Programmable gain amplifier flat to 10 MHz

[jbb]

Hi all

I’m making my way though the initial design of an impedance analyser project (it’s like an LCR meter). I have, perhaps foolishly, set myself a goal of operating from 50 Hz to 10 MHz.

Part of the design requires range switching of some sense signals. They could come in at a range of roughly 1 mV to 1 V RMS. So I’d like to have 1x, 10x, 100x gain or near offer.

The obvious tool for the job is a Programmable Gain Amplifier (PGA). However, it’s hard to find one which is flat to 10 MHz. Which means a -3rB bandwidth significantly higher than 10 MHz…

Does anyone have any suggestions?

I’m could make my own using a current feedback opamp and a multiplexer to adjust its gain. But I’m worried about stability problems because the multiplexer will add stray capacitance to the feedback network. Has anyone tried this before? Did it work out OK or was it a problem?

[Andree Henkel]

your requirement are somewhat heafty.
gain 100 * Bandwidth 10MHz means Gain-Bandwidth of 1Ghz
so I would advice that you build two stages:
1st stage switchable 1/10, second stage fixed at 10, and you select if you use output of 1st or 2nd stage

so no programmable amplifier, just two OPA + two analog switches + some resistors

[jonpaul]

use reed relays

j

[moffy]

The ADA4310-1: https://www.analog.com/media/en/technical-documentation/data-sheets/ADA4310-1.pdf
is a dual current feedback opamp that with a gain of +10 has a BW of around 150MHz. If you used the pair each with a gain of 10 in series you get your x1, x10, x100 just by switching the outputs.

[jbb]

Hmm

Reed relays definitely get me much better on resistance and off capacitance than silicon switches. Would need to think about how often they switch during operation but they’re definitely an option.

[ArdWar]

The ADA4310-1
is a dual current feedback opamp...

Just beware that there are distinction between "normal" and CFB amplifier. Too often those unaware just use them like regular opamp only to wonder why the magic smoke escapes.

[jbb]

Those current feedback amplifiers seem to be pretty fussy about their feedback networks, so I'm trying to be careful.  It makes sense that meaningful gain above 100 MHz goes hand in hand with the risk of oscillation.

I could maybe squeak out acceptable performance using a silicon switch, but I'm worried about stability; apparently even a little capacitance around the inverting input is Bad News and my LTSpice simulations show concerning humps up around 200 MHz...

I think reed relays would be the lowest risk right now.  Looks like I can get 0.2 Ohm on resistance and somewhere in the vicinity of 3 pF capacitance to ground, which is miles better than an analog switch IC.

[moffy]

CFB opamps are straight forward if you follow the guidelines, keep feedback impedances within recommended limits, follow basic good layout guidelines. They offer excellent gain at good BW without the usual voltage follower fixed GBW product.

[jonpaul]

scopes, high res DVM, prodessional preeamps all use switches or relays and not active devices to change ranges.

Jon

[David Hess]

It is usually a bad idea to switch the feedback networks for variable gain for the reason that you give; it increases input capacitance at the inverting input.  It also inevitably alters the frequency response.

More commonly a fixed gain input stage is followed by switched attenuation stages and fixed gain stages.

[Terry Bites]

I checked out ADGM1001's
Fab performance at a shocking price! $50, really?

[nfmax]

This is a circuit topology I have used with success for a switched-gain amplifier with very well controlled bandwidth, using two conventional OPAMPs and an analogue multiplexer. It avoids some of the issues with the multiplexer, and even turns them to advantage. Note the component selection is arbitrary for illustration, you will need to choose appropriately.

U1A is wired as a non-inverting amplifier, U1B is inverting. The analogue multiplexer U2 and resistors R2-R10 act as a switched potentiometer. When input X7 is selected, the feedback resistance of the non-inverting stage is at the minimum value, while the input resistance of the inverting stage is at maximum: this gives minimum gain. As lower-numbered channels are selected, the gain of each stage is increased. By choosing suitable resistance values, you can get gains evenly spaced in dB.

Note that the gain is not affected by the multiplexer on resistance, which simply appears in series with U1A output resistance, so its effect is reduced by the negative feedback around the OPAMP. What may be less obvious is that the multiplexer capacitance (i.e. between Xn and GND) actually helps.

As the gain of each stage is increased, its bandwidth will be reduced because of the limited gain-bandwidth of the OPAMP. However, the parasitic capacitances, in conjunction with the gain-setting resistors, act as a cascade of RC low-pass filters. At minimum gain, all these filters appear in the input resistance to U2B, thus rolling off its gain at high frequency. At maximum gain, these same filters are removed from U1B's feedback network and instead act to reduce the feedback applied at high frequency around U1A. A judicious choice of resistor values can give a circuit bandwidth almost unaffected by the gain setting.

I found a gain range of around 30dB was practical with this circuit: you can cascade two to get your required range.

I suggest using the best OPAMP models you can get to simulate the blazes out of the design, tweaking resistor values as necessary to get the best results, then breadboard. For my application, I was working with a bandwidth of 2.5MHz, but I needed to control the phase response to within ±5˚ over the bandwidth. I found it necessary to add a few pF additional capacitance to ground from one of the taps - this didn't show up from the simulation results.

Oh, and watch the power rail decoupling. The PSRR of most OPAMPs is pretty dire at high frequencies, and feedback thorugh the supply rail can mess things up something rotten!

[exmadscientist]

A couple notes from when I did something similar a few years back:

• This is hard. Really hard. I only had to go to 1MHz to meet spec, and you're asking for 10.
• You cannot use leads at 10MHz. Even at 1MHz leads are a disaster. There is a reason the HP/A/K LCR meter leads fixture has an upper limit of 100kHz.
• Fully differential signal paths are your friend. This is a lot easier to do now than it was 30 years ago, and it can help a lot.
• AD8129/8130 is a really interesting and unique part that might help you. It consumed too much power for my application, but be aware of it.
• People build 3-op-amp instrumentation amplifiers all the time. Less so, now that there are high-performance monolithic in-amps in the world, but it's still done. But did you know that you can build instrumentation amplifiers out of 3 monolithic in-amps? They work really well.

Bonne chance!

[jbb]

Yes, I am coming to the conclusion that 10 MHz is asking for a lot.

I am indeed planning to use the AD8130 in some other parts of the circuit - specifically for converting the differential output of the main sine wave synthesis DAC into a single ended drive signal.

[David Hess]

Why not use two x10 gain stages in series and then pick off the output, x1 through x100, that you want with a multiplexor?

x100 is 40dB so not completely unreasonable in a broadband amplifier.  Care will need to be taken to prevent the high level output from getting back into the input.  Use single operational amplifier packages and not duals or quads.

Alternatively, use a multiplexor but disconnect the outputs of amplifiers which are not being used, like with a shutdown function.  Many fast operational amplifiers support shutdown for exactly this reason.

[jbb]

That’s kind of tempting, actually. Unfortunately it leaves me with a tricky question; what will happen to the non inverting (ie “+”) input on shutdown mode? I’ve been playing with ADA4860 in LTSpice - it’s probably not the best amplifier for the job! - and its data sheet doesn’t say.

Does anyone know what happens to a current feedback amplifier non-inverting input when it’s shut down? Or have any commentary on whether a SPICE model is likely to provide useful results?

I guess it’s time to make up a test board to try some options out.
- switchable x1, x10 amp with reed relay
- a switchable /1, /10 attenuator followed by fixed x10 amp
- a cascaded x1, x10, x100 amplifier with selection mux and amplifier shutdowns

[BrianHG]

I would have used this good old classic 90mhz video amp which uses a resistor between 2 pins to set the gain between 0 and 400.

https://www.mouser.com/datasheet/2/308/ne592-d-1194344.pdf

It is possible to set an adjustable gain using a reed relay or jfet and 2 resistors.

[David Hess]

That’s kind of tempting, actually. Unfortunately it leaves me with a tricky question; what will happen to the non inverting (ie “+”) input on shutdown mode? I’ve been playing with ADA4860 in LTSpice - it’s probably not the best amplifier for the job! - and its data sheet doesn’t say.

Does anyone know what happens to a current feedback amplifier non-inverting input when it’s shut down? Or have any commentary on whether a SPICE model is likely to provide useful results?

The is an interesting question because the ADA4860 datasheet is incomplete.  When shut down, its input bias current increases from -250 nanoamps to 130 microamps, but no data is available about what its output does.  Usually the output becomes high impedance, and I suspect that is the case here.

The ADA4860-1 actually seems like one of the better choices because of its low cost and good performance.