Selectable Gains With Analog Switches

Sure that part number is right? Analog Devices doesn't seem to know the AD1604.

TIAs go berserk mainly in one situation: Too much capacitance from IN- to ground while there isn't enough from IN- to OUT.

Again, depends on two things:

a. Your resistor values.

b. MUX capacitances, especially the ones to ground or supply (which in terms of bandwidth calculations is like ground).

Why not use an active resistor or a digital pot?

With a bjt you can easily provide programmable variable gain. You can combine the ideas by using driving the bjt with a digital pot or switched array of resistors if you want a discrete version.

It also depends on where you put the analogue switch: for example: in a simple circuit, with a series combination of feedback resistor and analogue switch, if you put the switch next to the input of the op amp, the charge injection will upset the stability of the amplifier (you are injecting charge right into the input). If you put the switch on the output side, the charge injection is injecting into a low impedance node (the output) and is buffered from the input by the feedback resistor.

Both configurations are electrically identical, but one has problems and one does not (... or has less).

You are right, it is the wrong number. The correct number is ADG1604:

If the only dependence is on the resistors I am switching and the capacitances of the MUX, what is the rated bandwidth of 15MHz based on?

I the highest value resistor I am switching is 8 Kohm.

One thing to watch out for... If changing "gain" also changes GBW you may change your phase margin and get instability.

...Jim Thompson

I know about capacitances on virtual grounds and know how to calculate the noise and bandwidth limitations they impose. The data port of the mux has been designed to be on the op amp's output. This is well covered by Jerald G. Graeme.

For the TIA I have decided to use the ADG1404 which has a bandwidth of

For the voltage gain amplifier I have decided to switch the resistors from the inverting input to ground of a noninverting amplifer instead of feedback resistors to take the analog switch out of the feedback path. This would assure stability but I am not sure what that would do to available gain bandwidth constant for calculating gain versus bandwidth.

I do not have serial buses available to control digital pots. This board must be controlled by the digital port of a National Instruments PCI-6110 card. The selectable gains of the voltage amplifier must be from 1 up to 1000 while maintaining a bandwidth of at least 1 MHz. I have decided to use two stages of an LT1806 to do this.

Use a jfet ?

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

Jfets have linear resistances only at very low voltages.

That thing has HUGE capacitances. It's only good at DC or audio.

Again, depends on your values and schematic.

Just do the math, assuming a plain RC path: 8K and 100pF or more in capacitance, uhm, I don't think you'll even get to the MHz range much.

I assume Graham meant hard-switched JFETs.

Up to about 100mV pk.

V=IR. How much current were you planning to put through them ?

I know some JFETs go as low as 8 ohms typ. Not good enough ?

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

With the op amp's output on the D port of the ADG1604, or even the ADG1404, the capacitance of either of these switches will not make a pole with the 8K resistor.

My concern now is that it will make a pole with the output resistance of the LTC6244 causing instability. This will make necessary a series resistor and capacitor pair across the op amp's inputs for stability. The value of these will depend on the feedback resistance so now I am looking at the possibility having to switch these as well.

The switchable gains for the TIA were a nicety and not a necessity. I am giving up on the switchable gain for the TIA.

The switchable voltage gain is a necessity. I will look at switching resistors for it using jfets.

Correct ( in view of the word 'switch' ). You use as many as you need ( as bits of resolution ). Been there, done that. They can work well as linearish control devices too.

Graham

-- due to the hugely increased level of spam please make the obvious adjustment to my email address

The input-to-output capacitance will not make for much of an "off" position. Again, a switch with an Rdson of 1ohm is nice for heavy duty stuff but not at several kohm and several MHz.

Yup, you've go to calculate your phase margins. BTW, the cap then doesn't go across the inputs but from IN- to OUT, then R to load.

With a low-C mux you could do it.

Without a schematic none of us will really know what you are tring to accomplish. But have you thought about using an amp with gain control input? They cost more but with all those muxes you are putting on it might pan out price-wise.

BTW, did you bring that sinewave burst project to a success? Just curious.

In the young and wild days we used cheap current-steering DACs for gain control. It made some people's toe nails curl up during the design review.

CD and RD at:

The principle is the greater the feedback ratio the more stable a control system is. RD increases the ratio. CD causes that ratio to be in effect only at the needed frequencies.

Ok, got your email with schematic. If you really want to use FET switches, this is what I've used in similar situations:

Excellent performance, low capacitance, if the roughly 70ohms Rdson bother you, you can neutralize that by using one switch (on the saem chip) as a dummy in a reverse feedback path or something. Or maybe use the same chip on both 1k, on both 2k, and so on respectively if that would cancel out. The Rdson on the same chip tracks quite well.

The downside: SD5400 have become specialty parts, now only via LIS, and I've heard not such good things about actually getting parts from them. But if it's for just one or a few units it could be ok.

Bottomline the mux capacitances need to be in the low single-digit pF-range and that means highish Rdson. Only relays are better but you probably don't want those.

Ok, it sounded like you just wanted to place a capacitor across IN- and IN+. That wouldn't have worked.

I usually prefer the in-loop method but as people say, there are many roads to Rome ;-)

I have decided use a programmable gain amplifier instead of trying to design and build this myself. The AD8231,