• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 79 subscribers
  • Views 679 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

ADC10 overclocking

Ilmars
Guru 46710 points

Hi!

I am curious - how ADC10 behave when overclocked? Let's say from 6.3MHz (max specified ADC clock) to 12MHz or even 16MHz. From what I know, some (competitor's) uC's with SAR-based ADC allow to decrease resolution & increase sampling rate by overclocking. Does msp430 behave same way or results will be useless? TI datasheets specify 6.3MHz max ADC freq with note "For specified performance of ADC linearity" which is kinda interesting message. Any realworld experience or other kind of useful knowledge?

  • 0
    Guru 35605 points

    ILMARS JARGANS said:
    how ADC10 behave when overclocked?

    How does anything behave when operated outside its specified limits: The behaviour is undefined.

    You might get lucky and find that you just happen to get apparently-useful results - but you have no guarantee whatsoever that those results will be reliable or repeatable.

    And the fact that anyone else may have got lucky and found that they just happen to get apparently-useful results is  no guarantee whatsoever that anyone else will have the same luck.

     

  • 0 in reply to Andy Neil
    Guru 46710 points

    Andy Neil said:

    How does anything behave when operated outside its specified limits: The behaviour is undefined.

    Exactly. Undefined = not specified. Undefined does not universally mean useless. "Other" SAR ADC's at higher freq's predictably lose resolution, still giving useful readings. - I will not name it (other ADC) yet, but perhaps you already know which one I mean. I want to know how TI's ADC10 behave.

    In our world where everything can be measured there's no such thing as luck. Current information we have tells just one thing: TI specified maximum frequency at which ADC10 conform to specified 10-bit resolution and linearity. At higher frequencies most probably ADC10 meets other resolution/linearity specs which I am looking for. Why TI did not specify - maybe because they have other more important things to do, not because it's not possible to get meaningful and repeatable results with overclocked ADC10.

  • 0 in reply to Ilmars
    Guru 227245 points

    Ilmars said:
    Undefined does not universally mean useless.

    Indeed. It means 'we don't know or don't make any statement about its behaviour'. It also means that even if you believe you know what happens, this might only apply to your individual device and the next one you try behaves differently. Sure, you may uy a large lot and try all and determine the maximum variation. However, it doesn't mean that the next lot will behave the same. Within specification, TI guarantees that future device will meet these specs.

    But to answer your question: The ADC10/12 use charge distribution method. A smart quick and cheap method of ADC conversion. However, charge distribution is subject to low-pass effects caused by capacitance and resistance of the inner circuitry. Beyond a certain point, the outcome can be rather random. If a distributed charge has not enough tiem to flow, then anything can happen. Same happens for underclocking. Due to leakage currents, underclocking may change the charges also so much that the distribution algorithm fails and the outcome is rahte rrandom.
    For devices working with a comparison loop for successive approximation using a D/A and a comparator, over/underclocking will cause the sampling charge to leak away or thesettling time of the DAC not being met before next comparison, leading to an increasing loss of precision. On a charge distribution device, I expect a slight loss of precision suddenly turning into bit flips. But then, I don't see a reason for overclocking the ADC. The limiting factor is the input R/C combination which will require a certain sampling time anyway. For a good precision, this is usually way more than the conversion time.

  • 0 in reply to Jens-Michael Gross
    Guru 46710 points

    Jens-Michael Gross said:

    But to answer your question: The ADC10/12 use charge distribution method. 

    That's it! This is all I wanted to know. So answer is: ADC overclocking most probably is not useable. Thanx.

  • 0 in reply to Ilmars
    Guru 227245 points

    Ilmars said:
    So answer is: ADC overclocking most probably is not useable.

    I jsu tdid some google search and found the following:

    Mixed-Signal Systems: Analog Input and Output chapter 9.3 and especially 9.3.2.

**Attention** This is a public forum