Bandgap ADC

[Enthalpy]

Hello everybody!

Analog-to-Digital Converters (ADC) need an accurate voltage reference to convert an input voltage into a dimensionless number. The bandgap reference is common
Wikipedia
It uses the V_be of two bipolar transistors used as diodes and passing different current densities. V_be decreases with the temperature but the difference of both V_be increases, and at least in simple models, a weighed sum of about 1.24V for silicon is independent of the temperature and relates with the bandgap. Excellent special chips do that.

I propose here to let software make the weighed sum and more corrections for further improvements. I didn't check if this is already done.

• The same ADC converts the voltage to be measured and the V_be. Software makes the comparisons and corrections.
• My circuit uses a single transistor at different currents successively. No more mismatch between to transistors.
• Accurate current ratios are still necessary. They can result from a number of identical resistors, measured and chosen if discrete, adjusted if integrated. Easier than matched transistors.
• Software can, better than a continous-time circuit, identify and correct error sources, especially the series resistance and the ideality factor, by using more than two currents.
• Several transistors in series can exploit the ADC range better. For instance the MAT-04 contains 4 and thrives to resemble ideal models.
• The ADC still needs an auxiliary voltage reference whose accuracy isn't critical but should be quiet enough.
• The V_be should be measured often enough so software compensates the noise and drift of the auxiliary reference and removes most noise from the computed reference.
• This circuit adds the ADC's quantization noise several times, multiplied by 5 to 10 for the V_be difference. Process many conversions, and use a high resolution converter, or improve it for V_be with dither noise added there.

Marc Schaefer, aka Enthalpy

[John Cuthber]

1 hour ago, Enthalpy said:

I didn't check if this is already done.

This is close

http://www.ti.com/lit/ug/tiduaf0/tiduaf0.pdf

You need a counter and  comparator to "close the loop" to make  it ADC

[Enthalpy]

Thanks JC!

There are some common elements, but not the computation of a bandgap voltage by software.

I'd not be surprised if someone had already done exactly what I propose. The idea is immediate enough.