Basic Op Amp closed-loop gain question

An opamp has a specified gain-bandwidth product (high frequency where the gain=1) and a specified open-loop voltage gain, the gain at very low frequencies or DC. Assuming a usual unity-gain stable (not decompensated) amp, you can draw a straight line 1/f curve starting at Ft (gain 1) down to the huge low-frequency gain.

A common opamp might have Ft = 10 MHz and Vol=100 dB, or 100K voltage gain.

Some data sheets graph that for you. The LT Spice ideal opamp models let you plug in the gain-bw and the open-loop lf gain.

Given an open-loop opamp model, you can then add feedback networks.

Okay, thanks, I think this is the information I was looking for. Two questions:

(1) Is the GBP temperature independent? I see — on the data sheet — that open loop voltage gain is temperature dependent, so I am wondering about that.

(2) Can you clarify what is meant by "1/f curve"?

A_Vol is temperature-dependent, but GBW basically isn’t. This is because it’s set by local feedback around the gain stages, so what moves with A_Vol is not GBW but rather the low frequency corner, where the dc gain starts rolling off.

If you can turn up a copy, National Semi’s app note LB-1 is a good read. IIRC they also had an early app note called “the monolithic op amp: a universal “ somethingorother…maybe AN-4? Anyway, it has a lot of these sorts of things.

The open-loop gain rolls off like 1/f, i.e. 20 dB/decade.

Cheers

Phil Hobbs

PS: The LF357 is one of my favorite parts, on account of its uniquely low input capacitance and tolerable noise, but (a) it’s decompensated, so it isn’t well-behaved at unity gain, and (b) it was discontinued like 20 years ago.

Thank you for the information. So, when you say "straight line 1/f curve", do you have to use logarithmic scaling or something to get the straight line?

Truth is, I'm actually using LF356 op amps in my project — an analog computer. But the data sheet didn't show an open-loop voltage gain graph for LF356, so I felt like I had to modify my question accordingly.

A THAT analog computer, which I'm planning to buy soon, uses TL074H op amps for the integrators.

Probably a little, but we usually assume that any opamp circuit has a lot of gain to spare and the circuit behavior is dominated by the feedback network.

It means that open-loop gain is inverse on frequency. That's equivalent to -6 dB per octave.

Like this:

Some opamps don't have as nice a frequency:gain curve but most do.

Provided that you stay away from high frequencies the behaviour of most decent opamps today is pretty well behaved for sensible feedback loops.

Fig 24 of this datasheet gives the open loop gain of an LF356

What sort of frequencies are you intending to run it at?

There is a compromise of input and output impedance - although unity gain buffers on all outputs can make things a lot more nearly ideal.

Yes, if you remove the log function then both axes need to be logafied*

• Yes I know that's not a real word Bill.

In that case I'd go with LF156 data.

I wouldn't worry too much about what the data says for the op amps it uses. op amps go back a long way.

Even further than that. 741 was the first one that was dead easy to use.

The uA709 in 1965 was the first one that really was worth using - it just needed a bit of compensation. Amazingly Digikey still stock them!

A bit like the ubiquitous 555 they are still in use today.

I confess I was surprised where I found the 709 datasheet. I was expecting to find it in a historical archive somewhere.

Well, actually that figure only lists the LF155, LF156, and LF357. I suppose the LF356 is very similar to one of those...?

With my current analog computer, using 10nF caps and 10KΩ resistors for

My current analog computer does not use any unity gain buffers, since it was based on the Grappendorf schematics, but since then I have been wondering if it would be good to have one after each computational module. The THAT computer uses unity gain buffers after most computational modules, with the interesting exception of the integrators.

Back in the palmy days, National Semiconductor used a lot of part numbers like that. The LF156 is the military version (-55 to +125 C), the ‘256 is industrial (-40 to 100ish), and the ‘356 is commercial (0 to 70).

Life has got more complex since then, as have part numbers.

Cheers

Phil Hobbs

This one may have been a little harder to use because you would also need to light up your valve/tube filaments/heaters.

The National Semiconductor numbering convention as that the mil spec devices were LF155, LF156 and LF157 (good for -55C to 125C), the industrial spec devices were the LF255, LF256 and LF257 (good for -20C to 85C) and the commercial grade device were the LF355. LF356 and LF357 (good for 0C to 70C). As far as I know the chips were identical, but the higher grade packages were more robust, and they were tested over the appropriate temperature range after packaging.

The usual word is "logarithmic".

uA was the Fairchild Semiconducotr prefix. National Semiconductor used LM for bipolar parts and LF for the Fet-input parts.

This is more about legacy design - if it ain't broke don't fix it - than any particular virtue.

Bob Widlar did pretty well with the uA709. He did a whole lot better later in his career. His even older uA702 isn't in the same class. Around 1983 I told off one of my colleagues for using one, and he repented and found a better device, not that it made much difference to circuit.

But logafied has additional meaning. It means "done in a logarithmic manner".