking about "garden variety" voltage feedback
opamps with GBP less than 50 MHz. I've read somewhere that
manufactures work to keep this fairly constant, but I've never seen
any tolerances, min and max values or a nice histogram of values. I'd
be happy to hear that the variation is less than 10%, which is what my
limited experience would tend to suggest.

I've come across an instance where some TL072s had lower than typical GBP
causing units to fail bench test spec. For that batch we replaced them
with LF353s.

Graham

 

[whit3rd]

     Has anyone ever looked into the piece to piece variation in the
GBP of an opamp.

In general, the variation in gain-bandwidth should
be expected to scale with the quiescent current (various
current sources interact with a filter capacitor in
internally compensated op amps).

If it's a major concern, consider using a programmable
op amp, like LM4250 or LM346 (are those still around?).
The gain-bandwidth product, within limits, is proportional
to the current you feed to the program pin.

 

[[email protected]]

George, sorry to say that I don't know the answer... but I'm
interested in why you are asking.
Are you planning on operating so close to the GBP?

"Are you planning on operating so close to the GBP?"

Well maybe, The question grows out of discussions with a colleage. A
new instument has a final high level (10 V peak) gain stage with a 3dB
corner at 1 MHz, set by the opamp GBP. Upstream is a 100kHz low pass
filter (set by R's and C's) so the 1 MHz point is not that critical.
But I was then thinking of using the gain stage without the upstream
filter. We then wondered how consistent the GBP was.

George

 

[[email protected]]

Generally, all of the resistors and capacitors on the op-amp will have a
worst case absolute tolerance of about 30% batch to batch, typically within
about 10%, (but the ratio of the value of any two resistors or any two
capacitors on the same chip will usually be within less than a percent of
the desired ratio).  You should expect that with one manufacturer, the GBW
will probably vary by several percent part-to-part most of the time but you
might then get a batch 20% different, if they have to replace one of their
machines, etc.  If you rely on the value staying the same then you should
really test a sufficient quantity of parts and stockpile them, as there is
no guarantee that you will be able to get identical parts later (unless the
manufacturer gives you one).

Chris

Thanks Chris,
That makes sense. I don't care so much about the absolute value, as
long as it's consistent from unit to unit. So I can order enough
parts for a year or two, spot check them and tweak the circuit
accordingly, and then repeat in a year or two when we have to order
them again...(if they haven't gone obsolete.)

George

 

[[email protected]]

Generally, all of the resistors and capacitors on the op-amp will have a
worst case absolute tolerance of about 30% batch to batch, typically within
about 10%, (but the ratio of the value of any two resistors or any two
capacitors on the same chip will usually be within less than a percent of
the desired ratio).  You should expect that with one manufacturer, the GBW
will probably vary by several percent part-to-part most of the time but you
might then get a batch 20% different, if they have to replace one of their
machines, etc.  If you rely on the value staying the same then you should
really test a sufficient quantity of parts and stockpile them, as there is
no guarantee that you will be able to get identical parts later (unless the
manufacturer gives you one).

Chris

Thanks Chris,
That makes sense. I don't care so much about the absolute value, as
long as it's consistent from unit to unit. So I can order enough
parts for a year or two, spot check them and tweak the circuit
accordingly, and then repeat in a year or two when we have to order
them again...(if they haven't gone obsolete.)

George

 

[[email protected]]

I've come across an instance where some TL072s had lower than typical GBP
causing units to fail bench test spec. For that batch we replaced them
with LF353s.

Graham

Hmmm, fortunately for this project I can set the specifications at the
end. We shoot our arrow into the air and then draw the specification
circle around where it lands.

George

 

[[email protected]]

In general, the variation in gain-bandwidth should
be expected to scale with the quiescent current (various
current sources interact with a filter capacitor in
internally compensated op amps).

If it's a major concern, consider using a programmable
op amp, like LM4250 or LM346 (are those still around?).
The gain-bandwidth product, within limits, is proportional
to the current you feed to the program pin.

"programmable op amp, like LM4250 or LM346"

I've never used those, but I'll check out the spec sheets.
Thanks
George