Hello!

There is a relatively high frequency signal that should be decided whether its area under the curve exceeds a certain value or not. This will be done by by an FPGA fter digitization. At different threshold levels the FPGA will give different instructions to other equipment.

The problem is that since the frequency of the signal is quite high, this would require a relatively fast ADC, but since several pieces are required, this would increase the costs. So, as an alternative solution, it would be possible to take the absolute value of the signal, integrate it and digitize the integrated pulse only with a much slower and cheaper ADC.

The signals are something like this, there is a deviation in the amplitude (the maximum is perhaps around 10V, I have not seen anything higher), but the frequency is something like in the picture.

I used this circuit: Link

I simulated it in LTSpice, but the absolute value I get on the left side of R6 doesn't look exactly as it should, it even goes negative.

What should I change on the circuit to make it work normally? I guess this issue comes from the high frequency. Bonus question: at the right side, how should the value of R and C be chosen for the integrator to be suitable for such a fast signal? At least tenths of a second pass between the pulses, so they do not follow each other too quickly.

Thanks for the help!

 

One thing you can do if you suspect the absolute value circuit is make a behavioral voltage source that implements the function in a mathematically correct fashion. This will help you focus on other issues while you work on the absolute value circuit. I just happen to have what you need on hand.

I can also add a mathematical integration operator if you need it so you can test the concept.

 

In the circuit R2 is connected wrong. The op-amps are very slow for what you want to do.

 

Two significant problems:

The amps are two slow for a 10MHz signal.
The amps need a gain-bandwidth of greater than 10MHz times the amp closed-loop gain.

The signal voltage is too low for the diodes to ever become forward biased.

 

The behavioral sources in LTspice have no inherent bandwidth limitations. The purpose of simulation is understanding.

 

The max voltage is about 10V. What is the minimum voltage you care about? (how many bits?)
Do you really need to measure both the positive and negative voltages. Will measuring only the positive give you a close enough number?
How many voltages do you want to measure? Do you need to measure all at the same time?
Do you know in advance when this pulse will happen?

 

Only a vague hint at the frequency, which means no ideas as to what amplifiers to use.The cheating trick is to use both a positive peak detector and a negative peak detector and then determine the difference. but is it ten kilohertz or ten gigahertz? What is the estimated amplitude? And is it all Sine waves?