Practical uses for logarithmic forward voltage

Thread Starter

cmartinez

Joined Jan 17, 2007
8,253
No. The voltage across a diode is a logarithmic function of the current.
Here's a typical curve of diode forward voltage versus current.


My simulation gave 490mV and 575mV drop respectively for the two diodes.
Is there any practical use for that characteristic? I mean, are there situations in real life in which it is desirable/useful to have a logarithmic (as opposed to linear) voltage drop?
Maybe metrology or normalization of certain signals?
 

#12

Joined Nov 30, 2010
18,224
Not sure I understand you....but
Consider a bjt as a voltage follower. No resistor in the collector circuit, yes resistor in the emitter circuit, voltage output is (almost) linear compared to voltage on the base. It must be useful or it wouldn't have been invented. Right? Does this have anything to do with your query?
 

crutschow

Joined Mar 14, 2008
34,428
Yes there is. Diodes are often use to build amplifiers (log amps) where the output is a logarithmic function of the input.
They are used where the signal level can vary over a wide range (such as an RF signal) and you want to significantly amplify the smallest signal without saturating the amp with the largest signal.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,253
Not sure I understand you....but
Consider a bjt as a voltage follower. No resistor in the collector circuit, yes resistor in the emitter circuit, voltage output is (almost) linear compared to voltage on the base. It must be useful or it wouldn't have been invented. Right? Does this have anything to do with your query?
Say I have circuit #1, that requires a logarithmic voltage input ... not done in purpose but just a characteristic of that circuit... could a diode be used to apply the required logarithmic function to a linear voltage output from a previous circuit so as to condition that signal before feeding it to circuit #1?
What about the other way around? If I have a voltage output in a circuit #2, that just so happens to be logarithmic, can an antilogarithmic conversion be done using a diode (or some other component) in order to linearise its output?
 

#12

Joined Nov 30, 2010
18,224
I know how to do it with transistors but I don't know how to do it with only a diode.
Google log amplifier and anti-log amplifier. Op-amp with a transistor in the input feed or in the feedback loop.
 

RichardO

Joined May 4, 2013
2,270
To turn triangle waves into sine waves. :)

Or, more generally, to do a piecewise approximation of a transfer curve such as converting the non-linear response of a thermistor circuit to a linear output.
Note that the piecewise approximation can be first visualized as using ideal diodes but the logarithmic nature of the diodes smooths the breakpoints in the curve.
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,253
1. Easy analog multiplication. log(A)+log(B) = log(AxB).
2. Linear control of an exponential characteristic such as frequency change in octaves of a VCO.
The first one you mentioned looks very interesting... I might even have an application for that!
thanks!
 

GopherT

Joined Nov 23, 2012
8,009
The first one you mentioned looks very interesting... I might even have an application for that!
thanks!
Think about hearing quiet and loud noises. Now, how would you convert that huge dynamic range to a digital signal? How many bits of of ADC would you need? With an anti-log amp (compressor), you could compress those signals to a linear function (db/V) and store them. Then, on the output, you could expand them (log amp) to get the full dynamic range.

See Compander circuit (chip) = Compressor/Expander
http://www.datasheetcatalog.com/datasheets_pdf/M/C/3/3/MC33110D.shtml

General topic
https://en.wikipedia.org/wiki/Companding
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,253
Think about hearing quiet and loud noises. Now, how would you convert that huge dynamic range to a digital signal? How many bits of of ADC would you need? With an anti-log amp (compressor), you could compress those signals to a linear function (db/V) and store them. Then, on the output, you could expand them (log amp) to get the full dynamic range.

See Compander circuit (chip) = Compressor/Expander
http://www.datasheetcatalog.com/datasheets_pdf/M/C/3/3/MC33110D.shtml

General topic
https://en.wikipedia.org/wiki/Companding
You're right... there must be a good reason for describing sound intensity in dB... which is a logarithmic unit.
 

GopherT

Joined Nov 23, 2012
8,009
Other things like current from a photo-diode per distance from a light source (1 over d-squared) type relationships can be linearized with a log amp.
 

ErnieM

Joined Apr 24, 2011
8,377
One of my first projects was a music synthesizer where I used a voltage to frequency generator driven off a series of equally spaced voltages. By running the equally stepped voltages thru a log converter I got a scale of notes out.

The fact that the log generator was horribly temperature dependent and I eventually abandoned the work is simply besides the point
 

Thread Starter

cmartinez

Joined Jan 17, 2007
8,253
One of my first projects was a music synthesizer where I used a voltage to frequency generator driven off a series of equally spaced voltages. By running the equally stepped voltages thru a log converter I got a scale of notes out.

The fact that the log generator was horribly temperature dependent and I eventually abandoned the work is simply besides the point
Ah! I had not thought about musical notes as related to logarithmic scale before... and I'm also a musician myself!
 

OBW0549

Joined Mar 2, 2015
3,566
Is there any practical use for that characteristic?
Yes.

The PN junction logarithmic V-I relation figures prominently in much, if not all, of analog circuit design. Aside from describing the relationship of forward voltage to current in a diode or in a BJT's base-emitter junction, the logarithmic V-I relationship is used explicitly in analog multipliers/dividers, balanced modulators/demodulators, voltage-controlled variable-gain amplifiers, circuits for generating square/square root and cube/cube root of analog signals, and log/antilog circuits. Many of these uses have embraced digital signal processing over traditional analog processing in recent years, but there are still appropriate applications.

Some reading:

http://www.ti.com/ww/en/bobpease/assets/AN-31.pdf National Semiconductor Application Note AN-31, Op Amp Circuit Collection, pp. 30-32

http://www.ti.com/lit/ds/snls385a/snls385a.pdf National Semiconductor LM194/394 datasheet, Supermatched Pair

http://users.ece.gatech.edu/~lanterma/sdiy/datasheets/transistors/AN-222.pdf National Semiconductor Application Note AN-222, Super Matched Bipolar Transistor Pair Sets New Standards for Drift and Noise, pp. 10-12

and

http://www.thatcorp.com/Design_Notes.shtml THAT Corporation design notes for their line of Blackmer Cell VCVGA ICs.
 

nsaspook

Joined Aug 27, 2009
13,270
Natural processes are seldom linear and a linear view of the universes smacks us in head with every bad weather prediction or electrical blackout from non-linearities in AC current transmission power flow.
 

MrAl

Joined Jun 17, 2014
11,474
Hello there,

I did an article on another site but i'd have to find it now.
It shows how to use and scale the input/output to create a log amplifier.
One note though is that diodes are temperature sensitive, so something has to be done about that too.
 

WBahn

Joined Mar 31, 2012
30,055
Is there any practical use for that characteristic? I mean, are there situations in real life in which it is desirable/useful to have a logarithmic (as opposed to linear) voltage drop?
Maybe metrology or normalization of certain signals?
There are lots of times when the last thing that is useful to you is a linear response.

Linear systems, while very useful and relatively easy to analyze and design, are limited by the fact that they cannot produce new frequencies, which means that you cannot use a linear circuit to do something as "simple" as modulate a voice signal onto an AM carrier or demodulate it at the other end. To do either of these you MUST have a nonlinear element. For simple such circuits, a diode serves this purpose just fine. For more complicated circuits FET transistors are favored because of the are dominated by a square-law characteristic.
 
Top