Reply-To: "Kevin Aylward"
From: "Kevin Aylward"
References: <email@example.com> <firstname.lastname@example.org> <email@example.com>
Subject: Re: Opamp GBW tempco ?
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
Date: Thu, 12 Sep 2002 16:19:30 +0100
NNTP-Posting-Date: Thu, 12 Sep 2002 16:19:31 BST
"fred bartoli" wrote in message
> Kevin Aylward a écrit dans le message :
> > "fred bartoli" wrote in message
> > news:firstname.lastname@example.org...
> > >
> > > Kevin Aylward a écrit dans le message :
> > > dnGf9.1410$ZB3.email@example.com...
> > > >
> > > > "fred bartoli" wrote in message
> > > > news:firstname.lastname@example.org...
> > > > > Maybe a question for Kevin or Jim ?
> > > > >
> > > > > For my 100kHz IF amplifier that has to have a really (really
> > :
> > > > <0.01d)
> > > > > stable phase rotation at the (constant) IF frequency, I'm
> > evaluating
> > > > the
> > > > > sensitivity of all the components.
> > > > > One point I don't have answer is the tempco of opamps.
> > > > > The ones that I want to use are in the 100/200MHZ range,
> > input
> > > > (low
> > > > > voltage noise).
> > > > >
> > > > > Any order of magnitude for this tempco ?
> > > >
> > > > If the data sheet don't spec it, I reckon you have to bank on
> > 50%
> > > > or so at least as an open loop gain variation over temperature.
> > > > might be better, it depends on how well all the gm stages have
> > > > biased with a PTAT current.
> > > >
> > >
> > > That is about 0.2% to 0.5% per °C and is pretty large regarding my
> > needs. So
> > > that means that I probably will have to measure it and account for
> > the
> > > amplifier, and the situation may not be so bad if this tempco is
> > fairly
> > > repeatable.
> > > Do you have some ideas about that ?
> > >
> > It might take a bit of digging. Off hand, I have not looked much at
> > temperature variations of the Early effect.
> > Note that the GBW is also highly variable from unit to unit. e.g. a
> > might be quoted at typically 1Mhz, but it might be 0.47MHz to
> > The open loop dc gain might easily be a vary by factor of two.
> > The whole idea of the high open loop gain is that on closed loop the
> > variations are reduced tremendously.
> Indeed, but my app is 100kHz centered. Also, what I was interested for
> the temperature dependance, the case to case dependance being
> once for all.
> So, unless I'm missing something, Early effect will not have first
> effect on gain at these frequencies, but rather the main pole setting
> capacitor, which I guess is pretty stable, and the input stage gm,
> dependant on bias current stability as you stated, and also on
> I'm well ready to admit the 20%-50% range you gave over the full
> range, but one thing I find amazing is that it implies the same change
> bias current, which I find not good at all, even for a moderately
> design. Did I missed something ?
The bias arrangement in many ics is performed by a generation of a
single PTAT current that is mirrored to all devices, that is a current
that is proportional to absolute temperature. This is done automatically
by an internal band gap voltage reference, usually. So yes, the bias
current in an ic is usually dependant on current in this manner. This
actually helps some of the gain stages, since their gain varies in an
inverse manner. The 1st order correction is to keep the gain constant
with temperature. Its just that they may be other bits and bobs that are
not compensated, hence causing some net variation with temperature.
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.