op-amp signal delay

Seems that any "signal processing" device gives a delay; even the infamous gain ~0.98 emitter follower. Shoot, get even more symple...a wire gives delay (transmission line).

I agree with Jim T, meaning that TI totally overlooked the required delay device.

How about using a slower, precision opamp and putting an inductor in series with the input resistor, to do the integration at high frequencies? That might work if you need a final correct integral but don't care too much about the path.

((is she good looking?))

And all-pass, in the time domain, is a fixed delay. ...Jim Thompson

That is not true in general. An all-pass has a flat frequency response, yes, but its phase response may be quite different from being linear, so it won't look like a fixed delay. In fact, it's exactly that which makes all-pass sections useful to linearize the phase response of filters that are not normally well-behaved in that respect.

Jeroen Belleman

That's the open loop delay, and course this amp has mucho internal poles that are significant in the kind of time frames under consideration here. It's the feedback that complicates things.

Can I put a time delay into LTspice?

It would be nice to be able to model thermal loops.

George H.

Wow! Thanks (I'll check out your opamp model too...)

George H.

OK, can you make an all-pass with linear phase? (It would probably be OK if it was just over an order of magnitude or two.)

George H.

Not too bad for "government work"...

Actually quite good for frequencies well below 1/(2*pi*RC) ...Jim Thompson

Sure, a really long transmission line is one approach. I usually just do a Bode plot, and tweak the real system afterwards. (Same with PLLs.)

Cheers

Phil Hobbs

You can also do it with the absdelay() function (see the help).

Behavioural sources can use the special variable 'time', e.g. you can have a BV with V=(time > 1)*(time-1), which will give you a unit ramp starting from zero at time=1.

Cheers

Phil Hobbs

The transmission line model works, but be sure to ground both ends! It acts as if there's an ideal transformer inside.

The other thermal model is a string of RCs, sort of a super-lossy delay line, approximating thermal diffusion. As Phil says, measure the open-loop response of the heater+sensor system, and then hack a circuit that behaves similarly.

Approximately...

Electrical == Thermal

close enough, usually.

Thermal diffusion often consists of... just that, diffusion; the f^1/2 term is analytically painful. Same goes for skin effect (current diffusion, essentially).

You can approximate it with a chain of RL/RC networks (R+X arrayed in parallel, or R||X arrayed in series, whichever fits your system best) where the break frequencies (when Rn = Xn) are spaced by regular factors (e.g., Fn = F(n-1) * 3) and the resistances are spaced by some factor away from that (Rn = R(n-1) * sqrt(3), hence giving an average f^1/2 or whatever profile). The poles at the ends need to be spaced a bit wider and/or higher/lower in resistance to smoothly transition from the f^1/2 region to the flat or all-pole or whatever characteristic surrounds this region.

Example:

Tim

For a faster "amplifier", try something like an ECL bus receiver, similar to the ON MC100LVEL16DG.

Do you think one can put feedback around that beast (my application is an integrator)?

I wonder what its gain is.

It's not designed for it.

Used to be about ten volts per volt for the previous generations of line receivers.

Unlikely. It will oscillate/

Modern ECL gates run in the ballpark of 100.

Probably an integrator (pure capacitive feedback) would not oscillate, so I'd say 'yes'. There's some concern, though, that a 'bus receiver' function might be enhanced by a small amount of positive feedback, and the spec sheet might not mention that little 'feature'.

You'll want to ensure that the inputs are biased in the linear range at all times, of course (the Vbb pin will help).

If the differential outputs aren't identically loaded, there will be some current modulation at the power pins.