Peak rectifier time constant problem

Dear Lab members,
this is a rectifier circuit that is supposed to behave PPM-like:

(borrowed from the LM3916 datasheet, but it seems to be a standard circuit)

I need a bit of gain (25, if possible) in this application, though, which is no problem by increasing R2 and/or lowering R3 and R6, works fine in simulation. This completely messes up the decay time constant, though, and i can't figure out how / through which resistors etc. the 3.3uF cap discharges. Any ideas to get me on the right track?


;Matthias

Run the gain seperately. This is a precision rectifier with two half-rectifiers. Keep gain away from it and life wil become much simpler.

Keith

Stick another opamp in front to get your gain.

While my father could probably devise a way to do this plus gain in two (maybe one) opamp, such shenanigans belong in the days of $299 opamps. For another $0.13 (price of a half-chip) it isn't even worth calling him to ask.

Also, doing 20KHz rectification well is tough enough without asking gain too.

Hmmm.... change R5 to 4K, R6 to 8K, C1 to 5,500nFd (5uFd). That should "work" but input Z will be very low, and accuracy at 20KHz kinda poor.

[quote author="SSLtech"]Run the gain seperately. This is a precision rectifier with two half-rectifiers. Keep gain away from it and life wil become much simpler.

Keith[/quote]

Just to clarify for those that might not see it immediately: it is already a full-wave rectifier with respect to current at the second op amp inverting input. The second op amp's diodes etc are for the attack/release function.

I agree that it will be much more straightforward to put additional gain ahead---3MHz gain bandwidth product op amps don't have a lot of loop gain at 20kHz to begin with. Moreover, you want to a.c.-couple where you can to avoid amplifying input offset voltages, and the standard rectifier circuit as shown makes this tricky.



Another observation: This circuit is working hard with an LF347 section driving the 510 ohm attack tau resistor, although maybe the voltage required at the output is fairly small hence making it less of an issue.

Hmmmm..
C1=1uF, R3=5k, R6=10k, R2=220k, (R1, R5=100k)

[quote author="EZ81"]......
.... and i can't figure out how / through which resistors etc. the 3.3uF cap discharges.[/quote]

T=(R2+R4)*C2

Regards,
Milan

[quote author="moamps"]Hmmmm..
C1=1uF, R3=5k, R6=10k, R2=220k, (R1, R5=100k)

[quote author="EZ81"]......
.... and i can't figure out how / through which resistors etc. the 3.3uF cap discharges.[/quote]

T=(R2+R4)*C2

Regards,
Milan[/quote]

That's so isn't it---you can leave the rectifier portion at unity gain. But with R2 at 220k it is a longer decay tau, unless C2 goes to 1.5uF and then R4 is increased to 1.1k for the same attack tau. Whatever is loading C2 may be a consideration as well.

One could I suppose also more evenly split the gain increase between the stages. Assuming you mean you want x25 more gain than the existing circuit, you could do R1 = 100k, R5 = R3 = 20k, R6 = 8k, C1 = 2.7uF, R2 = 100k.

That isn't equal noise gain per stage, but it feels about right. Would still advise an offset trim. Input Z isn't too low, especially considering that you can't drive it all that hard with that much gain.

[quote author="bcarso"]
... But with R2 at 220k it is a longer decay tau, unless C2 goes to 1.5uF and then R4 is increased to 1.1k for the same attack tau. [/quote]

Hi,
R2 is 200k in the NS datasheet. (C2 is 3u3).
Assuming the cap tolerance, 220k is a good value.:wink:

... Assuming you mean you want x25 more gain than the existing circuit, you could do R1 = 100k, R5 = R3 = 20k, R6 = 8k, C1 = 2.7uF, R2 = 100k.

Click to expand...

Sorry, but I dont see any reason for first using a voltage gain then lowering current. Also, your gain isn't even close to A=25, IMO.

Regards,
Milan

Thanks for all answers, looks like another opamp with adjustable gain (a nice extra) will do the gain. there will also be a offset adjust trimmer.

[quote author="bcarso"] This circuit is working hard with an LF347 section driving the 510 ohm attack tau resistor, although maybe the voltage required at the output is fairly small hence making it less of an issue.[/quote]

Do you have any suggestions what to look for? A very common quad opamp that can drive the output to 5v max? The load on C2 is very high impedance (ADC of a uC) All i know about opamps is that 5532 are ok and 741 are &%$§&%, so i just used what the SSL metering used...

[quote author="moamps"] T=(R2+R4)*C2[/quote]
Milan, in simulation (as i don't have a working scope) R6 and R3 also seem to influence the time constant.

[quote author="moamps"]R2 is 200k in the NS datasheet. (C2 is 3u3).[/quote]
Thanks, i mixed this up in the schematic.

[quote author="moamps"][quote author="bcarso"]
... But with R2 at 220k it is a longer decay tau, unless C2 goes to 1.5uF and then R4 is increased to 1.1k for the same attack tau. [/quote]

Hi,
R2 is 200k in the NS datasheet. (C2 is 3u3).
Assuming the cap tolerance, 220k is a good value.:wink:

... Assuming you mean you want x25 more gain than the existing circuit, you could do R1 = 100k, R5 = R3 = 20k, R6 = 8k, C1 = 2.7uF, R2 = 100k.

Click to expand...

Sorry, but I dont see any reason for first using a voltage gain then lowering current. Also, your gain isn't even close to A=25, IMO.

Regards,
Milan[/quote]

The reason is distributing the loop gain for better accuracy, and not lowering the input impedance as much.

And, my statement was that if he wanted 25 times more gain than the original circuit, those values would be a way to do it.

As far as the NS datasheet----I don't know. I was referring to his schematic, where R2 is 100k, and the change resulting from moving to 220k. Based on the last post this was evidently a transcription error.

[quote author="EZ81"]Thanks for all answers, looks like another opamp with adjustable gain (a nice extra) will do the gain. there will also be a offset adjust trimmer.

[quote author="bcarso"] This circuit is working hard with an LF347 section driving the 510 ohm attack tau resistor, although maybe the voltage required at the output is fairly small hence making it less of an issue.[/quote]

Do you have any suggestions what to look for? A very common quad opamp that can drive the output to 5v max? The load on C2 is very high impedance (ADC of a uC) All i know about opamps is that 5532 are ok and 741 are &%$§&%, so i just used what the SSL metering used...

[quote author="moamps"] T=(R2+R4)*C2[/quote]
Milan, in simulation (as i don't have a working scope) R6 and R3 also seem to influence the time constant.

[quote author="moamps"]R2 is 200k in the NS datasheet. (C2 is 3u3).[/quote]
Thanks, i mixed this up in the schematic.[/quote]

The offset trimming will typically only be an issue if you try to realize the extra gain as PRR, Milan, and I have been proposing, and should be applied to the first and possibly second existing stages.

The two input R's to the second stage shouldn't have much effect on tau because the second op amp should always have the loop closed with one or the other component of its feedback network---that is, the input is a "virtual ground". In your simulator, see if the voltage at the inverting input is moving much---it shouldn't be if all is well. So, after a burst of audio and subsequent charge up of C2, make the audio go away---the op amp should make its output go low to roughly one diode drop below ground, in order to keep the inverting input at about ground potential.

As far as higher output current garden-variety Bifet amps, I don't know. Probably it is not really a big deal, because you will just get a bit of current limiting at worst which will effectively add to the attack time.

It would be good to know what the input Z of your A/D actually is---some might not be that terribly high-Z. Maybe yours is.

Thanks, i will just stay with the LF347 and then try if offset trimming is necessary.

It would be good to know what the input Z of your A/D actually is---some might not be that terribly high-Z. Maybe yours is.

Click to expand...

The ADC seems to be very high impedance. I just connected a 10uF foil cap to two inputs in parallel, loaded to 5v and the voltage drops about 1LSB (=~5mV) per 10s while sampling constantly :thumb: