[Halld-tracking-hw] ADC125 as a TDC...

[Gerard Visser]

Hi Beni,
	That looks good, glad to see the interpolation still makes sense on a real signal.
	Your question is certainly an important, valid point. But, "in the old days" or 
even in FDC wires today, you just use a LE discriminator and push the threshold 
down pretty low. There is certainly an error from time walk there. But it works. 
So I don't think that we can rule out accpting this error in the digital method, 
we have to try it and see. On the other hand, one doesn't usually use a 
high-order shaper in the timing path in an analog readout, so the errors due to 
time walk here are going to be more significant probably than in a traditional 
analog readout. I could imagine using some information from the complete leading 
edge of the pulse, e.g. find the slope somehow in terms of the next pulses over 
threshold, to make a time walk correction. For this particular event it looks 
like that would work well.
	The threshold to be used needs to be optimized too. I am sure that the time 
walk error becomes larger at some point, as the threshold is reduced. (Because 
the pulse starts out at shallow slope because of the high order shaping.) On the 
other hand, at larger thresholds (or in the DCOG algorithm, I believe) one 
starts to see not the first electrons, for short drift time signals. This is why 
I am pushing for some kind of leading edge measurement, it is the only way to 
grab those first electrons and see their timing.
	If you don't mind, could we just try this without time walk correction and see 
what we get? It looks like you should do some common-mode correction first 
though. [Ok so we think that will be hard to do online in the ADC125 but clearly 
it will affect the result significantly for this signal. Maybe it has to be 
done, in which case we better think about how to do it and if any changes in the 
ADC125 are needed to do it.]
	I started this thread of emails off list (sorry) but now we are moved to topics 
that would be better to have a record of the discussion... So I change the "To:" 
above.

	- Gerard
	

On 5/13/2011 11:26 AM, Beni Zihlmann wrote:
> Hi Gerard,
> I was finally able to implement your code into my hack to analyze some real
> FDC wire data. Attached you see an event with two wire signals including
> your interpolation.
> The red curves with solid dots are the FADC data and the green curves with crosses
> are the interpolation result using your code. I still have some issues I do
> not completely understand so I will come back to you at some point.
>
> One question I have though is the following. The code segment at the
> end of the function le_time() calculates the threshold crossing time. This
> is however not the time we really need. What we need is the start time
> of the signal. Otherwise we have a huge effect from time walk.
> Do you have a proposal for calculating this value?
>
> thanks,
> Beni