From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Adjustable current source
References: <3DFFF5AD.1D45@sneakemail.com> <3E00C854.7CC5@sneakemail.com> <3E0111D8.2D1D@sneakemail.com>
Date: Thu, 19 Dec 2002 10:44:51 -0800
NNTP-Posting-Date: Thu, 19 Dec 2002 10:44:53 EST
Organization: Bell Sympatico
Winfield Hill wrote:
> Winfield wrote...
> >Mike Monett wrote...
> >> Winfield Hill wrote:
> >>> 10pA? No problem.
> >> Actually, the current may be closer to 200 mA. ...
> > Working from a 10pA noise-budget approach, let's assume 8pA from the
> > resistor and the rest (6pA) from the amplifier/reference. We get 8pA
> > from a R >= 4kT/i_n^2 = 250-ohm resistor, which means a current-source
> > sense voltage of up to 0.2A * 250 = 50 volts. That's not so bad, but
> > it will also dissipate 10 watts at 200mA full-scale, which seems just
> > a bit strange for a quiet resistor!
> Hmm, 10 watts, that resistor's going to get hot! We can't use room
> temp for 4kT, which means the resistor must be a higher value, which
> means it'll dissipate even more power at 200mA, which means... OK,
> let's use a 500 ohm resistor, requiring 100V and dissipating 20W at
> 200mA. We'll use a large 100W wire-wound unit bolted to the chassis,
> say its temperature gets up to 77C = 350K, so it's Johnson noise is
> 6.4pA. That means our opamp/reference noise component can be 7.7pA,
> and the opamp/reference voltage noise requirement is a more relaxed
> 3.8nV. So going to a higher resistor value is helpful, even if it
> does require a 100V 20W low-noise reference rather than 50V and 10W.
> > The voltage-noise budget for the reference & opamp = 6pA * 250 ohms =
> > 1.5nV, which is tough, but not impossible. Create a "quiet" 50-volt
> > reference (say 500nV), heavily filter it down to the 1nV level, and
> > use a quiet 1.0nV or less voltage-noise opamp. We'll have to create
> > a pair of tracking bootstrapped supplies for the opamp. Maybe a few
> > other messy details. A veritable piece of cake. :-)
> Mike, I'm glad your requirement was only 10pA/root-Hz, had it been
> 5pA, or 2pA instead...
> - Win
Well, the current took another jump. I found an old driver circuit at
Micro Lambda to sweep a 2 to 18GHz YIG oscillator. The 20MHz/mA spec
seems common, so it takes 900mA to reach 18GHz. The circuit uses a 3 ohm
sense resistor and AD708 op amps in a noisy circuit, and the phase noise
must be terrible:
I suspect wideband spectrum analyzers like HP, Anritsu and Rohde accept
poor phase noise in the wideband sweep ranges, then lock the YIG for
narrowband work. Or they could lock the YIG over the entire range using
any of several methods, although Anritsu shows a block diagram using a
DDS which is normally very noisy. So part of the trick might be to roll
the current source frequency response off quickly and depend on the pll
to keep the phase noise down. This would involve a tradeoff between noise
and settling time.
Anyway, I gotta get Rohde's books on synthesizers. His spectrum analyzer
has phenomenal dynamic range and close-in phase noise, and it doesn't
seem possible to achieve this with simple methods.