The Cyber-Spy.Com Usenet Archive Feeds Directly
From The Open And Publicly Available Newsgroup
This Group And Thousands Of Others Are Available
On Most IS NNTP News Servers On Port 119.
Cyber-Spy.Com Is NOT Responsible For Any Topic,
Opinions Or Content Posted To This Or Any Other
Newsgroup. This Web Archive Of The Newsgroup And
Posts Are For Informational Purposes Only.
Reply-To: "fred bartoli"
From: "fred bartoli"
References: <email@example.com> <firstname.lastname@example.org> <email@example.com> <3D8235D8.DD57FF10@earthlink.net> <firstname.lastname@example.org> <email@example.com> <3D883297.9EEF28F7@earthlink.net>
Subject: Re: Opamp GBW tempco ?
Date: Wed, 18 Sep 2002 14:52:37 +0200
X-Newsreader: Microsoft Outlook Express 5.00.2314.1300
Organization: Guest of ProXad - France
NNTP-Posting-Date: 18 Sep 2002 14:51:50 MEST
Robert Baer a écrit dans le message :
> Ian Buckner wrote:
> > "fred bartoli" wrote in message
> > news:firstname.lastname@example.org...
> > >
> > >
> > > Robert,
> > > Opamps for my app may not be as silly as it looks at first.
> > > My gain stages are 30dB and 10dB, I have 3x30dB+3x10dB switched on
> > or off
> > > depending on the wanted gain and I'm seeking 0.01dB, 0.01° accuracy
> > over the
> > > full gain range for the IF-detector block. This is about a 30ppm
> > requirement
> > > for the the phase accuracy. Not a trivial task for a discrete
> For 0.01dB gain accuracy, over temperature would require rather
> precision resistors that also have rather low tempcos.
> And that translates to expensive wire wound resistors; a specialty
> item - i vaguely remember "Kelvin" as a brand name but i could be wrong
> and if correct they mey not be in business now (that was 30+ years ago).
> As far as phase shift goes, for that kind of accuracy, one might need
> trimmers to compensate for strays and have *very* careful layout, and
> capacitors will have to have rather low tempcos also.
> I think your phase "calculation" based on GBW is in error, as the
> phase margin of an opamp VS frequency ain't linear; it tends to be
> constant from DC to some value, and rolls "near" the 3dB point (in the
> simple case).
> The phase VS freq diagram may have glitches in it (a number of wiggles
> up and down) as it tends towards the 90 degree critical point; that is
> usually not at the gain=1 point.
> Also, it may turn away from perfect 180 degree much earlier than
> expected if one attempts to derive the phase margin from the open loop
> Bode plot.
> However, i still stand by my "Bill Gates pocket lint" comment; you
> will have more phase problems due to capacitances and *time delay* thru
> an opamp.
> A single discrete transistor is far superior in that case........
In fact this is not gain/phase absolute accuracy that is much important but
rather repeatability (or stability) as a lot of things will be calibrated
The 0.01dB/0.01° requirements is because the IF amplifiers will have gain
blocks switched in or out the IF chain so that I can have precision
measurements over the full dynamic range. The repeatability will give
consistent measurements over time *and* scale (IF gain) switching (in order
to avoid gain/phase discontinuities).
If I estimate the overall necessary tempco and divide by the number of
"sensitive" components I arrive to impractically low tempcos. So there's the
need for some real time (at the scale of the slow temperature changes)
self-calibration for the phase and gain characteritics of each IF block. The
detector outputs DC voltages so having absolute accuracy is pretty easy
(either some 20+ bits adc converter, or a PWM divider (which don't rely on
components for its accuracy)). Hence, what I really need is "short time"
stability (time counts in seconds or 10s of seconds) and that greatly
alleviates the burden on tempcos and also on GBW variations.
For the added phase errors I didn't take into account in my simplistic
calculations it simply translate in additionnal partial derivatives that
worsen the situation I showed already unsatisfactory. So there was no need
to include them and this only call louder for real time cal.
Also the "phase problems due to capacitances and *time delay* thru the
opamp" (which all translate to phase errors) becomes irrelevant as soon as
those errors are calibrated out. This is not the case for some other points
- much better power supply rejection for the opamp (very important point
- lower distorsion particularly at the middle to high level I want at the IF
- better stability of distorsion vs output level variation
- parasitic inter-stages feedback paths are much lower for RC active filters
than for LC tuned ones (freq is low) and this is also important as I can
only calibrate gain stages individually
- for the same reason, and because gain blocks are switched in and out, it
is important to have the highest input and the lowest output impedances to
avoid adverse loading effects (they can't be calibrated)
All these points plead in favour of (modern) opamps.
The only point in favour of transistor could be noise in the input stage but
there's no need to be better than the DUT, and if I needed outrageously
sensitive measurements, lowpassing the lock-in output will do a far better
job, of course at the expense of speed but I can live with that for those
So I think I'll stick to opamps with all due respect to Mr Bill Gates :-)
Go Back To The Cyber-Spy.Com
Usenet Web Archive Index Of
The sci.electronics.design Newsgroup