Comments by Dave in Ohio, and Dr. Mittleman, regarding the "faster controller" on the T-4000

(THIS IS A REPOST FROM THE COMPANION GOOGLE TERAHERTZ GROUP PAGE)
Dave in Ohio, wrote:
API recently announced a faster "controller" capable of producing 1000
sampled terahertz waveforms per second, one order of magnitude faster
than their previous 100 Hz. rate.

Since API did not announce any increase in the rep rate of the laser
(or femtosecond oscillator), we must assume the 10 fold increase in
waveforms comes at the expense of either reduced FFT size and/or a
lower number of averaged samples per pixel. The former would result
in coarser spectral resolution while the latter would decrease the
SNR.

Those trade-offs may not be important at all for some applications and
in such cases, the faster throughput could be a significant
advantage. The company did not mention imaging speed, but a 10 fold
increase would reduce the imaging time of a 12" x 12" area (using
millimeter pixels) from over 15 minutes to about a minute and a
half.

SNR relates directly to the amount of noise in a measured signal or
image. If the source is very close to the object and the object is
not too dense, then a sufficient SNR may be obtained even with a
reduced number of averaged samples per pixel. This would permit the
use of the faster waveform output rate.

The FFT size relates to the number of samples performed over the 320
psec. window. The more samples in the FFT, the finer the frequency
(spectral) resolution available for analysis. Generally, one wants
the resolution as high as possible when doing spectroscopy. For non-
spectroscopy applications, the FFT size would not be particularly
important.

So the new, faster waveform option gives the user much more latitude
in selecting the best TD-terahertz technique based on the needs of
each specific application.
-----
Dr. Daniel Mittleman responded:

I cannot be certain, but I am reasonably sure that the increase in speed is
a hardware improvement, i.e., faster scanning optics. This means that it
would correspond to the second of your options: a lower number of averaged
samples per data point in each measured waveform, and therefore a lower
number of averages per data pixel, if one is talking about imaging. Of
course, the time window may also be different, which would affect the FFT,
but as you point out, it's not obvious that this matters much.

--------------------------------------------------------------------
Dr. Daniel Mittleman Phone: (713) 348-5452
Rice University FAX: (713) 348-5686
ECE Dept., MS-366 E-mail: daniel@rice.edu
6100 Main St. URL: www-ece.rice.edu/~daniel/Mittleman.html
Houston, TX 77005
----------------------------
--~