trimpots at speed

[John Larkin]

I'm designing an optical-electrical converter (fiber-coupled pin
diode, opamps, stuff like that) and figured I might want to offer a
calibrated version, 1 v/mw or something. So I figure I'll just use a
small trimpot (4mm surfmount) in one of the gain stages. I'll use 800
MHz opamps and maybe get a couple hundred MHz overall bw.

So, is there anything wrong with using trimpots at this speed? What's
the fastest anybody *has* used trimpots? Any advice?

And no, I'd prefer not to use a VGA or anything like that if it can be
helped.

My first customer for this is doing EMP testing on all sorts of
military hardware, and needs fiber links to trigger things. He's
talking e-fields of 50,000 v/m and induced currents in the kiloamps.

John

 

[Rene Tschaggelar]

I'm designing an optical-electrical converter (fiber-coupled pin
diode, opamps, stuff like that) and figured I might want to offer a
calibrated version, 1 v/mw or something. So I figure I'll just use a
small trimpot (4mm surfmount) in one of the gain stages. I'll use 800
MHz opamps and maybe get a couple hundred MHz overall bw.

So, is there anything wrong with using trimpots at this speed? What's
the fastest anybody *has* used trimpots? Any advice?

And no, I'd prefer not to use a VGA or anything like that if it can be
helped.

My first customer for this is doing EMP testing on all sorts of
military hardware, and needs fiber links to trigger things. He's
talking e-fields of 50,000 v/m and induced currents in the kiloamps.

John,
you mean to use the pot as feedback in a transimpedance or opamp.
Considering that going with 50 Ohms becomes more important the higher
the frequency is, I see some problems with the pot to get a decent
impedance. Otherwise is is just a few additional nH and thus RF-fitting.

I once did a gain adjustment by using a pot in the supply path
to set the gain of MAR6 amplifiers. This worked at the expense
of VSWR and such. But it worked. Now I'd perhaps use a voltage
controlled attenuator, such as the RVA2500 plus some excess gain.

BTW, the calibrated version ... photodiodes are not necessarily
linear when the pulses are shorter than the bandwidth of the
diode, just in case. Exact optical power measurements are a
nightmare. Being 50% off is common.

Rene

 

[Phil Hobbs]

I'm designing an optical-electrical converter (fiber-coupled pin
diode, opamps, stuff like that) and figured I might want to offer a
calibrated version, 1 v/mw or something. So I figure I'll just use a
small trimpot (4mm surfmount) in one of the gain stages. I'll use 800
MHz opamps and maybe get a couple hundred MHz overall bw.

So, is there anything wrong with using trimpots at this speed? What's
the fastest anybody *has* used trimpots? Any advice?

And no, I'd prefer not to use a VGA or anything like that if it can be
helped.

My first customer for this is doing EMP testing on all sorts of
military hardware, and needs fiber links to trigger things. He's
talking e-fields of 50,000 v/m and induced currents in the kiloamps.

John

One pF in odd places can do a lot of strange things in a 200-MHz TIA, if
it's inside the feedback loop. Getting modestly well-calibrated dc
response is pretty simple, but having the ac gain be the same is the
trick. The transimpedance bandwidth is roughly sqrt(GBW*f_RC), where
f_RC is the 3 dB corner of the feedback resistor and the total
capacitance on the summing junction (C_diode + C_in), so to get 200 MHz
from an 800 MHz op amp, you need 1/(2*pi*Rf*C) > 50 MHz.

I'd probably put the pot in a low-impedance part of the circuit
following the TIA. Maybe use a voltage divider with a loaded pot as the
shunt element (e.g. 500 ohm pot with 50 ohms from each end to the
wiper). That way the problem is more like scope probe compensation and
less like Q-multiplier design.

Back in the palmy days, Bourns used to make 50- and 75-ohm variable
attenuators that worked up to > 100 MHz.

Rene's comments about accuracy are also worth thinking about, though I'm
not quite as pessimistic as he is. Still, you might need to think about
how to control etalon fringes in the fibre. An APC fibre stub between
the input connector and the PD would be a good possibility--it would
reduce the backreflection from the fibre end significantly, and give you
a clean and repeatable transition from fibre to photodiode.

Cheers,

Phil Hobbs

 

[John Larkin]

John,
you mean to use the pot as feedback in a transimpedance or opamp.

I'm thinking of using three current-mode opamps: tia stage, inverting
amp, output buffer. I could tweak the gain anywhere, probably the
middle stage.

Considering that going with 50 Ohms becomes more important the higher
the frequency is, I see some problems with the pot to get a decent
impedance. Otherwise is is just a few additional nH and thus RF-fitting.

I once did a gain adjustment by using a pot in the supply path
to set the gain of MAR6 amplifiers.

I've used supply-current trimming (with a dac) to set the input
impedance of a mmic. One of the Sirenza SiGe parts can be rimmed to
exactly 50 ohms input impedance. It changes the gain a bit, of course.

This worked at the expense
of VSWR and such. But it worked. Now I'd perhaps use a voltage
controlled attenuator, such as the RVA2500 plus some excess gain.

BTW, the calibrated version ... photodiodes are not necessarily
linear when the pulses are shorter than the bandwidth of the
diode, just in case. Exact optical power measurements are a
nightmare. Being 50% off is common.

That is an issue. Just replacing a fiber cable with an "identical" one
can change things a few dB; just unmating/remating one connection can
do the same thing, especially on the laser end.

I guess I'd look pretty silly selling a "calibrated" detector if the
cal spec was +-50% or something.

John

 

[Rene Tschaggelar]

Rene's comments about accuracy are also worth thinking about, though I'm
not quite as pessimistic as he is. Still, you might need to think about
how to control etalon fringes in the fibre. An APC fibre stub between
the input connector and the PD would be a good possibility--it would
reduce the backreflection from the fibre end significantly, and give you
a clean and repeatable transition from fibre to photodiode.

I've seen power meters of the expensive, bolometer
type to differ by factors of two when illuminated by
the identical beam. All claimed to be calibrated.
Upon closer thinking I found calibrating a light source
myself to be rather tricky. Compare to a blackbody ?
The unknown emissivity comes in. Attenuate and count the
photons ? What about the lost photons ?
Apparently the industry can live with a uncertainity
of 3dB, the problem is known I was told.

The APC connector is a good idea to reduce backscatter.
Is it polarisation dependent ?

Rene

 

[Active8]

On Thu, 03 Mar 2005 12:42:07 -0800, John Larkin wrote:

I guess I'd look pretty silly selling a "calibrated" detector if the
cal spec was +-50% or something.

So call it "calibratable" or "adjustable." Almost every time
something has changed in a CATV system, I've had to rebalance or
sweep.

Ever coil yer fiber around a pencil?

 

[Frank Miles]

I'm designing an optical-electrical converter (fiber-coupled pin
diode, opamps, stuff like that) and figured I might want to offer a
calibrated version, 1 v/mw or something. So I figure I'll just use a
small trimpot (4mm surfmount) in one of the gain stages. I'll use 800
MHz opamps and maybe get a couple hundred MHz overall bw.

So, is there anything wrong with using trimpots at this speed? What's
the fastest anybody *has* used trimpots? Any advice?

Trimpots have been used for a long time in oscilloscope vertical amplifiers.

What follows will probably be obvious to you: if you use them in a
low-impedance path, the inductance associated with the path length will
get you eventually; on a high-impedance node, it will be the stray capacitance
that will be problematic.

Be careful with design, part selection, and layout and you will almost
certainly be able to achieve >200MHz bandwidth. It's been done, and with
larger parts than the tiny parts now available.

-frank
--