Confocal Microscopy List

Re: averaging vs. accumulation for noise reduction - is there a difference?

Classic

List

Threaded

10 messages Options

Stanislav Vitha

Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hallo,
thanks everybody for their input regarding this topic.
I have some results from a simple test.

Long dwell time (200 us) seem to produce essentially the same signal to
background ratio as accumulation of 100 images with 2us dwell times:
signal:

1 x 200us dwell time:
area (pixels) mean STDEV signal-to-background
specimen 20812 181.76 34.44 14.63
background 20812 12.42 8.99

100 x 2 us dwell time, Sum of frames
area (pixels) mean STDEV signal-to-background
specimen 20812 185.28 33.97 14.73
background 20812 12.58 8.97

The specimen is a Chroma fluorescent slide occupying half of the field of view,
the other half does not have any sample (= background).
The Sum was calculated in ImageJ (Image-Stack-Z Project-Sum of Slices).

FYI,
I also tested for the linearity of the signal versus laser power in the photon
counring mode to see when the pile-up error would become noticeable.
Using 10 us dwell time, spectral detector was set to narrow width (2 nm band
pass)to restrict the photon flux. The signal seems to scale well with intensity,
with just a slight drop off at the highest laser powers (but some of that drop
off can be due to photobleahching; after acquisition I zoomed out and the
scanned area showed as a darker rectangle):
(results from an area of 16756 pixels)
Specimen background
laser power (%) mean STDEV mean STDEV
0.1 9.36 7.62 0.67 2.13
0.2 11.3 8.42 0.65 2
0.3 13.26 8.89 0.65 2
0.4 15.21 9.69 0.67 2.1
0.5 16.85 10.22 0.68 2.09
1 45.54 16.63 0.68 2.13
2 82.51 22.64 0.67 2.15
4 155.69 31.58 0.66 2.09
8 297.19 43.26 0.73 2.19
16 567.71 60.79 0.79 2.32
32 1049.43 82.74 0.86 2.36

So it seems that I can count about 100 photons per microsecond without
having to worry too much about the pile-up. Does anybody know how how the
Olympus's "Hybrid photon counting" works?

Spreadsheed file with more details, graphs are here:
http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-tests

This folder also contains some of the image files.

Sincerely,

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

Boswell, Carl A - (cboswell)

Re: averaging vs. accumulation for noise reduction - is there a difference?

Thanks for the data Stan. Nice to see someone going to the trouble of getting numbers to add to the discussion. However the subject line of the thread doesn't seem to be addressed here.

I've had this problem understanding "accumulation" vs. average for years. I've heard the argument that accumulation, or adding multiple images amounts to the same thing as averaging (accumulate multiple images, then DIVIDE the resulting pixel intensities by the number of images collected), as the noise does not accumulate to the same extent as the signal. However the two seem mathematically different, should produce different results and the terms should not be used interchangeably. What am I missing here?

Thanks,
C

Carl A. Boswell, Ph.D.
Molecular and Cellular Biology
Univ. of Arizona
520-954-7053
FAX 520-621-3709

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Stanislav Vitha
Sent: Monday, June 27, 2011 11:49 AM
To: [hidden email]
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hallo,
thanks everybody for their input regarding this topic.
I have some results from a simple test.

Long dwell time (200 us) seem to produce essentially the same signal to background ratio as accumulation of 100 images with 2us dwell times:
signal:

1 x 200us dwell time:
area (pixels) mean STDEV signal-to-background
specimen 20812 181.76 34.44 14.63
background 20812 12.42 8.99

100 x 2 us dwell time, Sum of frames
area (pixels) mean STDEV signal-to-background
specimen 20812 185.28 33.97 14.73
background 20812 12.58 8.97

The specimen is a Chroma fluorescent slide occupying half of the field of view, the other half does not have any sample (= background).
The Sum was calculated in ImageJ (Image-Stack-Z Project-Sum of Slices).

FYI,
I also tested for the linearity of the signal versus laser power in the photon counring mode to see when the pile-up error would become noticeable.
Using 10 us dwell time, spectral detector was set to narrow width (2 nm band pass)to restrict the photon flux. The signal seems to scale well with intensity, with just a slight drop off at the highest laser powers (but some of that drop off can be due to photobleahching; after acquisition I zoomed out and the scanned area showed as a darker rectangle):
(results from an area of 16756 pixels)
Specimen background
laser power (%) mean STDEV mean STDEV
0.1 9.36 7.62 0.67 2.13
0.2 11.3 8.42 0.65 2
0.3 13.26 8.89 0.65 2
0.4 15.21 9.69 0.67 2.1
0.5 16.85 10.22 0.68 2.09
1 45.54 16.63 0.68 2.13
2 82.51 22.64 0.67 2.15
4 155.69 31.58 0.66 2.09
8 297.19 43.26 0.73 2.19
16 567.71 60.79 0.79 2.32
32 1049.43 82.74 0.86 2.36

So it seems that I can count about 100 photons per microsecond without having to worry too much about the pile-up. Does anybody know how how the Olympus's "Hybrid photon counting" works?

Spreadsheed file with more details, graphs are here:
http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-tests

This folder also contains some of the image files.

Sincerely,

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

Armstrong, Brian

Re: averaging vs. accumulation for noise reduction - is there a difference?

In reply to this post by Stanislav Vitha

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Stan, thanks for the data. I think this may rule out some of the issues discussed previously in the thread.

However, if I remember correctly, in a post from McNamara he provided references for papers addressing the nature of the fluorophores in averaging vs scan time. So although this may be true for a Chroma slide it tells us little if anything about the outcome from a test using a sample utilizing fluorophores.

Cheers,

Brian D Armstrong PhD
Assistant Research Professor
Light Microscopy Core
Beckman Research Institute
City of Hope
Dept of Neuroscience
1450 E Duarte Rd
Duarte, CA 91010
626-256-4673 x62872

http://www.cityofhope.org/research/support/Light-Microscopy-Digital-Imaging/Pages/default.aspx

-----Original Message-----
From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Stanislav Vitha
Sent: Monday, June 27, 2011 11:49 AM
To: [hidden email]
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hallo,
thanks everybody for their input regarding this topic.
I have some results from a simple test.

Long dwell time (200 us) seem to produce essentially the same signal to
background ratio as accumulation of 100 images with 2us dwell times:
signal:

1 x 200us dwell time:
area (pixels) mean STDEV signal-to-background
specimen 20812 181.76 34.44 14.63
background 20812 12.42 8.99

100 x 2 us dwell time, Sum of frames
area (pixels) mean STDEV signal-to-background
specimen 20812 185.28 33.97 14.73
background 20812 12.58 8.97

The specimen is a Chroma fluorescent slide occupying half of the field of view,
the other half does not have any sample (= background).
The Sum was calculated in ImageJ (Image-Stack-Z Project-Sum of Slices).

FYI,
I also tested for the linearity of the signal versus laser power in the photon
counring mode to see when the pile-up error would become noticeable.
Using 10 us dwell time, spectral detector was set to narrow width (2 nm band
pass)to restrict the photon flux. The signal seems to scale well with intensity,
with just a slight drop off at the highest laser powers (but some of that drop
off can be due to photobleahching; after acquisition I zoomed out and the
scanned area showed as a darker rectangle):
(results from an area of 16756 pixels)
Specimen background
laser power (%) mean STDEV mean STDEV
0.1 9.36 7.62 0.67 2.13
0.2 11.3 8.42 0.65 2
0.3 13.26 8.89 0.65 2
0.4 15.21 9.69 0.67 2.1
0.5 16.85 10.22 0.68 2.09
1 45.54 16.63 0.68 2.13
2 82.51 22.64 0.67 2.15
4 155.69 31.58 0.66 2.09
8 297.19 43.26 0.73 2.19
16 567.71 60.79 0.79 2.32
32 1049.43 82.74 0.86 2.36

So it seems that I can count about 100 photons per microsecond without
having to worry too much about the pile-up. Does anybody know how how the
Olympus's "Hybrid photon counting" works?

Spreadsheed file with more details, graphs are here:
http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-tests

This folder also contains some of the image files.

Sincerely,

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

---------------------------------------------------------------------
*SECURITY/CONFIDENTIALITY WARNING:
This message and any attachments are intended solely for the individual or entity to which they are addressed. This communication may contain information that is privileged, confidential, or exempt from disclosure under applicable law (e.g., personal health information, research data, financial information). Because this e-mail has been sent without encryption, individuals other than the intended recipient may be able to view the information, forward it to others or tamper with the information without the knowledge or consent of the sender. If you are not the intended recipient, or the employee or person responsible for delivering the message to the intended recipient, any dissemination, distribution or copying of the communication is strictly prohibited. If you received the communication in error, please notify the sender immediately by replying to this message and deleting the message and any accompanying files from your system. If, due to the security risks, you do not wish to receive further communications via e-mail, please reply to this message and inform the sender that you do not wish to receive further e-mail from the sender.

---------------------------------------------------------------------

James Pawley

Re: averaging vs. accumulation for noise reduction - is there a difference?

In reply to this post by Stanislav Vitha

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>Hallo,
>thanks everybody for their input regarding this topic.
>I have some results from a simple test.
>
>Long dwell time (200 us) seem to produce essentially the same signal to
>background ratio as accumulation of 100 images with 2us dwell times:
>signal:
>
>1 x 200us dwell time:
> area (pixels) mean STDEV signal-to-background
>specimen 20812 181.76 34.44 14.63
>background 20812 12.42 8.99
>
>100 x 2 us dwell time, Sum of frames
> area (pixels) mean STDEV
>signal-to-background
>specimen 20812 185.28 33.97 14.73
>background 20812 12.58 8.97
>
>
>The specimen is a Chroma fluorescent slide occupying half of the
>field of view,
>the other half does not have any sample (= background).
>The Sum was calculated in ImageJ (Image-Stack-Z Project-Sum of Slices).
>
>FYI,
>I also tested for the linearity of the signal versus laser power in the photon
>counring mode to see when the pile-up error would become noticeable.
>Using 10 us dwell time, spectral detector was set to narrow width (2 nm band
>pass)to restrict the photon flux. The signal seems to scale well
>with intensity,
>with just a slight drop off at the highest laser powers (but some of that drop
>off can be due to photobleahching; after acquisition I zoomed out and the
>scanned area showed as a darker rectangle):
>(results from an area of 16756 pixels)
> Specimen background
>laser power (%) mean STDEV mean STDEV
>0.1 9.36 7.62 0.67 2.13
>0.2 11.3 8.42 0.65 2
>0.3 13.26 8.89 0.65 2
>0.4 15.21 9.69 0.67 2.1
>0.5 16.85 10.22 0.68 2.09
>1 45.54 16.63 0.68 2.13
>2 82.51 22.64 0.67 2.15
>4 155.69 31.58 0.66 2.09
>8 297.19 43.26 0.73 2.19
>16 567.71 60.79 0.79 2.32
>32 1049.43 82.74 0.86 2.36
>
>
>So it seems that I can count about 100 photons per microsecond without
>having to worry too much about the pile-up.

Then magic is involved. Or something other than what is normally
called photon counting. Do try and find out what they do.

Bit it isn't photon counting with a PMT. (Because even fast PMTs make
single PE pulses that are several ns long and so you would be dead
for at least 40% of the time at 100 MHz.

Also, I note that your STDEV is NOT equal to the sqr root of the
mean. I would guess that they have merely calibrated their PMT system
so that the number in the memory is about equal to the number of
single PE pulses that occurred in the pixel time.

JP

>Does anybody know how how the
>Olympus's "Hybrid photon counting" works?
>
>Spreadsheed file with more details, graphs are here:
>http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-tests
>
>This folder also contains some of the image files.
>
>
>Sincerely,
>
>Stan Vitha
>Microscopy and Imaging Center
>Texas A&M University

--
****************************************
Prof. James B. Pawley, Phone: 604-822-7801
3D Microscopy of Living Cells: Summer Course CELL: 778-919-3176
Info at: http://www.3dcourse.ubc.ca
"If it isn't diffraction, it is statistics":Microscopist's complaint, Anon.

Chen, De (NIH/NCI) [C]

Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

It seems, in this experiment, the count rate is way too high to see the shot noise effect. The Poissonian noise with uncertainty 1/sqrt(N), longer dowelling time will increase S/N. When signal is very weak, longer dowelling time will help.

________________________________________
From: James Pawley [[hidden email]]
Sent: Monday, June 27, 2011 9:10 PM
To: [hidden email]
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Stanislav Vitha

Re: averaging vs. accumulation for noise reduction - is there a difference?

In reply to this post by Stanislav Vitha

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I have a little update.
Some commented that the noise in my photon counting experiment is higher
than expected from Poisson distribution.

I checked the dark counts on the PMTs (kept in the dark overnight, all light
blocked, lasers off during the scan)
512 x 512 pixels scan, 200 us dwell time.

PMT 1 mean = 9.80 Min = 0 Max = 101 STDEV = 7.98
PMT 2 mean = 8.75 Min = 0 Max = 126 STDEV = 8.68
PMT 3 mean = 49.69 Min = 0 Max = 181 STDEV = 18.18

PMT1 was the one used in the previous test.

I expected that the dark count would have Poisson distribution, but again the
STDEV is more than twice the SQRT(mean).

It could be a questtion of the threshold set for photon counting, but I think
only gods on Olympus would know..

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

On Tue, 28 Jun 2011 11:33:13 -0400, Chen, De (NIH/NCI) [C]
<[hidden email]> wrote:

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>It seems, in this experiment, the count rate is way too high to see the shot
noise effect. The Poissonian noise with uncertainty 1/sqrt(N), longer dowelling
time will increase S/N. When signal is very weak, longer dowelling time will
help.
>
>________________________________________
>From: James Pawley [[hidden email]]
>Sent: Monday, June 27, 2011 9:10 PM
>To: [hidden email]
>Subject: Re: averaging vs. accumulation for noise reduction - is there a
difference?
>>Also, I note that your STDEV is NOT equal to the sqr root of the
>mean. I would guess that they have merely calibrated their PMT system
>so that the number in the memory is about equal to the number of
>single PE pulses that occurred in the pixel time.
>
>JP
>

Mario

Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Stan,

I suggest that you adjust voltages and offsets on
your PMTs, test the variations of dark noise
counts with PMT voltage, and make sure that you
are operating on a linear portion of PMT gain
range. The large Max counts you are seeing is a
bit strange. Cosmic rays could generate some
anomalies but given that the STDs are
consistently more than twice the expected values
some systematic problem is more likely. What I
find the most eye catching is your getting ~ 10
dark counts per 200 usec. dwell time. That
implies your PMT(1&2) are generating 50,000
counts per second (-> 250,000 per sec. for PMT
3!). That's likely to be about 60-200x higher
than what one gets with a typical 2 inch PMT, a
discrepancy worthy of examination. If the gain
settings are too high you might end up with a lot
of spurious counts. I would certainly check with
manufacturer to get the specs. of the PMTs.

Another factor that comes to mind is cooling of
your detectors. The noise from PMTs can be helped
a lot by cooling. If PMTs are so equipped, I 'd
check to make sure that the cooling is working.
Going from room temp. down to -20°C can reduce
thermionic emission by greater than a factor of a
hundred.

There was another consideration mentioned in your
posts that you could get about 200 counts per
pixel with a 2 usec dwell in your test specimen.

"100 x 2 us dwell time, Sum of frames
area (pixels) mean STDEV signal-to-background
specimen 20812 185.28 33.97 14.73"

This is means you are getting about 1 count per
10 nsec. Although this depends on the PMT, a
common pulse width is in the neighborhood of 5
nsec. This is only a rule of thumb but you need
to be close to 1/10 of this flux to avoid
coincidence counting error. Photon counting rates
higher than 40 counts per 2 usec. will
unavoidably start losing counts, i.e., show signs
of saturation. There are some tricks that can
help correct for this error or at least make a
reasonable estimate (e.g., correlate count(s)
with size of the current pulse per detected
photon; overlapping pulses will generate ~ twice
total charge of single photon event). Actually,
from your laser power table, it does appear that
your are showing some saturation effects when
comparing the values from 1% power and higher.

It is also interesting that in virtually all your
measurements, your STDev are very consistently
2.5-2.6 times the expected Poisson number. In
fact, this error is so consistent that it looks
systematic:

Laser Power STDEV/sqrt(mean counts)
0.1% 2.49
0.5% 2.49
1% 2.46
2% 2.49
4% 2.53
8% 2.51
16% 2.55
32% 2.55

I wonder if your system&A/D converter are
employing an offset and performing some kind of
scaling? Even your dark counts look like a factor
of 2.5 is being applied. I have to ask, is this
really photon counting? "Hybrid photon counting"
makes me think it is not counting photons but
estimating them. Sorry, haven't used the FV1000
or I might be able to provide an answer?

As for accumulating versus using very long dwell
times, if true photon counting is being used, the
statistics are straight forward, averaging and
accumulating are equivalent. The noise should
follow the sqrt rule. Long dwell times especially
at high power will enhance bleaching and
photodynamic damage, but to a first order one can
imagine that most decay processes will still sum
their rates exponentially so that noise will
track the accumulated signal the via sqrt law.

Let us know, Stan, how things turn out and thanks
for taking the time to do the measurements.

Mario

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>I have a little update.
>Some commented that the noise in my photon counting experiment is higher
>than expected from Poisson distribution.
>
>I checked the dark counts on the PMTs (kept in the dark overnight, all light
>blocked, lasers off during the scan)
>512 x 512 pixels scan, 200 us dwell time.
>
>PMT 1 mean = 9.80 Min = 0 Max = 101 STDEV = 7.98
>PMT 2 mean = 8.75 Min = 0 Max = 126 STDEV = 8.68
>PMT 3 mean = 49.69 Min = 0 Max = 181 STDEV = 18.18
>
>PMT1 was the one used in the previous test.
>
>I expected that the dark count would have Poisson distribution, but again the
>STDEV is more than twice the SQRT(mean).
>
>It could be a questtion of the threshold set for photon counting, but I think
>only gods on Olympus would know..
>
>Stan Vitha
>Microscopy and Imaging Center
>Texas A&M University
>
>On Tue, 28 Jun 2011 11:33:13 -0400, Chen, De (NIH/NCI) [C]
><[hidden email]> wrote:
>
>>*****
>>To join, leave or search the confocal microscopy listserv, go to:
>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>>*****
>>
>>It seems, in this experiment, the count rate is way too high to see the shot
>noise effect. The Poissonian noise with
>uncertainty 1/sqrt(N), longer dowelling
>time will increase S/N. When signal is very weak, longer dowelling time will
>help.
>>
>>________________________________________
>>From: James Pawley [[hidden email]]
>>Sent: Monday, June 27, 2011 9:10 PM
>>To: [hidden email]
>>Subject: Re: averaging vs. accumulation for noise reduction - is there a
>difference?
>>>Also, I note that your STDEV is NOT equal to the sqr root of the
>>mean. I would guess that they have merely calibrated their PMT system
>>so that the number in the memory is about equal to the number of
>>single PE pulses that occurred in the pixel time.
>>
>>JP
>>

--
________________________________________________________________________________
Mario M. Moronne, Ph.D.

[hidden email]
[hidden email]

Andreas Bruckbauer

Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Which laser was used for the measurement? When i tried something similar with an Ar+ laser a while ago i noticed higher noise which was more than expected from photon counting. Using the 640 laser the noise was much reduced and followed the squarerrot rule.

best wishes

Andreas

-----Original Message-----
From: Mario <[hidden email]>
To: CONFOCALMICROSCOPY <[hidden email]>
Sent: Fri, 1 Jul 2011 20:52
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?

*****

To join, leave or search the confocal microscopy listserv, go to:

http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

*****

Stan,

I suggest that you adjust voltages and offsets on

your PMTs, test the variations of dark noise

counts with PMT voltage, and make sure that you

are operating on a linear portion of PMT gain

range. The large Max counts you are seeing is a

bit strange. Cosmic rays could generate some

anomalies but given that the STDs are

consistently more than twice the expected values

some systematic problem is more likely. What I

find the most eye catching is your getting ~ 10

dark counts per 200 usec. dwell time. That

implies your PMT(1&2) are generating 50,000

counts per second (-> 250,000 per sec. for PMT

3!). That's likely to be about 60-200x higher

than what one gets with a typical 2 inch PMT, a

discrepancy worthy of examination. If the gain

settings are too high you might end up with a lot

of spurious counts. I would certainly check with

manufacturer to get the specs. of the PMTs.

Another factor that comes to mind is cooling of

your detectors. The noise from PMTs can be helped

a lot by cooling. If PMTs are so equipped, I 'd

check to make sure that the cooling is working.

Going from room temp. down to -20°C can reduce

thermionic emission by greater than a factor of a

hundred.

There was another consideration mentioned in your

posts that you could get about 200 counts per

pixel with a 2 usec dwell in your test specimen.

"100 x 2 us dwell time, Sum of frames

area (pixels) mean STDEV signal-to-background

specimen 20812 185.28 33.97 14.73"

This is means you are getting about 1 count per

10 nsec. Although this depends on the PMT, a

common pulse width is in the neighborhood of 5

nsec. This is only a rule of thumb but you need

to be close to 1/10 of this flux to avoid

coincidence counting error. Photon counting rates

higher than 40 counts per 2 usec. will

unavoidably start losing counts, i.e., show signs

of saturation. There are some tricks that can

help correct for this error or at least make a

reasonable estimate (e.g., correlate count(s)

with size of the current pulse per detected

photon; overlapping pulses will generate ~ twice

total charge of single photon event). Actually,

from your laser power table, it does appear that

your are showing some saturation effects when

comparing the values from 1% power and higher.

It is also interesting that in virtually all your

measurements, your STDev are very consistently

2.5-2.6 times the expected Poisson number. In

fact, this error is so consistent that it looks

systematic:

Laser Power STDEV/sqrt(mean counts)

0.1% 2.49

0.5% 2.49

1% 2.46

2% 2.49

4% 2.53

8% 2.51

16% 2.55

32% 2.55

I wonder if your system&A/D converter are

employing an offset and performing some kind of

scaling? Even your dark counts look like a factor

of 2.5 is being applied. I have to ask, is this

really photon counting? "Hybrid photon counting"

makes me think it is not counting photons but

estimating them. Sorry, haven't used the FV1000

or I might be able to provide an answer?

As for accumulating versus using very long dwell

times, if true photon counting is being used, the

statistics are straight forward, averaging and

accumulating are equivalent. The noise should

follow the sqrt rule. Long dwell times especially

at high power will enhance bleaching and

photodynamic damage, but to a first order one can

imagine that most decay processes will still sum

their rates exponentially so that noise will

track the accumulated signal the via sqrt law.

Let us know, Stan, how things turn out and thanks

for taking the time to do the measurements.

Mario

>*****

>To join, leave or search the confocal microscopy listserv, go to:

>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>*****

>

>I have a little update.

>Some commented that the noise in my photon counting experiment is higher

>than expected from Poisson distribution.

>

>I checked the dark counts on the PMTs (kept in the dark overnight, all light

>blocked, lasers off during the scan)

>512 x 512 pixels scan, 200 us dwell time.

>

>PMT 1 mean = 9.80 Min = 0 Max = 101 STDEV = 7.98

>PMT 2 mean = 8.75 Min = 0 Max = 126 STDEV = 8.68

>PMT 3 mean = 49.69 Min = 0 Max = 181 STDEV = 18.18

>

>PMT1 was the one used in the previous test.

>

>I expected that the dark count would have Poisson distribution, but again the

>STDEV is more than twice the SQRT(mean).

>

>It could be a questtion of the threshold set for photon counting, but I think

>only gods on Olympus would know..

>

>Stan Vitha

>Microscopy and Imaging Center

>Texas A&M University

>

>On Tue, 28 Jun 2011 11:33:13 -0400, Chen, De (NIH/NCI) [C]

><[hidden email]> wrote:

>

>>*****

>>To join, leave or search the confocal microscopy listserv, go to:

>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>>*****

>>

>>It seems, in this experiment, the count rate is way too high to see the shot

>noise effect. The Poissonian noise with

>uncertainty 1/sqrt(N), longer dowelling

>time will increase S/N. When signal is very weak, longer dowelling time will

>help.

>>

>>________________________________________

>>From: James Pawley [[hidden email]]

>>Sent: Monday, June 27, 2011 9:10 PM

>>To: [hidden email]

>>Subject: Re: averaging vs. accumulation for noise reduction - is there a

>difference?

>>>Also, I note that your STDEV is NOT equal to the sqr root of the

>>mean. I would guess that they have merely calibrated their PMT system

>>so that the number in the memory is about equal to the number of

>>single PE pulses that occurred in the pixel time.

>>

>>JP

>>

--

________________________________________________________________________________

Mario M. Moronne, Ph.D.

[hidden email]

[hidden email]

Stanislav Vitha

Re: averaging vs. accumulation for noise reduction - is there a difference?

In reply to this post by Stanislav Vitha

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

I used the Ar laser (488 nm);
I will run a test with a HeNe (I have 543 abd 633nm) as you suggest.
nevertheless, the high STDEV in dark count measurements indicate that there
is some systematic issue not related to laser stability, but rather related to
the way the photons are "counted" (or estimated).
The PMTs are not actively cooled. Higher dark counts are to be expected, but
they should still have Poisson distribution, I think.

I realize this thread is veering slightly off the original topic, buth here is more
on the photon counting on Olympus FV1000:

I still have not heard from Olympus how their "hybrid photon counting" actually
works. I am not happy with their lack of openness.

Nevertheless, looking into the system settings of the confocal software, there
is a tab for Photon counting settings.
Here is the screenshot from the help file in their freely downloadable viewer:
http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-
tests/FV1000Photoncount.jpg

On the actual microscope, PMTs work at 1050 Volts in Photon Counting mode,
the five photon count thresholds for PMT 1 are listed as:

level Value
(/1023) (0-63)
------------------------------
110 1
117 2
127 3
144 4
170 5

My conclusion is that we do not have true photon counting but rather photon
count estimation from the analog signals.

Stan Vitha
Microscopy and Imaging Center
Texas A&M University

Stan Vitha

On Sat, 2 Jul 2011 17:30:03 -0400, Andreas Bruckbauer
<[hidden email]> wrote:

>*****
>To join, leave or search the confocal microscopy listserv, go to:
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>*****
>
>
> Which laser was used for the measurement? When i tried something similar
with an Ar+ laser a while ago i noticed higher noise which was more than
expected from photon counting. Using the 640 laser the noise was much
reduced and followed the squarerrot rule.

>
>best wishes
>
>Andreas
>
>
>
>
>
>
>
>
>-----Original Message-----
>From: Mario <[hidden email]>
>To: CONFOCALMICROSCOPY <[hidden email]>
>Sent: Fri, 1 Jul 2011 20:52
>Subject: Re: averaging vs. accumulation for noise reduction - is there a

difference?

>
>
>*****
>
>To join, leave or search the confocal microscopy listserv, go to:
>
>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>
>*****
>
>
>
>Stan,
>
>
>
>I suggest that you adjust voltages and offsets on
>
>your PMTs, test the variations of dark noise
>
>counts with PMT voltage, and make sure that you
>
>are operating on a linear portion of PMT gain
>
>range. The large Max counts you are seeing is a
>
>bit strange. Cosmic rays could generate some
>
>anomalies but given that the STDs are
>
>consistently more than twice the expected values
>
>some systematic problem is more likely. What I
>
>find the most eye catching is your getting ~ 10
>
>dark counts per 200 usec. dwell time. That
>
>implies your PMT(1&2) are generating 50,000
>
>counts per second (-> 250,000 per sec. for PMT
>
>3!). That's likely to be about 60-200x higher
>
>than what one gets with a typical 2 inch PMT, a
>
>discrepancy worthy of examination. If the gain
>
>settings are too high you might end up with a lot
>
>of spurious counts. I would certainly check with
>
>manufacturer to get the specs. of the PMTs.
>
>
>
>Another factor that comes to mind is cooling of
>
>your detectors. The noise from PMTs can be helped
>
>a lot by cooling. If PMTs are so equipped, I 'd
>
>check to make sure that the cooling is working.
>
>Going from room temp. down to -20Â°C can reduce
>
>thermionic emission by greater than a factor of a
>
>hundred.
>
>
>
>There was another consideration mentioned in your
>
>posts that you could get about 200 counts per
>
>pixel with a 2 usec dwell in your test specimen.
>
>
>
>"100 x 2 us dwell time, Sum of frames
>
> area (pixels) mean STDEV signal-to-background
>
>specimen 20812 185.28 33.97 14.73"
>
>
>
>This is means you are getting about 1 count per
>
>10 nsec. Although this depends on the PMT, a
>
>common pulse width is in the neighborhood of 5
>
>nsec. This is only a rule of thumb but you need
>
>to be close to 1/10 of this flux to avoid
>
>coincidence counting error. Photon counting rates
>
>higher than 40 counts per 2 usec. will
>
>unavoidably start losing counts, i.e., show signs
>
>of saturation. There are some tricks that can
>
>help correct for this error or at least make a
>
>reasonable estimate (e.g., correlate count(s)
>
>with size of the current pulse per detected
>
>photon; overlapping pulses will generate ~ twice
>
>total charge of single photon event). Actually,
>
>from your laser power table, it does appear that
>
>your are showing some saturation effects when
>
>comparing the values from 1% power and higher.
>
>
>
>It is also interesting that in virtually all your
>
>measurements, your STDev are very consistently
>
>2.5-2.6 times the expected Poisson number. In
>
>fact, this error is so consistent that it looks
>
>systematic:
>
>
>
>Laser Power STDEV/sqrt(mean counts)
>
>0.1% 2.49
>
>0.5% 2.49
>
>1% 2.46
>
>2% 2.49
>
>4% 2.53
>
>8% 2.51
>
>16% 2.55
>
>32% 2.55
>
>
>
>I wonder if your system&A/D converter are
>
>employing an offset and performing some kind of
>
>scaling? Even your dark counts look like a factor
>
>of 2.5 is being applied. I have to ask, is this
>
>really photon counting? "Hybrid photon counting"
>
>makes me think it is not counting photons but
>
>estimating them. Sorry, haven't used the FV1000
>
>or I might be able to provide an answer?
>
>
>
>As for accumulating versus using very long dwell
>
>times, if true photon counting is being used, the
>
>statistics are straight forward, averaging and
>
>accumulating are equivalent. The noise should
>
>follow the sqrt rule. Long dwell times especially
>
>at high power will enhance bleaching and
>
>photodynamic damage, but to a first order one can
>
>imagine that most decay processes will still sum
>
>their rates exponentially so that noise will
>
>track the accumulated signal the via sqrt law.
>
>
>
>Let us know, Stan, how things turn out and thanks
>
>for taking the time to do the measurements.
>
>
>
>Mario
>
>
>
>
>
>>*****
>
>>To join, leave or search the confocal microscopy listserv, go to:
>
>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>
>>*****
>
>>
>
>>I have a little update.
>
>>Some commented that the noise in my photon counting experiment is higher
>
>>than expected from Poisson distribution.
>
>>
>
>>I checked the dark counts on the PMTs (kept in the dark overnight, all light
>
>>blocked, lasers off during the scan)
>
>>512 x 512 pixels scan, 200 us dwell time.
>
>>
>
>>PMT 1 mean = 9.80 Min = 0 Max = 101 STDEV = 7.98
>
>>PMT 2 mean = 8.75 Min = 0 Max = 126 STDEV = 8.68
>
>>PMT 3 mean = 49.69 Min = 0 Max = 181 STDEV = 18.18
>
>>
>
>>PMT1 was the one used in the previous test.
>
>>
>
>>I expected that the dark count would have Poisson distribution, but again

the

>
>>STDEV is more than twice the SQRT(mean).
>
>>
>
>>It could be a questtion of the threshold set for photon counting, but I think
>
>>only gods on Olympus would know..
>
>>
>
>>Stan Vitha
>
>>Microscopy and Imaging Center
>
>>Texas A&M University
>
>>
>
>>On Tue, 28 Jun 2011 11:33:13 -0400, Chen, De (NIH/NCI) [C]
>
>><[hidden email]> wrote:
>
>>
>
>>>*****
>
>>>To join, leave or search the confocal microscopy listserv, go to:
>
>>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
>
>>>*****
>
>>>
>
>>>It seems, in this experiment, the count rate is way too high to see the shot
>
>>noise effect. The Poissonian noise with
>
>>uncertainty 1/sqrt(N), longer dowelling
>
>>time will increase S/N. When signal is very weak, longer dowelling time will
>
>>help.
>
>>>
>
>>>________________________________________
>
>>>From: James Pawley [[hidden email]]
>
>>>Sent: Monday, June 27, 2011 9:10 PM
>
>>>To: [hidden email]
>
>>>Subject: Re: averaging vs. accumulation for noise reduction - is there a
>
>>difference?
>
>>>>Also, I note that your STDEV is NOT equal to the sqr root of the
>
>>>mean. I would guess that they have merely calibrated their PMT system
>
>>>so that the number in the memory is about equal to the number of
>
>>>single PE pulses that occurred in the pixel time.
>
>>>
>
>>>JP
>
>>>
>
>
>
>
>
>--
>
>_________________________________________________________________

_______________

>
>Mario M. Moronne, Ph.D.
>
>
>
>[hidden email]
>
>[hidden email]
>
>
>

Andreas Bruckbauer

Re: averaging vs. accumulation for noise reduction - is there a difference?

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

If it is "true photon counting" I would expect no dark counts (no photons). Otherwise you would need to subtract the stdev of the dark signal from the stedev of the measured signal geometrically and then i would expect it to follow poisson statistics.

best wishes

Andreas

-----Original Message-----
From: Stanislav Vitha <[hidden email]>
To: CONFOCALMICROSCOPY <[hidden email]>
Sent: Tue, 5 Jul 2011 16:49
Subject: Re: averaging vs. accumulation for noise reduction - is there a difference?

*****

To join, leave or search the confocal microscopy listserv, go to:

http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

*****

I used the Ar laser (488 nm);

I will run a test with a HeNe (I have 543 abd 633nm) as you suggest.

nevertheless, the high STDEV in dark count measurements indicate that there

is some systematic issue not related to laser stability, but rather related to

the way the photons are "counted" (or estimated).

The PMTs are not actively cooled. Higher dark counts are to be expected, but

they should still have Poisson distribution, I think.

I realize this thread is veering slightly off the original topic, buth here is

more

on the photon counting on Olympus FV1000:

I still have not heard from Olympus how their "hybrid photon counting" actually

works. I am not happy with their lack of openness.

Nevertheless, looking into the system settings of the confocal software, there

is a tab for Photon counting settings.

Here is the screenshot from the help file in their freely downloadable viewer:

http://microscopy.tamu.edu/instruments/light-microscopy/fv1000-

tests/FV1000Photoncount.jpg

On the actual microscope, PMTs work at 1050 Volts in Photon Counting mode,

the five photon count thresholds for PMT 1 are listed as:

level Value

(/1023) (0-63)

------------------------------

110 1

117 2

127 3

144 4

170 5

My conclusion is that we do not have true photon counting but rather photon

count estimation from the analog signals.

Stan Vitha

Microscopy and Imaging Center

Texas A&M University

Stan Vitha

On Sat, 2 Jul 2011 17:30:03 -0400, Andreas Bruckbauer

<[hidden email]> wrote:

>*****

>To join, leave or search the confocal microscopy listserv, go to:

>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>*****

>

>

> Which laser was used for the measurement? When i tried something similar

with an Ar+ laser a while ago i noticed higher noise which was more than

expected from photon counting. Using the 640 laser the noise was much

reduced and followed the squarerrot rule.

>

>best wishes

>

>Andreas

>

>

>

>

>

>

>

>

>-----Original Message-----

>From: Mario <[hidden email]>

>To: CONFOCALMICROSCOPY <[hidden email]>

>Sent: Fri, 1 Jul 2011 20:52

>Subject: Re: averaging vs. accumulation for noise reduction - is there a

difference?

>

>

>*****

>

>To join, leave or search the confocal microscopy listserv, go to:

>

>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>

>*****

>

>

>

>Stan,

>

>

>

>I suggest that you adjust voltages and offsets on

>

>your PMTs, test the variations of dark noise

>

>counts with PMT voltage, and make sure that you

>

>are operating on a linear portion of PMT gain

>

>range. The large Max counts you are seeing is a

>

>bit strange. Cosmic rays could generate some

>

>anomalies but given that the STDs are

>

>consistently more than twice the expected values

>

>some systematic problem is more likely. What I

>

>find the most eye catching is your getting ~ 10

>

>dark counts per 200 usec. dwell time. That

>

>implies your PMT(1&2) are generating 50,000

>

>counts per second (-> 250,000 per sec. for PMT

>

>3!). That's likely to be about 60-200x higher

>

>than what one gets with a typical 2 inch PMT, a

>

>discrepancy worthy of examination. If the gain

>

>settings are too high you might end up with a lot

>

>of spurious counts. I would certainly check with

>

>manufacturer to get the specs. of the PMTs.

>

>

>

>Another factor that comes to mind is cooling of

>

>your detectors. The noise from PMTs can be helped

>

>a lot by cooling. If PMTs are so equipped, I 'd

>

>check to make sure that the cooling is working.

>

>Going from room temp. down to -20Â°C can reduce

>

>thermionic emission by greater than a factor of a

>

>hundred.

>

>

>

>There was another consideration mentioned in your

>

>posts that you could get about 200 counts per

>

>pixel with a 2 usec dwell in your test specimen.

>

>

>

>"100 x 2 us dwell time, Sum of frames

>

> area (pixels) mean STDEV signal-to-background

>

>specimen 20812 185.28 33.97 14.73"

>

>

>

>This is means you are getting about 1 count per

>

>10 nsec. Although this depends on the PMT, a

>

>common pulse width is in the neighborhood of 5

>

>nsec. This is only a rule of thumb but you need

>

>to be close to 1/10 of this flux to avoid

>

>coincidence counting error. Photon counting rates

>

>higher than 40 counts per 2 usec. will

>

>unavoidably start losing counts, i.e., show signs

>

>of saturation. There are some tricks that can

>

>help correct for this error or at least make a

>

>reasonable estimate (e.g., correlate count(s)

>

>with size of the current pulse per detected

>

>photon; overlapping pulses will generate ~ twice

>

>total charge of single photon event). Actually,

>

>from your laser power table, it does appear that

>

>your are showing some saturation effects when

>

>comparing the values from 1% power and higher.

>

>

>

>It is also interesting that in virtually all your

>

>measurements, your STDev are very consistently

>

>2.5-2.6 times the expected Poisson number. In

>

>fact, this error is so consistent that it looks

>

>systematic:

>

>

>

>Laser Power STDEV/sqrt(mean counts)

>

>0.1% 2.49

>

>0.5% 2.49

>

>1% 2.46

>

>2% 2.49

>

>4% 2.53

>

>8% 2.51

>

>16% 2.55

>

>32% 2.55

>

>

>

>I wonder if your system&A/D converter are

>

>employing an offset and performing some kind of

>

>scaling? Even your dark counts look like a factor

>

>of 2.5 is being applied. I have to ask, is this

>

>really photon counting? "Hybrid photon counting"

>

>makes me think it is not counting photons but

>

>estimating them. Sorry, haven't used the FV1000

>

>or I might be able to provide an answer?

>

>

>

>As for accumulating versus using very long dwell

>

>times, if true photon counting is being used, the

>

>statistics are straight forward, averaging and

>

>accumulating are equivalent. The noise should

>

>follow the sqrt rule. Long dwell times especially

>

>at high power will enhance bleaching and

>

>photodynamic damage, but to a first order one can

>

>imagine that most decay processes will still sum

>

>their rates exponentially so that noise will

>

>track the accumulated signal the via sqrt law.

>

>

>

>Let us know, Stan, how things turn out and thanks

>

>for taking the time to do the measurements.

>

>

>

>Mario

>

>

>

>

>

>>*****

>

>>To join, leave or search the confocal microscopy listserv, go to:

>

>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>

>>*****

>

>>

>

>>I have a little update.

>

>>Some commented that the noise in my photon counting experiment is higher

>

>>than expected from Poisson distribution.

>

>>

>

>>I checked the dark counts on the PMTs (kept in the dark overnight, all light

>

>>blocked, lasers off during the scan)

>

>>512 x 512 pixels scan, 200 us dwell time.

>

>>

>

>>PMT 1 mean = 9.80 Min = 0 Max = 101 STDEV = 7.98

>

>>PMT 2 mean = 8.75 Min = 0 Max = 126 STDEV = 8.68

>

>>PMT 3 mean = 49.69 Min = 0 Max = 181 STDEV = 18.18

>

>>

>

>>PMT1 was the one used in the previous test.

>

>>

>

>>I expected that the dark count would have Poisson distribution, but again

the

>

>>STDEV is more than twice the SQRT(mean).

>

>>

>

>>It could be a questtion of the threshold set for photon counting, but I think

>

>>only gods on Olympus would know..

>

>>

>

>>Stan Vitha

>

>>Microscopy and Imaging Center

>

>>Texas A&M University

>

>>

>

>>On Tue, 28 Jun 2011 11:33:13 -0400, Chen, De (NIH/NCI) [C]

>

>><[hidden email]> wrote:

>

>>

>

>>>*****

>

>>>To join, leave or search the confocal microscopy listserv, go to:

>

>>>http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy

>

>>>*****

>

>>>

>

>>>It seems, in this experiment, the count rate is way too high to see the shot

>

>>noise effect. The Poissonian noise with

>

>>uncertainty 1/sqrt(N), longer dowelling

>

>>time will increase S/N. When signal is very weak, longer dowelling time will

>

>>help.

>

>>>

>

>>>________________________________________

>

>>>From: James Pawley [[hidden email]]

>

>>>Sent: Monday, June 27, 2011 9:10 PM

>

>>>To: [hidden email]

>

>>>Subject: Re: averaging vs. accumulation for noise reduction - is there a

>

>>difference?

>

>>>>Also, I note that your STDEV is NOT equal to the sqr root of the

>

>>>mean. I would guess that they have merely calibrated their PMT system

>

>>>so that the number in the memory is about equal to the number of

>

>>>single PE pulses that occurred in the pixel time.

>

>>>

>

>>>JP

>

>>>

>

>

>

>

>

>--

>

>_________________________________________________________________

_______________

>

>Mario M. Moronne, Ph.D.

>

>

>

>[hidden email]

>

>[hidden email]

>

>

>