op-amp signal delay

Measurement adjusted for cable delay, of course?

Not quite the infamous "Damn Fast Buffer", but still pretty kickin'.

Since you mentioned it, how does the SPICE model compare?

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

I don't know the details, but a basic transistor delay (Rbb and Cmiller) doesn't do it, it takes a series of them. A series of RC delays act like a lumped-constant LC delay line, except with attenuation (but the gain covers that). Some contribution comes from wiring pathlength, too, but that can't be a big part of 0.7 ns, unless it were in the probe wiring.

The current-mode amps don't have a constant GBW like normal amps do. They have higher BW at high gains than you'd expect. The slew rates also tend to go up for large swings. They do need a lot of swing at the inverting input: that voltage swing, into the thevenin equivalent of the feedback resistors, creates the current that drives everything internally. A diff pair has a limited output current, whatever the current source furnishes. A current-mode amp gets its first-stage signal current from the resistors on the inverting pin, and that increases with the error.

They also get a lot of their speed by brutally loading the NI input signal at larger swings, as I have learned the hard way. Well, "there are no fast, high impedance signals."

I don't have a clue about the time delay.

We use the THS4303, a current-mode amp with a fixed gain of 10... the FB resistors are inside. Its closed-loop gain-10 BW is 1.8 GHz.

ps - AoE 3 is on order!

Two microwave coax cables with SMA connectors, one each in and out of the board. The direct-path trace was made by unscrewing both SMA connectors from the board and connecting them with a barrel adapter. So, yes, "adjusted" to be the same.

Manufacturers' SPICE models stink, but you already knew that. Since almost all transistors work by charge diffusion, there's some irreducible amount of real delay involved.

The slew issue is probably most of it. 830 V/us is pretty slow for a 3 GHz op amp. There's a Widlar or Pease app note from the Cambrian Epoch (like 1970) that explains that a normal two-stage op amp with pole splitting and no emitter degeneration has a maximum slew rate of about

Cheers

Phil Hobbs

The output overshoots, which suggests multiple internal poles, and that's delay. The duration of the overshoot is similar to the rising-edge delay.

Yes, you're right about CFB op-amps. However this is a VFB op-amp. Ultimately I'm making a fast inverting integrator with this op-amp, so CFB is out, right?

Very good! Did you notice Amazon dropped the price dramatically (25% off), then raised it dramatically (now 5% off). They go all over the maps with prices.

Fortunately they guarantee that pre-orders get the lowest price that it's offered at between the times of order and shipping. (Mine's on order too.)

Cheers

Phil Hobbs

-Lasse

Oh, then, never mind.

Ultimately I'm making a fast inverting

That's interesting. It could be done, maybe. Ground the + input, make an inverting amp with a cap in series with Rf, and compensate downstream. Probably more trouble than it's worth.

I buy all sorts of stuff from Amazon. Cajun spices. Greek sour cherry preserves. Chocolate in large shipments.

Actually, I misspoke, the way one does it with a CFB is to place the input resistor and integrating feedback capacitor in the normal way, but then add the required "-input" resistor in series to the R - C node, setting the CFB transconductance gain. If this VFB isn't fast enough I'll consider that.

See Op-Amp-Config.zip on the Device Models & Subcircuits Page of my website.

Excess Phase = Time Delay ...Jim Thompson

A delay has to be at least two pole... RC reacts right away.

(Wow Win is back, and SED has been civil, big grins :)

Naaaah! Delay is _physical_ transit time. It manifests itself (electrically) as excess phase (AKA ALL_PASS).

It takes discipline to ignore the bait that the cretins throw out >:-} ...Jim Thompson

OK. (you don't have to yell :^) that's certainly right. I was trying to do this spice sim of a thermal control loop and two poles didn't have enough time delay.

Oh I totally appreciate the effort. (I'm always throwing away reply's..."Do I need to send this?") I look down the list of threads and they are all things I want to read... few flame wars.

(Now if we could get Bill S. to hold back a bit....)

Let's talk about electronics, George H.

Study my model, it's educational ;-) ...Jim Thompson

Jim Thompson being a leading cretin, when it comes to incivility and generally barbaric behavior. I don't think that anybody else who posts here makes a habit of announcing that he's planning to kill people.

On a sunny day (5 Feb 2015 14:29:03 -0800) it happened Winfield Hill wrote in :

Amazon is evil, they are largely fake, yesterday I was looking online for some cookies, mazomne page pooped up with them cookies but big sign 'not in stock', and I am sure they never had...

So what they do is they fake search results and then artificialy create a web page, and sort of set a trap to let you sign up. And the mamazone boss works for NSA.

I do not and will not order anything from them, not even for free!

Thanks, Lasse. Just ordered my copy at co.uk :-)