2MHz Peak detector

[guusl]

Dear forum members,

I'm trying to measure the amplitude (between 10mV and 2V) of a sine wave in a frequency range of 100Hz to 2MHz. I found this circuit that does this. I'm using different opamps though: the THS4631 for the left one and the OPA124 for the right one. These give me a bandwidth of 300kHz before the output shows an overshoot and then drops with higher frequencies. I think this has something to do with the opamp driving a capacitive load.
I've tried different capacitor-values and putting a resistor in series with the output of the left opamp (I read this should compensate for the capacitive load) but nothing has helped enough to reach 2MHz.
The THS4631 has a gain bandwidth product of 210MHz and a high slew rate (900V/us), I thought this would be enough to reach 2MHz, is it?

Are there any modifications to this circuit that would improve the bandwidth? Is there another circuit that would allow me to do this?

Thanks

[Kleinstein]

In this circuit the first OP-amp needs to be quite fast to overcome the read zone of about 2 diode drops. So the slew rate can be critical. The slew rate of FET OP-amps are measured at a rather high input differential - so you won't see the advertised 900 V/µs.

There a similar circuits that lower (e.g. reduce to about 0.2 V) the dead zone somewhat. Another point may be the transistor - there may be some delay in turning off. The more typical circuit just uses a diode, and needs a fast diode, like a small schottky diode. Also the capacitor may have to be rather small (e.g. 10 pF range). If needed a second similar (bus slower) stage can be used to cope with drift due to the small capacitor.

The alternative is using a simple diode - capacitor circuit without OP-Amps for feedback, but with compensation for temperature drift and maybe nonlinearity. At low frequency this is less accurate, but it's much faster.

[GK]

Accurate peak detection at MHz frequencies isn't trivial. If you're just measuring clean sinewaves though maybe an RMS detector IC (plenty available) would better suit the application, working on the fact that the peak amplitude is 1.414 x the RMS value.
 

[guusl]

Hi, thanks for your inputs. I have more or less given up the idea of using only opamps. I've found an article however that would promise me a high frequency peak detector by using a comparator: http://www.edn.com/design/analog/4347624/High-speed-peak-detector-uses-ECL-comparator Any thoughts? I also can't seem to figure out what the purpose is of the right part of this circuit (R3, C3, R4, R5) I know it introduces an level shift on the signal but I can't seem to figure out why that would be needed. 

[yramgu]

I think you'd be better off using a dedicated IC. Analog Devices has some (AD8307, etc...), Maxim (MAX9932, etc...), Linear Tech as well. Plenty of choice and much easier to do. As GK said "classic" implementation of peak detectors at >MHz frequencies is far from being trivial

[moffy]

Looked at the same problem a while back. It isn't that easy. The article you are referencing is one I looked at too, and seems to be the fastest. The level shifting is because the output is ECL compatible: approx Voh = -1v and Vol = -1.8v