Op-amp peak detector with gain

You need the diode in the feedback loop to avoid seeing a significant effect from the forward voltage drop.

Graham

(Previously

about

have

effect

Thanks for the reply Graham. In my circuit, won't the first op-amp's output just rise to about 0.6V higher and automatically compensate for the diode to balance the circuit, as it would do in a standard (unity gain) peak detector as shown below:-

IC=LM324 +-------------+---------+ | | | | | |\\ | | |\\ +--|-\\ | +--|-\\ 1N4148 | >---+---o Vout = Vin(pk) x 2 | >-->|--+---|+/ Vin o-------|+/ | |/ |/ 0.01uF--- --- | === GND

would

as

A typo in my last post with the unity gain peak detector. Vout=Vin(pk), NOT Vout = Vin(pk) x 2

... Steve

Yes, silly me ! I haven't actually seen that configuration previously so I missed that point.

You'll need to consider drift due to the input bias current of the 2nd op-amp btw.

Please, please don't use am LM324 either. What are your specific requirements for bandwidth and precision ?

Graham

would

to

don't

circuit, as

op-amp

requirements

This circuit will only be used for test purposes, and precision isn't too important. I'm developing a 2-quadrant 12V,

I'd have thought that the LM324's inpur bias current will cause quite a high rate of droop ( or rise ) in the peak detector voltage on the cap. Yes, it's pnp input. So, in the absence of an input, your output will float toward the supply rail.

Are they really OK with inputs as low as ground on a single supply btw ?

Graham

too

for

sensing

gain

parts

high

it's pnp

supply

Thanks for the continued interest Graham. The LM324 datasheet claims that the common-mode input range extends to the negative supply rail, as does the output if it is not heavily loaded. It also claims fairly low input bias currents, about 100nA for the '324A. This is why I'm using the '324, (besides the prev mentioned fact that it's the only quad that I have on hand). However, I don't want the output to rise in the absence of inputs. This could potentially be a problem. I was probably going to add a high-value, (1M to 10M, yet to be determined), bleed resistor across the cap. Do you think the voltage across the capacitor would still rise due to the input bias current, even if I used a 10M resistor, allowing for the leakage of the cap?

... Steve

It appears that mt ISP is having problems. 10 minutes and my reply hasn't appeared, so here it is again:-

too

for

sensing

gain

parts

high

it's pnp

supply

Thanks for the continued interest Graham. The LM324 datasheet claims that the common-mode input range extends to the negative supply rail, as does the output if it is not heavily loaded. It also claims fairly low input bias currents, about 100nA for the '324A. This is why I'm using the '324, (besides the prev mentioned fact that it's the only quad that I have on hand). However, I don't want the output to rise in the absence of inputs. This could potentially be a problem. I was probably going to add a high-value, (1M to 10M, yet to be determined), bleed resistor across the cap. Do you think the voltage across the capacitor would still rise due to the input bias current, even if I used a 10M resistor, allowing for the leakage of the cap?

... Steve

The 10M will limit the effect to 100.10^-9 x 10.10^ 6 = 1000.10^-3 or 1V !

You need a fet input op-amp.

Graham

isn't

currents

I'm

more

the

(maximum)

is

(Also,

a

Yes,

the

?

input

'324,

inputs.

to

!

When you put it like that:- (10^-9 x 10.10^ 6 = 1000.10^-3 or 1V !) Incidentally, I hadn't realised that the input stage of the '324 was PNP. I don't have room for 4 single FET input op-amps. I could use a pair of CA3230's, if I can find them. Off the top of your head, do you know of a quad FET input op-amp with a common mode input range and output that extends to 0V, (in a 5V single-supply circuit)? I'll also do a quick Google ... Steve

You specifically need 4 ?

Unless DC precision is an a big issue, I'd look at TI's offerings. I'm sorry but I'm not that familiar with the single supply types.

Graham

PNP.

extends

sorry but

Yes, I do need four - two for the precision rectifier and two for the peak detector. In my prototype, I used separate CA3130E's. I just checked out a few of the LFxxx series and the LT1058, but their input/output ranges aren't suitable. Thanks for the tip and for all of your help. I'm grateful. I'll check out the TI site in the morning. I'll also check back here in case someone has made any suggestions. It's getting late here, (NSW Australia), so I'd better hit the sack.

... Steve

(Previously

about

have

We've already covered this point in the earlier posts. The diode will not introduce any error. ... Steve

(Previously

about

have

I should have added that a fast, low-leakage would make a difference, though. ... Steve

message

would

I left out the "diode". I'm getting tired. ... Steve

You'll eventually need to discharge that cap, too. Some opamps, like the 324, have positive input bias current so will charge the cap until it hits the positive rail.

John

message

would

to

don't

Yep, that's the biggest drawback. (We also covered this earlier). I've spent the last hour or two searching for a suitable rail-to-rail input and output FET-input quad op-amp. The AD8544 is looking good so far, with the AD844 running a close second.

... Steve

SDC a écrit :

No need for a 'fancy' opamp. You can easily make it like that:

+-[10K]-+----[10K]----------------+----------+ | | | | === | + | |\\ | GND | |\\ | `-- |-\\ | +--|-\\ 1N4148 |/ | >---+---o Vout = Vin(pk) x 2 | >-->|--+---+---|\\ .---|+/ Vin o-------|+/ | | v | |/ |/ | | +---' 0.01uF--- RL | --- | 1M | | | === === === GND GND GND

Or better if you have enough supply:

+-[10K]-+----[10K]-----------------+----------+ | | | | === | 2N7000 + | |\\ | GND | |\\ | `---|-\\ | +--|-\\ 1N4148 |- | >---+---o Vout = Vin(pk) x 2 | >-->|--+---+---||

to

don't

Hi Fred, I only have a 5V supply for this circuit, not high enough for a MOSFET, so the first of your suggestions is more suitable. It looks good to me. I don't know why I didn't think of it. It's almost 2am here, so I'd better get to bed and continue tomorrow. (I said this a couple of hours ago) Thank you for your help. ... Steve