Anti log amplifier help?

I would like to see if it is possible to design an anti-log (exponential) HF preamplifier.
One idea might be a conventional amp with a log amp in the feedback loop.
Can anyone with simulation skills or good ideas help me in that direction?
I love discrete transistors circuits, if this can be done that way.

Regards, Dana.

danadakk said:

https://www.analog.com/media/en/training-seminars/tutorials/MT-077.pdf

Antilogarithmic Amplifier | Derivation - Engineering Projects

Antilogarithmic amplifier circuit description and derivation for voltage output. In this article we had also describe and anti logarithmic

bestengineeringprojects.com

https://www.analog.com/en/education/education-library/nonlinear-circuits-handbook.html

Regards, Dana.

Click to expand...

Excellent book the last one!
Actually this "Dead Zone" in part1 Basic operations pp25 of the book, might be useful, as it has thresholds that can be set. It is not an exponential amplifier, but it will do in my case, as it does not amplify signals under a certain signal level, if I am getting this right.
A circuit is given with opamps but I have no idea of the starting values for HF (2-30MHz) or what opamps to use for these frequencies.

neazoi said:

Excellent book the last one!
Actually this "Dead Zone" in part1 Basic operations pp25 of the book, might be useful, as it has thresholds that can be set. It is not an exponential amplifier, but it will do in my case, as it does not amplify signals under a certain signal level, if I am getting this right.
A circuit is given with opamps but I have no idea of the starting values for HF (2-30MHz) or what opamps to use for these frequencies.

Click to expand...

UPDATE:
But how the dead zone circuit will behave under the presence of a strong amplitude sine in one frequency and a very low amplitude signal on another frequency? At the time that the amplifier conducts, it would amplify both signals, isn't that true?

On the other hand, an exponential amplifier will amplify both signals too, but at all times. It just amplifies the larger signals more than the lower.

Sounds like you didn't actually consider your requirements.
1. anti-log is has an unipolar characteristic with output >0, not suited for AC signals, e.g. RF
2. it has a strongly nonlinear characteristic. If you feed two signals to it, it acts as a mixer and doesn't maintain the original input spectrum

Can you describe a bit more what your intended circuit is for ?
Eg. what is input signal, its characteristics, and what processing
do you want done on that, eg. goal of output.

The circuit you are looking at is a summation of two signals
processed linear over region outside dead zone, followed by
simple sum of them. So not antilog....?



What is the GBW you need for the signal path ? That will determine OpAmp
choice. Given you will be working in RF range you will need very careful layouts'
and power distribution. Datasheets for these types of OpAmps have strong layout
recommendations for the respective OpAmp that one should follow. Will you be
working in a 50 ohm environment ?

Attached additional reference that you might find useful.

Lastly here is a table of wideband OpAmps, take a look at a couple of datasheets
to get an idea of layout considerations. if, for example, you want a G of ten in the
signal path, and you want 30 Mhz 3 db kinds of response, you need to look at parts
whose GBW > 300 Mhz. I have worked with OpAmps in this area (hybrids) eons ago
and they were a beast to stabilize. Even capacitors were a problem, careful selection/
evaluation needed for them as well.



Regards, Dana.

FvM said:

Sounds like you didn't actually consider your requirements.
1. anti-log is has an unipolar characteristic with output >0, not suited for AC signals, e.g. RF
2. it has a strongly nonlinear characteristic. If you feed two signals to it, it acts as a mixer and doesn't maintain the original input spectrum

Click to expand...

Both statements are very useful to know, thanks. I suspected about the second. See this image, I have taken it out of a Tektronix 491. This represents the square law position where the differences between smaller and larger signals are shown larger. This does work on RF, but I do not remember where the filter is, before or after it, I have to look at it.
Perhaps there is a filter before, just to minimize the 2nd case you refer to.

danadakk said:

Can you describe a bit more what your intended circuit is for ?
Eg. what is input signal, its characteristics, and what processing
do you want done on that, eg. goal of output.

The circuit you are looking at is a summation of two signals
processed linear over region outside dead zone, followed by
simple sum of them. So not antilog....?

View attachment 181579

What is the GBW you need for the signal path ? That will determine OpAmp
choice. Given you will be working in RF range you will need very careful layouts'
and power distribution. Datasheets for these types of OpAmps have strong layout
recommendations for the respective OpAmp that one should follow. Will you be
working in a 50 ohm environment ?

Attached additional reference that you might find useful.

Lastly here is a table of wideband OpAmps, take a look at a couple of datasheets
to get an idea of layout considerations. if, for example, you want a G of ten in the
signal path, and you want 30 Mhz 3 db kinds of response, you need to look at parts
whose GBW > 300 Mhz. I have worked with OpAmps in this area (hybrids) eons ago
and they were a beast to stabilize.

https://www.analog.com/en/parametricsearch/11087#/

Regards, Dana.

Click to expand...

Yes, this is the circuit I was looking from that book. Input level can be as low as a few uV (-100dbm or so) and as high as -10dbm, even more. Frequency is 2-30MHz. I do not need much gain, 3-9db at max I think should be more than enough. I bet the same thing can be done with simple BJTs as amplifying elements instead of the opamps (I may be wrong though).

Then simple CE designs would suffice, maybe a CB front end for Z translation if
you need that. Followed by a summer. But the dead zone behavior not sure
how to tackle that with simple discrete approach.

Here is a classic Motorola ap note - https://cdn.macom.com/applicationnotes/AN215a.pdf

Since you want 1 uV on low end makes one wonder if a dual gate MOSFET might
give you some flexibility in design/dead band......


Regards, Dana.

danadakk said:

Then simple CE designs would suffice, maybe a CB front end for Z translation if
you need that. Followed by a summer. But the dead zone behavior not sure
how to tackle that with simple discrete approach.

Here is a classic Motorola ap note - https://cdn.macom.com/applicationnotes/AN215a.pdf

Since you want 1 uV on low end makes one wonder if a dual gate MOSFET might
give you some flexibility in design/dead band......


Regards, Dana.

Click to expand...

Thanks a lot.
What puzzles me is how will such a dead-zone circuit will behave under the presence of a strong amplitude sine in one frequency and a very low amplitude signal on another frequency? At the time that the amplifier conducts, it would amplify both signals, the high one but also the low one, isn't that true?
The amplifier sees input signal levels. Once it is triggered by the presence of the high level signal, it will also amplify the low level one.
In contrast, an exponential amplifier just amplifies differently, low and high level signals.
I hope I understand the above things correctly?

In the circuit you are interested in each signal processed with an amp with dead band
before summation. So the low level signal, if its amplitude in the dead band, does not
get passed. 1 uV I would posit is right at noise levels in circuit, so not sure what region
of it you want "dead banded".

Basically you have two linear amps, with the special property of dead band region, being
summed. And the low level signal close to or in noise region when it is out of dead band
region.

I think you need to do some simulations to get a handle on this, at the least a starting
point.


Regards, Dana.

neazoi said:

I love discrete transistors circuits, if this can be done that way.

Click to expand...

This biasing arrangement of 2 mosfets produces curves that appear exponential.

Diode voltage-versus-Amperes curve. Exponential.



I once tested real diodes, reading volts and Amps. I graphed the data then finagled with equations in a search for a formula I could use in my homebrew simulator. (I steered away from the Shockley formula.) This reasonable match emerged:

A=(V * 1.25)^20

BradtheRad said:

Diode voltage-versus-Amperes curve. Exponential.

View attachment 181586

I once tested real diodes, reading volts and Amps. I graphed the data then finagled with equations in a search for a formula I could use in my homebrew simulator. (I steered away from the Shockley formula.) This reasonable match emerged:

A=(V * 1.25)^20

Click to expand...

Or a set of anti parallel ones, as used in the Tek 491. The problem with such approaches is the high signal level used. Anyway, here is the anti parallel diodes simulation.