Xref: utzoo rec.ham-radio:18086 sci.electronics:10350
Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!rice!uw-beaver!milton!whit
From: whit@milton.acs.washington.edu (John Whitmore)
Newsgroups: rec.ham-radio,sci.electronics
Subject: Re: How does one build LOG AMPS (>60dB dynamic range)
Summary: National Semiconductor AN-264 explains it
Message-ID: <2154@milton.acs.washington.edu>
Date: 27 Feb 90 04:29:42 GMT
References: <596@massey.ac.nz>
Reply-To: whit@milton.acs.washington.edu (John Whitmore)
Organization: University of Washington, Seattle
Lines: 37

In article <596@massey.ac.nz> GMoretti@massey.ac.nz (Giovanni Moretti) writes:
>I've build a spectrum analyser (0-100MHz) that has a 10.7MHz IF (from
>HAM-RADIO years ago) and it works quite well but I'm not happy with the
>log-amp/detector.  

	Most log amps use the relationship that the base-emitter voltage
of a transistor is proportional to the logarithm of the emitter current
(sometimes approximated as the collector current).  A good low-noise transistor
will follow this "ideal diode" curve for about eight orders of magnitude
in the collector current (power ratio 160 dB).  Routine use of this
effect is complicated by the temperature dependence of the equation,
	Ie= Is * exp( q* Vbe/( k * T) )
and a good logarithm amplifier either needs a second (matched) transistor
at the same temperature to set the operating point, or a thermostat.
The National Semiconductor Linear Applications databook has one
thermostat-style solution, in AN-264. The thermostat they use employs
three transistors, with one transistor doing the logarithm conversion,
a second measuring the temperature, and a third acting as the heater.
The thermostat approach does NOT require matched transistors.
	The second approach, with matched transistors is slightly
less accurate (mainly because of mismatched emitter resistances at
elevated currents), but works at lower total power dissipation.  I think
the National databook has one of these, also; AN-29, Figure 23.
	I would suggest using the cookbook circuits, since at the lowest
current ranges these are not terribly easy to frequency-compensate (i.e.
the op amps will likely oscillate), and these circuits have
already been debugged.
	The second part of your problem is the precise rectification of
the IF-frequency signal; all I can think of is a good hot-carrier
diode, and keeping 60 dB of accuracy there is not going to be easy.
Good luck.
	Your local National Semiconductor distributor (or the
representatives in Australia: telephone (03) 729-6333) can
supply the application notes I quoted above.

I am known for my brilliance,                John Whitmore
 by those who do not know me well.