High frequency peak detector....

hey guys i am making a sine wave peak detector.....and have made the following circuit....

https://www.circuitlab.com/circuit/y453h7/precision-active-peak-detector/

It is working fine at 1KHz frequency.....but at 1MHz....a lot of attenuation of signal happen......
so when I send 1Vpeak sin wave of 1MHz the output shows approx 0.478 V dc
but if 1Vpeak at 1KHz is fed....the output is around 0.98V...

i am not able to get the problem...please help....

You need a faster opamp but even then you will probably need a faster circuit, I cannot find an exact circuit at the moment (and cannot draw one on my phone) but I favour dual opamp designs which prevent the opamp going into negative saturation and hence slew rate issues, a bit like **broken link removed** but without the extra JFET circuitry. I hope that helps you find a better circuit.

Keith

Try adding a .1 MF cap across C1, your one might not be working well at 1 MHZ. Also what feed back resistor are you using? . The TL 082 is not really a high frequency amplifier, so it might not be producing all that much at its output. You might be able to fiddle the response a bit by attenuating the input at low frequencies and not at high frequencies. Do this by feeding your input signal into a 10K resistor to the input and connecting another 10K to earth from this input. Reset the low frequency gain by means of R2 to something convenient. Go to 1MHZ, and put a small capacitor across the series 10K to raise the gain at this frequency, Try 15 PF as a starting point, adjust this cap to try and get the high frequency gain OK. It will have no effect on the response at 1KHZ.
Frank

here are the simulation results of slightly different configuration...

working fine at low freq (10KHz)


but same circuit is not producing output for 1MHz signal....


In 1MHz case....how can a signal be low....I am using CA3140(4MHz OpAmp) and 1N4148 fast switching diode...

I suspect, as I said earlier, that you need a better circuit which avoids the opamp going into negative saturation. Have you looked at the effect of opamp slew rate in your circuit?

Keith

- - - Updated - - -

Just to add - look at the opamp output to see what is going on - it should give a clue to the problem.

here are the circuits at both frequencies....low(10KHz) and high(1MHz)

10KHz...

Good output : twice of input, since the gain is almost 2 according to non inverting op amp amplifier...

1MHz...


as we can see that output in later case has decreased to a very small value.....

A faster opamp would probably be better.
Smaller resistors may help too. e.g. change the 10K resistors to 1K.
The amplitude of the input signal is only 100mV. Does it work better at higher levels e.g. 1V or 2V?

Okay...Now I have changed the OPAMP to a 1GHz opamp....OPA695 and things are looking a lot better now...But now there is a error in amplitude...and this is increasing with increase in input signal voltage....

Input signal = 100mV Output voltage = 93.8mV error=6.2mV


Input Voltage = 3V Output voltage = 2.6V error=400mV


so why the error is increasing with input voltage....????

- - - Updated - - -

I think that the error is coming because the opamp is mostly in saturation region and it might take time for it to reach input voltage level when required....

here is the picture showing op amp output on CRO...


if the problem is this itself, then is there any way to keep the op amp away from saturation region...???

I did some quick simulations on two simple architectures of peak detectors. To achieve better precision in high speeds its good to use double-diode diode structure as is in the following picture (bottom-left)



As the simulation results show, precision PD is more reliable in higher frequencies and almost exactly follows the peak value. Still it should be noted that if capacitance of C1 is high then also this structure can't follow the peak value exactly and there would be error between the real peak and detected one. So, don't forgot to change the C1 and R1 values accordingly.

I am unhappy about R1 (and C1), R1 only discharges the C1, at the moment T = 10 microsecs, Why can't it be 100 microsecs, then R1 can be 100K, so less "droop" between the incoming RF peaks. Likewise the current required to charge C1, results in slew rate errors, so the value of C1 should as low as possible, consistent with having a good time constant compared to the incoming RF pulses. I wonder how the circuit would perform if D2, was replaced by the base emitter junction of an emitter follower using a medium power RF transistor (2n3866).
Frank