Any links to "Are Op Amps Really Linear?" by Barrie Gilbert?

[jez]

The links for the article are all dead.... Does anyone know of a working link or maybe could email me a copy?

[jcx]

Analog Avenue: Are Op Amps Really Linear? (6/10/98) <META NAME="DESCRIPTION" CONTENT="CMPnet's network of technical information for the Electronics Professional, including EETimes, EDTN Design Center, Electronic Buyers News and Semiconductor Business
text, but the drawings aren't available

the analysis is correct in context - but not relevant to modern audio op amp application given the availability of much higher GBW - the input tanh distortion % drops as the square of the ratio of signal frequency to GBW

it is quite practical to use 50x higher GBW op amps than Gilbert's example - giving 2500x less of the distortion he discussed

the article is also misused by John Curl as "evidence" that Otala's PIM theory and "flat open loop gain" over audio frequencies acquired a new "convert" - not so - just read all of the article, not selectively as Curl has

the search may also lead to some of Walt Jung's early articles - he did initially embrace the "flat gain" prescription from Otala but in later publications came to the correct conclusion that one can have high loop gain/low corner frequency and still have low PIM

Marshall Leach also traveled the same intellectual path - ultimately rejecting Otala's "flat loop gain" over audio frequency as being a necessity for low PIM

High loop gain (and high GBW), 2-pole compensation,
input stage linearization techniques including local degeneration, "cascomp", FET inputs, or Gilbert's multi-tanh compensation (now available in some AD op amp) all change the conditions of Gilbert's article analysis

[jez]

Thanks very much for the link jcx! I'd looked all over the interweb thingy for that....
Modern high bandwidth op-amps do indeed ameliorate these issues greatly.
In spite of the "proof" of the authors you mention (and also Cherry, Self and our own Bob Cordell who all seem to agree with the anti Otala/Lohstro line) it feels somehow intrinsically right to have a flat open loop gain to beyond audio frequencies before applying NFB.

[janneman]

Originally Posted by jez

[snip]In spite of the "proof" of the authors you mention (and also Cherry, Self and our own Bob Cordell who all seem to agree with the anti Otala/Lohstro line) it feels somehow intrinsically right to have a flat open loop gain to beyond audio frequencies before applying NFB.

Look at it this way. Take an amp with a flat OL gain of 50dB from DC to 30kHz. Certainly one that you would approve of I assume.

Now increase the OL gain at 30Hz to 80 dB, and then roll it of so it gets to the same 50 dB gain at 30kHz. (Leave all other things the same of course).
This amplifier has more gain at lower frequencies and for the same closed loop gain has better linearity at all audio frequencies. If the first amp was TIM free, so would the second amp. Yet, because the shape of the curve doesn't look right for you, you would not approve of it? Where would you base that on?

jan didden

[Samuel Groner]

Gilbert's multi-tanh compensation (now available in some AD op amp).

Which opamps?

Samuel

[jez]

Originally Posted by janneman

Look at it this way. Take an amp with a flat OL gain of 50dB from DC to 30kHz. Certainly one that you would approve of I assume.

Now increase the OL gain at 30Hz to 80 dB, and then roll it of so it gets to the same 50 dB gain at 30kHz. (Leave all other things the same of course).
This amplifier has more gain at lower frequencies and for the same closed loop gain has better linearity at all audio frequencies. If the first amp was TIM free, so would the second amp. Yet, because the shape of the curve doesn't look right for you, you would not approve of it? Where would you base that on?

jan didden

I'm well aware of the theory Jan
Doug Self takes that one to extremes!
It just "feels right" for the open loop gain to be reasonably flat across the audio bandwidth.
It becomes a real problem when amps have loads of OLG at very low frequencies like 100Hz but are rapidly running out of gain at high audio frequencies and a bit beyond.... Just where we need it for the NFB to be effective in reducing the higher order artefacts of crossover distortion.
The situation you describe of course doesn't suffer from this but I've seen far too many that do.
I'm with John Curl on this one

[SY]

And yet, when you look at the THD versus frequency of nearly any half-decent amp (I'll exclude "designer" high end gear, which is often poorly conceived), you don't see an enormous rise near the high end of the audible range. IM, same deal. Looking at distortion residuals, the crossover distortions of most commercial gear are minuscule. So the theory does work in practice, "feelings" and 1968-vintage anecdotes aside.

[DF96]

Was it Einstein who said that intuition is merely a set of prejudices laid down during childhood? Flat OLG may "feel right", but it is likely to be wrong. One possible get out: if distortion products at high frequency can be made more palatable by the simultaneous presence of low frequency products (e.g. via masking) then flat OLG may sound better. If this is the case then it must be understood that this is an example not of 'better', but 'worse in a more pleasing way'.

Rapidly running out of gain at high audio frequencies is a problem, but it has nothing to do with open loop bandwidth.

[janneman]

Originally Posted by jez

[snip]The situation you describe of course doesn't suffer from this but I've seen far too many that do.

Examples, perhaps?

jan didden

[jcx]

Originally Posted by Samuel Groner

Which opamps?

Samuel

AD8099:
John Curl's Blowtorch preamplifier

http://www.google.com.ua/patents/dow...erview_r&cad=0

I believe the ADA4898 also uses the same input stage linearization - closer reading suggests it isn't strictly Gilbert's multi-tanh input but gives a related flattening of the input gm