[time-nuts] Sub Pico Second Phase logger

Bruce Griffiths bruce.griffiths at xtra.co.nz
Mon Dec 15 22:31:27 UTC 2008


Joseph M Gwinn wrote:
> Bruce,
>
>
> time-nuts-bounces at febo.com wrote on 12/15/2008 04:34:34 PM:
>
>   
>> Joseph M Gwinn wrote:
>>     
>>> Bruce,
>>>
>>>
>>> time-nuts-bounces at febo.com wrote on 12/11/2008 08:36:47 PM:
>>>
>>>
>>>       
>>>> Joe
>>>>
>>>> Attached is noise spectrum (1kHz and below) of AP192 with nothing
>>>> connected to inputs.
>>>> Sampling rate 96KSPS.
>>>> Frequency bin equivalent noise bandwidth ~ 3Hz.
>>>> Noise has similar spectrum to flicker noise with a noise corner of
>>>> around 300Hz or so.
>>>>
>>>>         
>>> The noise floor ain't so bad, -130 dB (dBm? dbV?) at maybe 10 Hz, even 
>>>       
> if 
>   
>>> it's 20 dB worse than at 1 KHz.
>>>
>>> Has anyone measured the Allan Deviation?
>>>
>>> Joe
>>>
>>>
>>>
>>>       
>> Joe
>>
>> Noise plot is with respect to full scale (4Vrms).
>>     
>
> OK, call it dbV+12db.
>
>
>   
>> I need to build a noise source to check the absolute level.
>> Will use the amplified Johnson noise of a 150K resistor.
>>
>> By Allan deviation do you mean calculate it from the sequential 96KSPS
>> ADC output samples?
>>     
>
> Yes, although some decimation may be needed to keep compute times under 
> control, at least for the larger values of tau.
>
>  
>   
>> I can do this, but since the dominant noise source is white the Allan
>> deviation will scale with the measurement bandwidth.
>>     
>
> Would modified Allan deviation be better?
>
> I'm more interested in the general shape of the Allan curve than its 
> absolute value, one issue being the effect of thermal variations in your 
> laboratory dungeon.  We had speculated as to the relative size of thermal 
> effects in these sound cards, and this would give us some idea.
>
>
> Joe
>
>   

Joe

Modified ADEV, ADEV etc are possible, although the maximum usable record
length probably depends more on the limits of Plotter and Windows 2K.


I'll look into doing this.
Real time filtering and decimation may be impractical, in the short term
at least, as most signal processing libraries only process 16 bit samples.
Most real time spectrum analysis programs are similarly afflicted in
that they only process 16 bit samples.

Bruce



More information about the time-nuts mailing list