Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
Locked
|
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Hi listers, I was reading a multiphoton fluorescence imaging paper earlier today, and found in the method section that the authors used 15 us pixel clock (I assume this is the dwell time, thus the pixel
rate is 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: 1)
Is it generally true that the AO sampling rate has to be at least twice of signal frequency (National Instruments suggested >10X of signal rate), which in this case is limited by the
amp rate, i.e. 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. 2)
The pixel rate cannot exceed the half of the signal rate (amp bandwidth) to catch all spatial varying signal from the sample. What is considered as appropriate pixel rate given the
amp bandwidth? Here, the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. Can I assume either they have a typo somewhere, or they compromised the measurement by effectively applying a low pass filter to the images, or I am just
being completely idiotic. 3)
In general , why single photon counting modules (APD or PMT based) is not as popular as analog PMTs in multiphoton fluorescence microscopy? Do they not have less constrains for example
bandwidth than analog ones? I look forward to hearing from you guys. This really bugged me quite a few hours. Thanks in advance, |
 
|
Michael Giacomelli |
Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
|
|
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** 1) You don't have to satisfy Nyquist; you can undersample if you want, but then you are throwing away resolution. 2) If you are limited by analog bandwidth, then as you scan along the fast axis, the PSF of each pixel is elongated along the scan axis by the low pass filtering in the electronics. 15 us dwell time is actually 66 kHz, so they are slightly oversampled (Nyquist would be 40 kHz), but the bandwidth on amps is usually 3dB, so there are probably frequencies well above 20 kHz. 3) Single photon counting extracts more signal per photon, but typically can only handle extremely low numbers of photons per second. Operated in analog mode, most PMTs are several orders of magnitude faster than when in counting mode. If you are not concerned about illumination power, usually you get a better image per unit time from analog mode. (however, if you are limited by power, counting is much better) Hope that helps. Mike On Wed, Oct 26, 2016 at 3:56 PM, Yan, Lu <[hidden email]> wrote: > ***** To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on > http://www.imgur.com and include the link in your posting. ***** > > Hi listers, > > > > I was reading a multiphoton fluorescence imaging paper earlier today, and > found in the method section that the authors used 15 us pixel clock (I > assume this is the dwell time, thus the pixel rate is 1/15us=~667 kHz?), a 1 > MHz pulsed laser, AO (to drive scanning) sampling rate ~1MHz, but a 20 kHz > PMT amp. I got a little confused, so my questions are: > > > > 1) Is it generally true that the AO sampling rate has to be at least > twice of signal frequency (National Instruments suggested >10X of signal > rate), which in this case is limited by the amp rate, i.e. 20kHz? In this > paper particularly, they are massively oversampling which is OK I guess. > > > > 2) The pixel rate cannot exceed the half of the signal rate (amp > bandwidth) to catch all spatial varying signal from the sample. What is > considered as appropriate pixel rate given the amp bandwidth? Here, the amp > bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. > Can I assume either they have a typo somewhere, or they compromised the > measurement by effectively applying a low pass filter to the images, or I am > just being completely idiotic. > > > > 3) In general , why single photon counting modules (APD or PMT based) is > not as popular as analog PMTs in multiphoton fluorescence microscopy? Do > they not have less constrains for example bandwidth than analog ones? > > > > I look forward to hearing from you guys. This really bugged me quite a few > hours. > > > > Thanks in advance, > Lu |
|
|
0000001ed7f52e4a-dmarc-request |
Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
|
|
In reply to this post by Lu Yan
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Lu, Your question: Why single photon counting modules (APD or PMT based) is not as popular as analog PMTs in multiphoton fluorescence microscopy? The answer is the size (area) of the detector, you want large area detectors to collect all the scattered light, at least when you use two-photon excitation to image deep into tissue with non-descanned detectors. Best wishes Andreas From: [hidden email] Sent: 26/10/2016 20:57 To: [hidden email] Subject: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth Hi listers, I was reading a multiphoton fluorescence imaging paper earlier today, and found in the method section that the authors used 15 us pixel clock (I assume this is the dwell time, thus the pixel
rate is 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: 1)
Is it generally true that the AO sampling rate has to be at least twice of signal frequency (National Instruments suggested >10X of signal rate), which in this case is limited by the
amp rate, i.e. 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. 2)
The pixel rate cannot exceed the half of the signal rate (amp bandwidth) to catch all spatial varying signal from the sample. What is considered as appropriate pixel rate given the
amp bandwidth? Here, the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. Can I assume either they have a typo somewhere, or they compromised the measurement by effectively applying a low pass filter to the images, or I am just
being completely idiotic. 3)
In general , why single photon counting modules (APD or PMT based) is not as popular as analog PMTs in multiphoton fluorescence microscopy? Do they not have less constrains for example
bandwidth than analog ones? I look forward to hearing from you guys. This really bugged me quite a few hours. Thanks in advance, |
|
|
In reply to this post by Michael Giacomelli
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Mike, Thanks for your reply. I am sorry for the mis-calculation, I meant to write 0.0667MHz... :J Anyway, for the second comment you put, would it be 'undersanpled' instead of 'oversampled' if your pixel rate is 66kHz and amp 20 kHz? Supposed that you have a line sample (1x8) with intensity variation pixel by pixel, e.g. something like this: 'o x o x o x o x', where o denotes fluorophore on-site, and x means no signal. If you scan at 66 kHz, then would you be able to catch the second pixel 'x' and probably the third pixel 'o' because your pmt output is only 20 kHz, i.e. more than three times slower than pixel scanning rate? So effectively, will you not probably get an image something like ' o o o x x x o o'? Should the Nyquist pixel rate be 10 kHz? Thanks, Lu -----Original Message----- From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Michael Giacomelli Sent: Wednesday, October 26, 2016 4:30 PM To: [hidden email] Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** 1) You don't have to satisfy Nyquist; you can undersample if you want, but then you are throwing away resolution. 2) If you are limited by analog bandwidth, then as you scan along the fast axis, the PSF of each pixel is elongated along the scan axis by the low pass filtering in the electronics. 15 us dwell time is actually 66 kHz, so they are slightly oversampled (Nyquist would be 40 kHz), but the bandwidth on amps is usually 3dB, so there are probably frequencies well above 20 kHz. 3) Single photon counting extracts more signal per photon, but typically can only handle extremely low numbers of photons per second. Operated in analog mode, most PMTs are several orders of magnitude faster than when in counting mode. If you are not concerned about illumination power, usually you get a better image per unit time from analog mode. (however, if you are limited by power, counting is much better) Hope that helps. Mike On Wed, Oct 26, 2016 at 3:56 PM, Yan, Lu <[hidden email]> wrote: > ***** To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on > http://www.imgur.com and include the link in your posting. ***** > > Hi listers, > > > > I was reading a multiphoton fluorescence imaging paper earlier today, > and found in the method section that the authors used 15 us pixel > clock (I assume this is the dwell time, thus the pixel rate is > 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) > sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: > > > > 1) Is it generally true that the AO sampling rate has to be at least > twice of signal frequency (National Instruments suggested >10X of > signal rate), which in this case is limited by the amp rate, i.e. > 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. > > > > 2) The pixel rate cannot exceed the half of the signal rate (amp > bandwidth) to catch all spatial varying signal from the sample. What > is considered as appropriate pixel rate given the amp bandwidth? Here, > the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. > Can I assume either they have a typo somewhere, or they compromised > the measurement by effectively applying a low pass filter to the > images, or I am just being completely idiotic. > > > > 3) In general , why single photon counting modules (APD or PMT based) is > not as popular as analog PMTs in multiphoton fluorescence microscopy? > Do they not have less constrains for example bandwidth than analog ones? > > > > I look forward to hearing from you guys. This really bugged me quite a > few hours. > > > > Thanks in advance, > Lu |
|
|
In reply to this post by 0000001ed7f52e4a-dmarc-request
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Andreas, Thanks much for your answering. Now I get that detector area matters, but what did you mean by non-descanned detectors? Are they not all descanned in multiphoton microscopy like confocal does? Thanks, Lu From: Confocal Microscopy List [mailto:[hidden email]]
On Behalf Of Andreas Bruckbauer ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Lu, From: [hidden email] ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Hi listers, I was reading a multiphoton fluorescence imaging paper earlier today, and found in the method section that the authors used 15 us pixel clock (I assume this is the dwell time, thus the pixel
rate is 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: 1)
Is it generally true that the AO sampling rate has to be at least twice of signal frequency (National Instruments suggested >10X of signal rate), which in this case is limited by the
amp rate, i.e. 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. 2)
The pixel rate cannot exceed the half of the signal rate (amp bandwidth) to catch all spatial varying signal from the sample. What is considered as appropriate pixel rate given the
amp bandwidth? Here, the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. Can I assume either they have a typo somewhere, or they compromised the measurement by effectively applying a low pass filter to the images, or I am just
being completely idiotic. 3)
In general , why single photon counting modules (APD or PMT based) is not as popular as analog PMTs in multiphoton fluorescence microscopy? Do they not have less constrains for example
bandwidth than analog ones? I look forward to hearing from you guys. This really bugged me quite a few hours. Thanks in advance, |
|
|
Michael Giacomelli |
Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
|
|
In reply to this post by Lu Yan
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Lu, No, if the sampling rate (66 khz) is higher than twice the analog bandwidth (20 khz), then the image is oversampled (at least assuming the optical resolution is much higher than the analog bandwidth limit). However, since the analog bandwidth in this case is probably just the 3dB point, the image will contain some information above 20khz, although it will be attenuated. In your example, it depends on how rapidly you are scanning and what your optical resolution is. If you really mean 15 us per point, then your pattern frequency is 1/30us = 33.333Khz, which is a little bit higher than your 3dB point. Many detectors roll off 6dB per factor of two past the 3dB point, so you would probably be attenuated about 7-8dB. If the points you are imaging are bright, you would still see them. Mike On Wed, Oct 26, 2016 at 5:10 PM, Yan, Lu <[hidden email]> wrote: > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > Hi Mike, > > Thanks for your reply. I am sorry for the mis-calculation, I meant to write 0.0667MHz... :J > > Anyway, for the second comment you put, would it be 'undersanpled' instead of 'oversampled' if your pixel rate is 66kHz and amp 20 kHz? Supposed that you have a line sample (1x8) with intensity variation pixel by pixel, e.g. something like this: 'o x o x o x o x', where o denotes fluorophore on-site, and x means no signal. If you scan at 66 kHz, then would you be able to catch the second pixel 'x' and probably the third pixel 'o' because your pmt output is only 20 kHz, i.e. more than three times slower than pixel scanning rate? So effectively, will you not probably get an image something like ' o o o x x x o o'? Should the Nyquist pixel rate be 10 kHz? > > Thanks, > Lu > > > -----Original Message----- > From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Michael Giacomelli > Sent: Wednesday, October 26, 2016 4:30 PM > To: [hidden email] > Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth > > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > 1) You don't have to satisfy Nyquist; you can undersample if you want, but then you are throwing away resolution. > > 2) If you are limited by analog bandwidth, then as you scan along the fast axis, the PSF of each pixel is elongated along the scan axis by the low pass filtering in the electronics. 15 us dwell time is actually 66 kHz, so they are slightly oversampled (Nyquist would be 40 kHz), but the bandwidth on amps is usually 3dB, so there are probably frequencies well above 20 kHz. > > 3) Single photon counting extracts more signal per photon, but typically can only handle extremely low numbers of photons per second. > Operated in analog mode, most PMTs are several orders of magnitude faster than when in counting mode. If you are not concerned about illumination power, usually you get a better image per unit time from analog mode. (however, if you are limited by power, counting is much > better) > > Hope that helps. > Mike > > On Wed, Oct 26, 2016 at 3:56 PM, Yan, Lu <[hidden email]> wrote: >> ***** To join, leave or search the confocal microscopy listserv, go to: >> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on >> http://www.imgur.com and include the link in your posting. ***** >> >> Hi listers, >> >> >> >> I was reading a multiphoton fluorescence imaging paper earlier today, >> and found in the method section that the authors used 15 us pixel >> clock (I assume this is the dwell time, thus the pixel rate is >> 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) >> sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: >> >> >> >> 1) Is it generally true that the AO sampling rate has to be at least >> twice of signal frequency (National Instruments suggested >10X of >> signal rate), which in this case is limited by the amp rate, i.e. >> 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. >> >> >> >> 2) The pixel rate cannot exceed the half of the signal rate (amp >> bandwidth) to catch all spatial varying signal from the sample. What >> is considered as appropriate pixel rate given the amp bandwidth? Here, >> the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. >> Can I assume either they have a typo somewhere, or they compromised >> the measurement by effectively applying a low pass filter to the >> images, or I am just being completely idiotic. >> >> >> >> 3) In general , why single photon counting modules (APD or PMT based) is >> not as popular as analog PMTs in multiphoton fluorescence microscopy? >> Do they not have less constrains for example bandwidth than analog ones? >> >> >> >> I look forward to hearing from you guys. This really bugged me quite a >> few hours. >> >> >> >> Thanks in advance, >> Lu |
|
|
Craig Brideau |
Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
|
|
In reply to this post by Lu Yan
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Multi-photon usually uses non-descanned detection, which has the primary dichroic between the scan mirrors and the objective. The fluorescent light from the sample is therefore still moving when it reaches the detector. As a result it is important that the detector be large so it captures all the light that is moving around. Confocal uses descanned detectors, where the primary dichroic is between the laser and the scan mirrors. In this configuration the fluorescent light from the sample passes back through the scan mirrors before being collected. This makes the light stationary, which is vital for focusing through a confocal pinhole. CraigOn Wed, Oct 26, 2016 at 3:16 PM, Yan, Lu <[hidden email]> wrote: ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/ |
|
|
In reply to this post by Michael Giacomelli
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Mike, Thanks for your explanation. I understood that if the sampling rate is higher than twice the analog bandwidth (20kHz), because it simply means sampling more than two points per analog pulse. But in the example, as I understood, the sampling rate (set in the DAQ) is >1MHz per channel (obviously fulfill the 'oversampling' condition), but the pixel rate is 66 kHz, in this case, should I think of 'undersampled' or 'oversampled'? For the detector, can I interpret the analog bandwidth 20 kHz in this way (in the time domain): (1) The detector receives fluorescence photons and generate the photocurrent pulse with a raise time and fall time. The output signal pulse (the first pulse) has long tail (probably longer than 0.05 ms?) so that any other pulses coming in within 1/20kHz=0.05ms time window after the first pulse will not be detected or will be attenuated accordingly. (2) Therefore, if the first pulse is coming from a pixel (see, pixel #1), and the laser quickly moves (quicker than 0.05 ms, e.g. 30 us) to pixel #2 where there is no fluorophore. Now, the DAQ channel samples the detector again, and assigns the signal strength to pixel #2. Since the detector has 20 kHz bandwidth, if (1) holds, then would the channel get some value due to the long tail of the first pulse (coming from previous pixel)? If this is true, then the image is not the correct representation of the object, i.e. pixel #2 actually does not have fluorophores but is showed up in the image as if there are some. Thanks, Lu -----Original Message----- From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Michael Giacomelli Sent: Wednesday, October 26, 2016 5:40 PM To: [hidden email] Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth ***** To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Lu, No, if the sampling rate (66 khz) is higher than twice the analog bandwidth (20 khz), then the image is oversampled (at least assuming the optical resolution is much higher than the analog bandwidth limit). However, since the analog bandwidth in this case is probably just the 3dB point, the image will contain some information above 20khz, although it will be attenuated. In your example, it depends on how rapidly you are scanning and what your optical resolution is. If you really mean 15 us per point, then your pattern frequency is 1/30us = 33.333Khz, which is a little bit higher than your 3dB point. Many detectors roll off 6dB per factor of two past the 3dB point, so you would probably be attenuated about 7-8dB. If the points you are imaging are bright, you would still see them. Mike On Wed, Oct 26, 2016 at 5:10 PM, Yan, Lu <[hidden email]> wrote: > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > Hi Mike, > > Thanks for your reply. I am sorry for the mis-calculation, I meant to > write 0.0667MHz... :J > > Anyway, for the second comment you put, would it be 'undersanpled' instead of 'oversampled' if your pixel rate is 66kHz and amp 20 kHz? Supposed that you have a line sample (1x8) with intensity variation pixel by pixel, e.g. something like this: 'o x o x o x o x', where o denotes fluorophore on-site, and x means no signal. If you scan at 66 kHz, then would you be able to catch the second pixel 'x' and probably the third pixel 'o' because your pmt output is only 20 kHz, i.e. more than three times slower than pixel scanning rate? So effectively, will you not probably get an image something like ' o o o x x x o o'? Should the Nyquist pixel rate be 10 kHz? > > Thanks, > Lu > > > -----Original Message----- > From: Confocal Microscopy List > [mailto:[hidden email]] On Behalf Of Michael > Giacomelli > Sent: Wednesday, October 26, 2016 4:30 PM > To: [hidden email] > Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser > rep. rate, and detector/amp bandwidth > > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > 1) You don't have to satisfy Nyquist; you can undersample if you want, but then you are throwing away resolution. > > 2) If you are limited by analog bandwidth, then as you scan along the fast axis, the PSF of each pixel is elongated along the scan axis by the low pass filtering in the electronics. 15 us dwell time is actually 66 kHz, so they are slightly oversampled (Nyquist would be 40 kHz), but the bandwidth on amps is usually 3dB, so there are probably frequencies well above 20 kHz. > > 3) Single photon counting extracts more signal per photon, but typically can only handle extremely low numbers of photons per second. > Operated in analog mode, most PMTs are several orders of magnitude > faster than when in counting mode. If you are not concerned about > illumination power, usually you get a better image per unit time from > analog mode. (however, if you are limited by power, counting is much > better) > > Hope that helps. > Mike > > On Wed, Oct 26, 2016 at 3:56 PM, Yan, Lu <[hidden email]> wrote: >> ***** To join, leave or search the confocal microscopy listserv, go to: >> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on >> http://www.imgur.com and include the link in your posting. ***** >> >> Hi listers, >> >> >> >> I was reading a multiphoton fluorescence imaging paper earlier today, >> and found in the method section that the authors used 15 us pixel >> clock (I assume this is the dwell time, thus the pixel rate is >> 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) >> sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: >> >> >> >> 1) Is it generally true that the AO sampling rate has to be at least >> twice of signal frequency (National Instruments suggested >10X of >> signal rate), which in this case is limited by the amp rate, i.e. >> 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. >> >> >> >> 2) The pixel rate cannot exceed the half of the signal rate (amp >> bandwidth) to catch all spatial varying signal from the sample. What >> is considered as appropriate pixel rate given the amp bandwidth? >> Here, the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. >> Can I assume either they have a typo somewhere, or they compromised >> the measurement by effectively applying a low pass filter to the >> images, or I am just being completely idiotic. >> >> >> >> 3) In general , why single photon counting modules (APD or PMT based) is >> not as popular as analog PMTs in multiphoton fluorescence microscopy? >> Do they not have less constrains for example bandwidth than analog ones? >> >> >> >> I look forward to hearing from you guys. This really bugged me quite >> a few hours. >> >> >> >> Thanks in advance, >> Lu |
|
|
In reply to this post by Craig Brideau
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****
Craig,
This makes so much sense of ‘non-descanned’ and ‘descanned’ detectors. Thank you. Lu From: Confocal Microscopy List [mailto:[hidden email]]
On Behalf Of Craig Brideau ***** To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on
http://www.imgur.com and include the link in your posting. *****
Multi-photon usually uses non-descanned detection, which has the primary dichroic between the scan mirrors and the objective. The fluorescent light from the sample is therefore still moving when it reaches the
detector. As a result it is important that the detector be large so it captures all the light that is moving around. Confocal uses descanned detectors, where the primary dichroic is between the laser and the scan mirrors. In this configuration the fluorescent
light from the sample passes back through the scan mirrors before being collected. This makes the light stationary, which is vital for focusing through a confocal pinhole. Craig On Wed, Oct 26, 2016 at 3:16 PM, Yan, Lu <[hidden email]> wrote:
|
|
|
George McNamara |
Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth
|
|
In reply to this post by Lu Yan
*****
To join, leave or search the confocal microscopy listserv, go to: http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on http://www.imgur.com and include the link in your posting. ***** Hi Lu, Many (maybe most) multiphoton fluorescence microscopes use Coherent Chameleon Ultra or the their Spectra-Physics equivalent. These are run at 80 MHz (80x more pulses per second than the repetition rate you mentioned for the paper you read), so one pulse every 12.5 nanoseconds. Average power (at the peak wavelength) ~3 Watts, each pulse is ~100 femtosecond, so ~3x10^5 W "peak power. My (vague - it's been 3+ years) recollection of the FLIM detectors in the Leica SP5 i managed in Miami was that Becker & Hickl TCSPC electronics were intended to operate at a photon flux of less than or equal to 1 million photons per second (on average, 1 per microsecond), to minimize "pile up" (which reminds me of Houston traffic accidents). I may have this number wrong - Wolfgang often posts on the listserv so you can find his email address from one of those or from Becker&Hickl. Upshot: the paper you read may be using various funky hardware or settings that is not representative of what most MPEF microscope systems have. George p.s. Becker&Hickl also discouraged the use of resonant scanner mode for TCSPC FLIM. The SP5 had 3 standard Leica PMTs and two "FLIM" PMTs. The SP5 was purchased ~2005, so before HyD or HyD-FLIM detectors. It also Leica external NDDs (NDD2 if I recall correctly) standard PMTs, and two APDs on the confocal X1 port on the far end of the scanhead (these ran through SP5 software, to an ISS FCS board). On 10/26/2016 9:00 PM, Yan, Lu wrote: > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > Hi Mike, > > Thanks for your explanation. I understood that if the sampling rate is higher than twice the analog bandwidth (20kHz), because it simply means sampling more than two points per analog pulse. But in the example, as I understood, the sampling rate (set in the DAQ) is >1MHz per channel (obviously fulfill the 'oversampling' condition), but the pixel rate is 66 kHz, in this case, should I think of 'undersampled' or 'oversampled'? > > For the detector, can I interpret the analog bandwidth 20 kHz in this way (in the time domain): > > (1) The detector receives fluorescence photons and generate the photocurrent pulse with a raise time and fall time. The output signal pulse (the first pulse) has long tail (probably longer than 0.05 ms?) so that any other pulses coming in within 1/20kHz=0.05ms time window after the first pulse will not be detected or will be attenuated accordingly. > > (2) Therefore, if the first pulse is coming from a pixel (see, pixel #1), and the laser quickly moves (quicker than 0.05 ms, e.g. 30 us) to pixel #2 where there is no fluorophore. Now, the DAQ channel samples the detector again, and assigns the signal strength to pixel #2. Since the detector has 20 kHz bandwidth, if (1) holds, then would the channel get some value due to the long tail of the first pulse (coming from previous pixel)? If this is true, then the image is not the correct representation of the object, i.e. pixel #2 actually does not have fluorophores but is showed up in the image as if there are some. > > Thanks, > Lu > > > -----Original Message----- > From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Michael Giacomelli > Sent: Wednesday, October 26, 2016 5:40 PM > To: [hidden email] > Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser rep. rate, and detector/amp bandwidth > > ***** > To join, leave or search the confocal microscopy listserv, go to: > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy > Post images on http://www.imgur.com and include the link in your posting. > ***** > > Hi Lu, > > No, if the sampling rate (66 khz) is higher than twice the analog bandwidth (20 khz), then the image is oversampled (at least assuming the optical resolution is much higher than the analog bandwidth limit). However, since the analog bandwidth in this case is probably just the 3dB point, the image will contain some information above 20khz, although it will be attenuated. > > In your example, it depends on how rapidly you are scanning and what your optical resolution is. If you really mean 15 us per point, then your pattern frequency is 1/30us = 33.333Khz, which is a little bit higher than your 3dB point. Many detectors roll off 6dB per factor of two past the 3dB point, so you would probably be attenuated about 7-8dB. If the points you are imaging are bright, you would still see them. > > Mike > > > On Wed, Oct 26, 2016 at 5:10 PM, Yan, Lu <[hidden email]> wrote: >> ***** >> To join, leave or search the confocal microscopy listserv, go to: >> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy >> Post images on http://www.imgur.com and include the link in your posting. >> ***** >> >> Hi Mike, >> >> Thanks for your reply. I am sorry for the mis-calculation, I meant to >> write 0.0667MHz... :J >> >> Anyway, for the second comment you put, would it be 'undersanpled' instead of 'oversampled' if your pixel rate is 66kHz and amp 20 kHz? Supposed that you have a line sample (1x8) with intensity variation pixel by pixel, e.g. something like this: 'o x o x o x o x', where o denotes fluorophore on-site, and x means no signal. If you scan at 66 kHz, then would you be able to catch the second pixel 'x' and probably the third pixel 'o' because your pmt output is only 20 kHz, i.e. more than three times slower than pixel scanning rate? So effectively, will you not probably get an image something like ' o o o x x x o o'? Should the Nyquist pixel rate be 10 kHz? >> >> Thanks, >> Lu >> >> >> -----Original Message----- >> From: Confocal Microscopy List >> [mailto:[hidden email]] On Behalf Of Michael >> Giacomelli >> Sent: Wednesday, October 26, 2016 4:30 PM >> To: [hidden email] >> Subject: Re: Nyquist theorem and DAQ sampling rate, pixel rate, laser >> rep. rate, and detector/amp bandwidth >> >> ***** >> To join, leave or search the confocal microscopy listserv, go to: >> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy >> Post images on http://www.imgur.com and include the link in your posting. >> ***** >> >> 1) You don't have to satisfy Nyquist; you can undersample if you want, but then you are throwing away resolution. >> >> 2) If you are limited by analog bandwidth, then as you scan along the fast axis, the PSF of each pixel is elongated along the scan axis by the low pass filtering in the electronics. 15 us dwell time is actually 66 kHz, so they are slightly oversampled (Nyquist would be 40 kHz), but the bandwidth on amps is usually 3dB, so there are probably frequencies well above 20 kHz. >> >> 3) Single photon counting extracts more signal per photon, but typically can only handle extremely low numbers of photons per second. >> Operated in analog mode, most PMTs are several orders of magnitude >> faster than when in counting mode. If you are not concerned about >> illumination power, usually you get a better image per unit time from >> analog mode. (however, if you are limited by power, counting is much >> better) >> >> Hope that helps. >> Mike >> >> On Wed, Oct 26, 2016 at 3:56 PM, Yan, Lu <[hidden email]> wrote: >>> ***** To join, leave or search the confocal microscopy listserv, go to: >>> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy Post images on >>> http://www.imgur.com and include the link in your posting. ***** >>> >>> Hi listers, >>> >>> >>> >>> I was reading a multiphoton fluorescence imaging paper earlier today, >>> and found in the method section that the authors used 15 us pixel >>> clock (I assume this is the dwell time, thus the pixel rate is >>> 1/15us=~667 kHz?), a 1 MHz pulsed laser, AO (to drive scanning) >>> sampling rate ~1MHz, but a 20 kHz PMT amp. I got a little confused, so my questions are: >>> >>> >>> >>> 1) Is it generally true that the AO sampling rate has to be at least >>> twice of signal frequency (National Instruments suggested >10X of >>> signal rate), which in this case is limited by the amp rate, i.e. >>> 20kHz? In this paper particularly, they are massively oversampling which is OK I guess. >>> >>> >>> >>> 2) The pixel rate cannot exceed the half of the signal rate (amp >>> bandwidth) to catch all spatial varying signal from the sample. What >>> is considered as appropriate pixel rate given the amp bandwidth? >>> Here, the amp bandwidth is only 20 kHz, but the pixel rate is calculated to be 667 kHz.. >>> Can I assume either they have a typo somewhere, or they compromised >>> the measurement by effectively applying a low pass filter to the >>> images, or I am just being completely idiotic. >>> >>> >>> >>> 3) In general , why single photon counting modules (APD or PMT based) is >>> not as popular as analog PMTs in multiphoton fluorescence microscopy? >>> Do they not have less constrains for example bandwidth than analog ones? >>> >>> >>> >>> I look forward to hearing from you guys. This really bugged me quite >>> a few hours. >>> >>> >>> >>> Thanks in advance, >>> Lu -- George McNamara, PhD Houston, TX 77054 [hidden email] https://www.linkedin.com/in/georgemcnamara https://works.bepress.com/gmcnamara/75/ http://www.ncbi.nlm.nih.gov/myncbi/browse/collection/44962650 |
| Free forum by Nabble | Edit this page |