Variable gain input for high-speed ADC

[GeoAVR]

Hi all,

I want to build a data acquisition system for 0-10MHz signals but I am having trouble with the variable gain required before the ADC.

I am using a High speed differential output opamp to drive the ADC but since the input signal has variable Vpp voltage I would like to implement a variable gain amplifier.

Is a digital potentiometer enough to select gains: 0.1 , 1, and 10 through the resistor network that sets the gain of the Opamp or is there a different approach for high speed signal acquisition?

Additionally, since my source has a 50 Ohm resistance, changing the resistors might change the input resistance of the system. How do I solve this?

Thank you in advance.

 

[barry]

The LMH6881 from Texas Instruments is one solution.

 

[GeoAVR]

Yes, but it doesn't provide attenuation.

 

[D.A.(Tony)Stewart]

With external switched resistors you can program attenuation or gain.

With 1.8GHz GBW product, you can choose even higher gain and more attenuation with discrete resistors and FET or CMOS switches to ground.

https://www.ti.com/lit/ds/symlink/ths3201.pdf

 

[GeoAVR]

But how do I control which resistor will be on and which one off?

Is there an analog switch IC that will leave a signal of 10MHz passing through its channels without distortion?

 

[barry]

So, attenuate your input by 10, and then select gains of 1, 10 and 100. This is not hard!!

 

[D.A.(Tony)Stewart]

I agree using the noninverting input with an attenuator of 9:1 R ratio you get 1/10 gain or -20dB. Then with the non-inverting input the gain is 1+ inverting gain.

The inverting gain Rf/Rin- is solved with switched Rin- to Vref or gnd.
Gain Rin-
0dB Open
20 Rf/10
40 Rf/100

Terminate Vin+ with 1M with 9M series to input... Or 1/10th of this for 1M input impedance.


choose Rf= 10M for low bias current types
Use Rf=1M otherwise.

input offset voltages at Rf/100 are after by input bias current x Rin

stray capacitance of 10 pF can affect frequency response on high impedance , so beware of bandwidth requirements.

 

[GeoAVR]

The problem with the attenuator is that it adds noise and one extra stage to the circuit. Additionally, when the input signal is low, attenuation is not needed. Why attenuate smth and amplify it later ?

I was thinking a circuit like : input -> amplifier -> filter -> ADC -> microcontroller
The microcontroller will control the gain in order to keep the signal inside the ADC range. ie. 1V.
So if I have 100mV, 1V, 10V for input I have to set the desired gain accordingly.

I want to keep noise and distortion to the minimum in this circuit.
Is there any idea how professional equipment changes the gain/range of the signal?

 

[barry]

Other than resistor thermal noise, which is pretty low, what other noise does an attenuator introduce? And in your post #3 you were complaining that "it doesn't provide attenuation". Well, which is it? Do you want attenuation or not?

And then you go on to say "the microcontroller will control the gain". How will it do that, exactly? You mean like with a programmable amplifier or attenuator?

The way professional equipment changes the gain is with, get ready for it: programmable gain amplifiers and attenuators.

 

[GeoAVR]

I mean since I will have an opamp in the circuit, I could use the same opamp for both amplification and attenuation by digitally selecting the Rf resistor. Does this make sense and how this is implemented?

I want to avoid a resistor attenuator at the input since: it adds noise, extra components, I want to have the input resistance at 50 Ohm.

For 10V signal I want to attenuate it by 10 so it would be ok.
For a 100mV signal I only want to amplify it by 10, not attenuate it by 10 and then amplify it by 100. I dont think that is the correct practice.

Therefore I want a universal solution that fits all the input signal ranges.

Thank you

 

[D.A.(Tony)Stewart]

your signal levels are far above noise levels and 10V needs a 2W 50 Ohm non inductive resistor.
Go with the input attenuator.

That's what they do on 100k$ spectrum Analyzers is use a decade pad on front end and decade gain on intermediate.

Ensure your Op Amp can drive 100 Ohms.

 

[Borber]

What you need is called autoranging. Some digital multimeters has it. After level check they automatically select proper attenuation to adjust signal to A/D range. Combination of attenuators and amplifiers (VGA) is usual.

 

[D.A.(Tony)Stewart]

You can terminate the input with a thick film 2 W 50 R load then use the inverting input to control gain/attn. the use a series 50R on the output.

Gain Rin-
-20 10*Rf
0dB Rf
20 Rf/10
40 Rf/100
60 Rf/1000

but this requires series switches between Input and Vin-

 

[GeoAVR]

I think I start to understand.

So there are 2 options.
1. Do smth similar to the following schematic and just have some relays to select the resistance Rf and Rg in order to control the gain/attenuation.


I would prefer to control only the Rf but from the equations I see that the Gain and input resistance are not independent.



2. The other option is an input attenuator, followed by a programmable amplifier.
But I was thinking to put the attenuator then a buffer and then the amplifier and in order to have Rin=50 Ohm the attenuator resistance must be 50 Ohm right? So I need two resistors ( 5 + 45 0hms) for 1:9 ratio (x10 attenuation).

I am I right?
What is the best option for best results?

Thank you