Voltage-controlled amp design is hard

An MDAC as the variable gain element sounds good. DAC8812 or some such.

Analog Devices still makes real multipliers.

Do you need continuous gain scaling, or would steps work? A few steps would be easy.

I like THAT Corp. VCAs for audio work but I'm guessing those are way off the mark for your work. ?

Like JL I was going to suggest a real analog multiplier... for $700 you've got some money to burn.

How about diddling with the APD bias voltage to change the gain, that should go from 1 to ~100, maybe 300?

George H.

How about a diaphragm, as in a camera, to reduce sensitivity? Moving strip with ever smaller holes? Keep gain at maximum?

There's a second source for the THAT Corporation 2181 VCA, +20/-20 supplies, -60dB to +40 gain control range, 20MHz bandwidth. Looks like a pretty nice part though don't know if it's low enough noise over the range for your application.

$1.10 each in quantities over 100:

Some alternatives to VCA include PWM (precise but low bandwidth); Amplitude modulating RF and demod; or diode bridge vario-losser (wide bandwidth but imprecise and limited to 30-40dB range).

You mentioned bandwidth to few hundred kHz but does it extend to DC?

If ac coupled can you tolerate control voltage feedthrough (what audio compressors call "thump")?

Do you need predictable gain tracking control voltage or will it be inside a feedback loop? Would gain range switching be enough?

piglet

Or compress the range into one making the TIA into a log-amp?

By tradition I thought photodiodes and logamps went together like peaches and cream :)

piglet

I've made an Analog Devices multiplier a mainstay in many different types of designs, e.g., the 10MHz AD734. But that's a high-voltage +/-15V part, and Barrie Gilbert hasn't seen fit to update his design for low-voltages.

Hmm, maybe a pwm thing with a lockin/ low pass on the back end? And some switched section for different ranges.

I'd guess he wants DC, but best wait for Phil.

George H.

On Jul 30, 2018, Winfield Hill wrote (in article ):

I doubt that Barrie Gilbert designs commodity multipliers these days, so I?d be barking at the relevant Product Managers at Analog.

Joe Gwinn

AD835 does Z = X * Y, in pure volts, with X and Y range +-1 volt. Amazing part. Maybe too noisy for Phil.

There are some parts with +-10 inputs too.

Yup, DC-300kHz or thereabouts.

Last time I checked, analogue multipliers had about 1-2 uV 1-Hz noise, which is worse than the OTA thing. Is the 734 dramatically better?

Cheers

Phil Hobbs

We need a lower-frequency, lower-power version!

Yeah something like that. There's the DC offset too... if you care about that. I use the AD734, but I happiest putting big signals into the input.

George H. (How about changing the bias voltage.. that seems obvious, so you must have rejected it for some reason.)

Oh, just buy a fan.

It's initially for retrofitting, so I can't change the optical system. Also wasting photons is equivalent to the amplifier getting noisier.

Cheers

Phil Hobbs

Thanks. Diode-bridge attenuators have the same shot-noise problem as transconducance things. What I really want is an op amp with a variable feedback resistor, hence the MDAC/dpot idea. 256 steps really ought to be enough--the maximum range of gains is only 100:1.

Cheers

Phil Hobbs

The customer would say "Not responsive to requirements." ;) Plus log amps, being transconductance devices, have the same noise issue unless you run a zillion diodes in series.

If you want the log, just run the photodiode open-circuit with a noninverting buffer on it. A split diode and a diff amp lets you compute the ratio easily, which is useful for beam pointing applications--the output voltage doesn't depend very much on the laser power.

Cheers

Phil Hobbs

The gain range isn't enough--these diodes have a SNR sweet spot at about M=20, mostly because the TIA itself is so quiet.

With an avalanche gain of 20, its noise floor is equivalent to the Johnson noise of a feedback resistor 400 times higher, i.e. 400M.

Cheers

Phil Hobbs