Hi Guys,

I'm designing a high speed (20MHz) transimpedance amplifier with switchable gains of 1500, 15000 and 150000. I plan to use 4 amps in the following arrangement to achieve this:

I want to use only a single input ADC to read the output of this amp, which means I need a way to switch between points A, B, and C. My current idea is to connect all the outputs together, put each amp's power supply on a semiconductor switch, and switch off power to all the amps that are out of range of the measurement. Is this a silly idea or could it work? The paths to ground via the feedback and input resistors on the two amps that are switched off do not concern me, what I am wondering really is will the amps that are powered off provide enough impedance at high frequency to the power rails, inputs and ground? Is there anything I can check in the datasheets? Output impedance characteristics of op-amps when they're off aren't easily found. Last resort, i'll just build it and try.

Cheers!

 

Is this a silly idea or could it work?

Aside from the challenges of making a 20 MHz with such high gain, your idea might work depending upon the particular amplifiers involved. What amplifiers are you planning to use?

 

There is no rule of thumb for this; all parts are different. The powered-off impedance to either rail could be as low as a few hundred ohms. Probably higher, but I wouldn't assume anything above a few K and it will vary with signal amplitude.

Maxim makes video amplifiers with switchable outputs for multiplexing applications. They should have the bandwidth you need at low gain, but a gain of 100 at 20 MHz is a GBW of 2 GHz. That will be a hard part to find. Numbers like that sound like the LNA for a satellite dish.

When you find it, a better path might be to have one opamp with three feedback resistors and switch the resistors in and out with small reed relays. Again, analog switches with enough (how much?) off isolation at 2 GHz are ... difficult.

Now, about the 30 GHz input amp GBW...

ak

 

I'm planning to use an OPA657 as the transimpedance amp, and 3x OPA847 as the voltage gain amps (the OPA847 isn't stable at gains of less than 12, however I'll set the noise gain to something like 20 to stabilise it).

 

There is no rule of thumb for this; all parts are different. The powered-off impedance to either rail could be as low as a few hundred ohms. Probably higher, but I wouldn't assume anything above a few K and it will vary with signal amplitude.

Maxim makes video amplifiers with switchable outputs for multiplexing applications. They should have the bandwidth you need at low gain, but a gain of 100 at 20 MHz is a GBW of 2 GHz. That will be a hard part to find.

When you find it, a better path might be to have one opamp with three feedback resistors and switch the resistors in and out with small reed relays. Again, analog switches with enough (how much?) off isolation at 2 GHz are ... difficult.

Now, about the 30 GHz input amp GBW...

ak

The frequency response of a transimpedance amplifier is not calculated by the simple GBWP/Gain formula:

http://www.ti.com/lit/an/sboa122/sboa122.pdf

I have de-rated the 1.6GHz OPA657 to 1GHz GBWP over its full temperature range (as this amp will be used) and I still clear the bandwidth requirement by 3MHz (23.03MHz) with an input capacitance of 200pF.

I also really don't want to be putting reed relays into the feedback path of the transimpedance amp. The best I could find was this:

http://www.farnell.com/datasheets/1728704.pdf

and 0.4pF x 2 equates to ~1k parasitic impedance. Not good.

 

NICE TI paper.

Completely agree about relays being non-ideal, but they often have better high freq off isolation relative to analog switches.

Also, doesn't 0.8pf and 1K work out to 200 MHz, not 20 MHz?

Maxim and maybe Linear Tech or ADI have video switches with a T switch configuration that might have lower overall (circuit plus packaging) feedthrough.

ak

 

NICE TI paper.

Completely agree about relays being non-ideal, but they often have better high freq off isolation relative to analog switches.

Also, doesn't 0.8pf and 1K work out to 200 MHz, not 20 MHz?

Maxim and maybe Linear Tech or ADI have video switches with a T switch configuration that might have lower overall (circuit plus packaging) feedthrough.

ak

Ahh yes you are right, it's 10k combined, not 1. That does make a difference, perhaps these relays could be used for switching gains after all. Or, I could go with the topology above and use a relay to switch the outputs off whilst also shutting down the op-amp too.

I'll also have a look into those video switches you mentioned, possibly used in a similar way as described above. Thanks for the help!

 

How fast must the switching action be?

ak

 

How fast must the switching action be?

ak

There's no constraint on switching speed, slow is fine. Once the gain has been switched, it will stay like that for a while.

 

I'm still leaning toward one amp with three gains rather than three amps. Feels like something easier to debug now and maintain down the road. I get that your circuit is very sensitive to stray capacitance, and layout issues might kill this approach.

Fir the three-amp approach, another alternative is diodes, the original high freq switch.

ak