cancel
Showing results for 
Search instead for 
Did you mean: 

ADC: precise measurement on small packages (no reference voltage inputs)?

e-design
Associate II
Posted on September 04, 2008 at 07:57

ADC: precise measurement on small packages (no reference voltage inputs)?

5 REPLIES 5
e-design
Associate II
Posted on May 17, 2011 at 12:43

Hello,

in a new project I want to use a STM32F103T8U6 for precise measurement (voltage, current & temperature) and power regulation (DPWM).

For the measurement I want to use the 12-bit ADC of the STM32.

Unfortunately there are no input pins for the reference voltage when using small packages (less than 100 pins).

Has anyone tried to use a precise LDO for VDDA at nearly the same voltage level as VDD?

Or do you have any other suggestions how to make precise measurement with the 12-bit ADC (less than 0.5 % initial accuracy and low temperature coefficient / ageing).

Regards,

Florian

andreas2
Associate II
Posted on May 17, 2011 at 12:43

If you can spare an ADC input you could always convert the output of an external voltage reference and use it for correcting the other samples.

e-design
Associate II
Posted on May 17, 2011 at 12:43

Hi Andreas,

yes I can spend an ADC-input for the reference measurement.

But I think it is very important to take the reference measurement nearly at the same time as the regular measurements.

What do you think is the best way to achieve this?

- Dual-Mode: reference measurement with one ADC, regular measurement with the other ADC (taking the measurement exactly at the same time)

- Scan-Mode: reference measurement and regular measurement consecutively with the same ADC (taking the measurements with a short time skew between them)

When using the second approach (Scan-Mode) it seems to me very important that I have a very clean power supply, so there is no deviation between the two scans.

(Still wondering why STM did not connect the reference voltage to some pins on the smaller devices...)

Regards, Florian

lanchon
Associate II
Posted on May 17, 2011 at 12:43

it shouldn't be that critical.

> When using the second approach (Scan-Mode) it seems to me very important that I have a very clean power supply

Vref should be fairly stable during conversion or a successive approximation ADC won't perform well. this means that if you simultaneously measure the signal and the ext ref on two ADCs while Vref is changing, both measurements will be inaccurate.

on the other hand if Vref is fairly stable, you won't loose much if you serialize the measurements. my recommendation is to avoid complexity. if you have software devel time to spare, it might pay you much more to invest it in other strategies (like pausing the processor during measurements) than to parallelize.

design a good supply with low impedance decoupling at the sensitive frequency range. for that range use low ESR caps and don't relay on parallel caps more than a decade apart in capacitance.

e-design
Associate II
Posted on May 17, 2011 at 12:43

Hi launchon,

you are right with your VREF considering.

Unfortunately on smaller devices there is no separate VREF input (VREF is derived from VDD).

If VDD is a little bit dirty you get bad results from the ADC.

I'll do some filterung for the VDDA input (series resistor, ferrit bead and capacitors) to get it stable.

And I'll generate the VDD supply by a LDO (low dropout regulator).

Regards, Florian