Re: Electronoics of FC ???

From: Howard Shapiro (hms@shapirolab.com)
Date: Wed Jul 05 2000 - 19:48:11 EST


Alexander Shvalov asks-


>  Could someone help with literature on FC electronics
>  (block - scheme or/and electronic scheme, though,
>  I'm afraid the recent is the subject to copyright) of
>  how signal from Fluorescence is processed.
>
>  How it works? :
>   Pulse board, Pulse processor, Status board,
>   FSC Amplifier, SSC/FL2 .. FL1/FL2 ?
>
>  or send electronic version
>
>  We plan to add signal processing system to
>  fluorescent channel Scanning Flow Cytometer
>  unit, (principally new approach to the light
>  scattering)...
>  and would like to understand basic principles
>  of signal processing from FC fluorescent
>  Photoamplifier data processing.
>
>  Question: what reflects the fluorescence amplitude?
>  Integral of the signal from Photoamplifier?
>  Amplitude? or some classic combination of
>  those parameters with width of the signal.

I have tried to provide good qualitative descriptions of signal processing
in flow cytometers in the various editions of "Practical Flow Cytometry",
in a more recent paper (Shapiro HM, Perlmutter NG, Stein PG: A flow
cytometer designed for fluorescence calibration.  Cytometry.  1998;
33:280-287), and in the new version of the Flow Cytometry volumes in
Methods in Cell Biology", now in press from Academic Press.  Most of that
stuff is subject to copyright, but the current 3rd Edition of "Practical
Flow Cytometry" (Wiley-Liss, New York, 1995), which I wish were cheaper,
still won't cost you as much as a photomultiplier tube.

Complete schematics for linear signal processing electronics were contained
in the 2nd Edition of "Practical Flow Cytometry"; Wiley allows me to
reproduce and distribute these since they decided to omit construction
details from the 3rd and subsequent editions and add more on
applications.  I will send you a copy if you want it; however, these
schematics date back to the mid-1980's, and you would almost certainly want
to do things in a different fashion now.

A 1980's vintage flow cytometer used the following components:

Detectors - photodiode for forward scatter (PMT if higher gain needed);
PMT's for orthogonal scatter and fluorescence.

Preamplifiers - current-to-voltage conversion using an op-amp, with
baseline restoration provided through a feedback circuit integrating the
DC-coupled signal, with the peaks clipped off by diodes, and subtracting
this from the input signal.  Additional stages of voltage gain may be
included, and are required when diode detectors are used; when PMT's are
used, it is advisable to control gain using PMT voltage adjustments,
because this introduces less noise than voltage amplification.

"Front End" Electronics - incorporate a comparator to sense when a cell is
present, i.e., when a trigger signal level rises above an adjustable
threshold, and timing electronics to deal with coincidences and generate
control signals for peak detectors, integrators, and/or pulse width
measurement circuits, and signals to initiate analog-to-digital conversion
of the held peak, integral, and width values.

Peak detectors - these hybrid analog circuits are well known if not always
well behaved; peak signals are stored in capacitors, and a reset pulse
closes a switch which grounds the capacitor.

Integrators - these can be switched-capacitor integrators, or, in some
cases, peak detectors, the input signals to which are passed through a
low-pass filter, such that the peak height of the filtered signal becomes
proportional to the integral of the raw input signal.

Pulse width measurement circuits - basically integrators with a linear ramp
input.

To deal with large dynamic range signals, logarithmic amplifiers were
interspersed between the preamplifier and the peak detector or
integrator.  Fluorescence overlap compensation was done by placing the
circuits for linear combination of signals between the preamplifier output
and the input to the logarithmic amplifier.

Most commercial flow cytometers still work this way.  However, by the
mid-1990's, high-resolution (16 or more bits) analog-to-digital converters
had become available, allowing logarithmic amplifiers and compensation
circuits to be eliminated; compensation and log conversion could then be
accomplished by digital computation.  This was done in the Beckman Coulter
EPICS XL and in instruments built in my lab and others'.  This approach,
while improving the precision and accuracy of compensation and log
transformation, still requires the use of hybrid analog peak detectors,
integrators, etc., and there may be problems with dynamic range because
these circuits do not perform well at very low input signal levels.

Most recently, there has been a trend toward digital processing of the
signal pulses, using 14-bit converters with digitization rates of 1.25
megahertz and higher.  The Luminex 100 flow cytometer, designed for
multiplexed biochemical assays on dyed beads, incorporates digital pulse
processing, and Becton-Dickinson showed prototype digital pulse processing
electronics for the FACSVantage sorter at the recent ISAC meeting in
Montpellier.  Schematics of these systems are proprietary.  However, they
basically include a simple preamplifier and a fast A-D converter for each
channel (the same old diodes and PMT's are used as detectors), and one or
more DSP chips to deal with the signals.  Threshold determination,
triggering and gating, pulse height and width determination and
integration, compensation, and log transformation are all done by the DSP
processor(s), which can also do baseline restoration and even compensate
for laser noise by sampling the laser output. When multiple DSP's are used,
one is set up to supervise and distil data from the others. The DSP data is
transferred to the PC (or Mac) used for data analysis through a fast link
such as Ethernet, USB, or IEEE 1394. If you are going to develop a new
system, this is the way to do it.  Unfortunately, there is very little - or
at least I've found very little - in the DSP literature about pulse
processing; it all seems to be oriented to periodic signals and fast
Fourier transforms.  You have to figure it out for yourself (or find a DSP
expert and get him or her to help you figure it out).

-Howard



This archive was generated by hypermail 2b29 : Sat Mar 10 2001 - 19:31:25 EST