diode for log gain circuit

[Scott Kelley]

Knowing the voltage across a diode, I need to determine the current through
it with a fair degree of accuracy over a range of about 1000x. Is any one
type of diode better than another in terms of the accuracy of its
logarithmic voltage gain with increasing forward current? This will be a
production circuit, so consistancy between devices would also be good to
avoid the need for individual calibration.

Thanks,
Scott Kelley

 

[PeteS]

Knowing the voltage across a diode, I need to determine the current through
it with a fair degree of accuracy over a range of about 1000x. Is any one
type of diode better than another in terms of the accuracy of its
logarithmic voltage gain with increasing forward current? This will be a
production circuit, so consistancy between devices would also be good to
avoid the need for individual calibration.

Thanks,
Scott Kelley

Production spread on most diodes will give you a lot of grey hairs on
this quite quickly. A solution might be to use a current mirror (with
the diode being the 'programming side' as AoE so eloquently puts it)
and using the other side as the input to a log amp (as an example).

Much depends on the current you have (what range) - a Wilson mirror
gets around a couple of issues, and used to be commercially available
(not cheap though - as mirrors must have guaranteed Vbe match, they
tend to be expensive. See the MAT-xx series from Analog devices for an
example of tightly matched transistors).

As I said, a lot will depend on the maximum current you are using.

Cheers

PeteS

 

[PeteS]

Production spread on most diodes will give you a lot of grey hairs on
this quite quickly. A solution might be to use a current mirror (with
the diode being the 'programming side' as AoE so eloquently puts it)
and using the other side as the input to a log amp (as an example).

Much depends on the current you have (what range) - a Wilson mirror
gets around a couple of issues, and used to be commercially available
(not cheap though - as mirrors must have guaranteed Vbe match, they
tend to be expensive. See the MAT-xx series from Analog devices for an
example of tightly matched transistors).

As I said, a lot will depend on the maximum current you are using.

Cheers

PeteS

Something I didn't mention - this will double the current (there will
be equal current in the two transistors unless you can get or design
matched ratio pairs where you can get a ratio of currents).

Cheers

PeteS

 

[CWatters]

Knowing the voltage across a diode, I need to determine the current through
it with a fair degree of accuracy over a range of about 1000x. Is any one
type of diode better than another in terms of the accuracy of its
logarithmic voltage gain with increasing forward current? This will be a
production circuit, so consistancy between devices would also be good to
avoid the need for individual calibration.

Are you going to have temperature compensation?