General > General Technical Chat
Anybody seen this Analog x Digital multiplier architecture before?
(1/2) > >>
RoGeorge:
I was replacing today some tinypic broken links with local uploads on various online places, and was amused by my own hubris (did some years ago a circuit, I called it Rogeorge Cell, as in Gilbert Cell, but mine ;D ).

However, I never seen such a multiplier, so I thought I might ask if this idea was published or used already.

It is a digitally controlled potentiometer placed between a direct signal and its inverted version, like in the next figure.  The input and the signal path is fully analogue, while the multiplication is by a digitally set value.  By varying the position of the potentiometer's cursor (the digital set value) a four quadrant multiplication can be made between the analog signal and the digitally set value.

The same idea can be used to build a 4 quadrant multiplier for something other than voltage, if the divider is made out of something different than a digitally controlled potentiometer.  I'm thinking here about an optical signal multiplier, or maybe an RF signal.


The block diagram of the multiplier.  Here it was drawn as a digitally controlled potentiometer placed between the direct signal and the inverted signal, but the +, the - and the ratio adder can be anything else with the same functionallity.



And this is a prototype with the most left breakout board is for the OpAmps (quadruple OpAmp OPA1654), the next is the digitally controlled potentiometer (double potentiometer, 10k linear, I2C controlled, TLP0202), and the last one is an Arduino nano board (Atmega324) to controll the potentiometer by I2C.



https://hackaday.io/project/7542-rogeorge-cell/
Sorry for the shameless plug.   :-[
tom66:
I think the biggest disadvantage that an architecture like this would have is the bandwidth of a digital potentiometer is not fixed - since the output capacitance is usually constant and the resistance of the digipot increases or decreases with the selected code, you'd expect some variation of frequency response vs output code/gain. That said, having the complementary signal might have the effect of cancelling this out - maybe it needs to be modelled? And at low frequencies this effect could well be negligible, it very much depends on what you are trying to do.

Interesting architecture though and I'd share your curiosity as to whether anyone else has done something along these lines.
newbrain:
Good thinking, here are a couple of considerations.

Digipots' innards are not that far away from the R-2R part of a DAC; here, they are exploited to build a sort of multiplying DAC, with an added single ended to differential conversion to attain 4 quadrant operation.

Multiplying DACs (with 4 quadrant capabilities) have been around for some time, see, as examples, this old DS for MAX501 and this AN from TI; also AD has (had?) parts based on the same principle.


rhb:
Here's an MSEE thesis on the subject:

http://essay.utwente.nl/58276/

Have Fun!
Reg
David Hess:
That is a decades old idea used with current steering (bipolar) DACs.  Check out the 1986 Linear and Conversion Applications Handbook from PMI for several examples:

https://www.analog.com/media/en/technical-documentation/application-notes/28080533AN106.pdf

Tektronix used an analog version with two Gilbert cells with cross connected outputs to make wide bandwidth attenuators for an oscilloscope's "variable" function.

Update: Tektronix initially called this configuration a "para-phase amplifier".
Navigation
Message Index
Next page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod