LTC6240

"Vladimir Vassilevsky" schrieb im Newsbeitrag news:uTTdj.32947$ snipped-for-privacy@newssvr21.news.prodigy.net...

Hello Vladimir,

This is very normal. Just take a look to both datasheets regarding noise. Both opamps have the same curves for v-noise and i-noise versus frequency.

This doesn't mean you will have any problem.

All the RR-amps have very low open loop gain, if the output hits the rails.

I tried the demo circuit and it correctly simulates Vnoise and overshoot.

There is nothing strange in the specs. See my comments above.

Best regards, Helmut

because

is

This isn't a typical behaviour for an opamp. Besides, what could be the cause for this effect?

Although the curves look similar in the general shape, the OPA132 is much noisier at the low frequencies then LTC6240.

This means that the internal structure is not very trivial and the simplified approaches (like Bode plots) can't be used to estimate the stability.

and

LTC6240 behaves very differently in the terms of phase if the output is closer then 1V from a rail. There are some plots in the datasheet. Heck, I also can see it in a real circuit.

Sad smile :) It does not account for the peculiar phase response and for the nonlinear phase effects. This is important for me; the goal is the ultimate performance.

has

There is. Look more attentively.

Vladimir Vassilevsky DSP and Mixed Signal Consultant

If by dual channel you means both sexes at the input stage, that is not the case. You can tell by the limits on VCM.

Not relevant to your situation, but perhaps relevant to anyone else coming across this post is that you should buy rail to rail input capability if you don't need it. I'm leary of such parts since the gain paths do change with CM.

I meant hipass + lopass channels.

6240 is advertized as RR out, not RR in. However the RR out can cause problems, too.

VLV

schrieb im Newsbeitrag news: snipped-for-privacy@t1g2000pra.googlegroups.com...

Hello,

The LTC6240 part is "only" a RR-output amplifier, but not a RR-input/output amplifier. This results in one Mosfet differential pair at the input stage and so it avoids the common changing offset problem of all RR-input/output amplifiers.

Best regards, Helmut

RR = rail to rail

Generally rail to rail output is not a problem since loop gain falls as the output reaches the rails. This part has a phase splitter of some sort to give a true push-pull output. There could be phase shift in the splitter that makes the op amp a bit funky.

Has the state of LTC apps degenerated to the point where it is not worth a call?

Newsbeitragnews: snipped-for-privacy@t1g2000pra.googlegroups.com...

I put in the "not relevant" in my post because I knew the part was not rail to rail input, just rail to rail output. However, you can't state this often enough. Don't buy what you don't need. Less is more.

I wonder if the output stage transconductance is varying enough with output voltage that the right hand plane zero is causing problems. You could try to resistively loading down the output and see if the problem goes away.

Vladimir Vassilevsky a écrit :

It isn't for a BJT input stage but for a fet input stage it is.

That current noise is essentially capacitive. Look: at 10KHz it's about 10fA/rtHz and 7nV/rtHz (5nV/rtHz per input stage transistor) which is about a 500k impedance or 3.18pF GS capacitance which is about right for the stated CM and diff input capacitance.

Capacitances increase with lower voltages.

What kind of ultimate performances are you after?

I guess not. You have a simplified diagram, and I don't see anything shocking on the performances WRT to the given architecture.

--
Thanks,
Fred.

Excuse my silliness, but how exactly Cgs is related to the current noise?

• 2.5 Vpp output

• DC to 50kHz @ 0.5dB

• Rout < 500 Ohm,

• Total noise at the output: ~ 0.35uVrms @ 0.1....100Hz

• Drain no more then 5mA.

• Gain ~1/2, reasonable DC accuracy.

VLV

the

Thank you for the tip; I can see now. The noise of the DS resistance is capacitively coupled to G. So, at high frequencies the coupling is approaching something like Cdg/(Cgs + Cstray), whereas there is essentially no coupling at low frequency.

VLV