I think this question has been addressed incorrectly IMHO.
Changing the digital gain does not do anything to the recorded signal
except scale it up -noise and all. (IMHO, this a useless feature and I
have never ever used it on _any _ confocal I've used). Now the pmt is
intrinsically a *very* linear device. This was well established in the
1960's by extensive tests by EMI. The linearity of pmts was essential
for the precise stellar measurements being made by astronomers.
Non-linear behavior may be introduced by poor application design, in
particular dynode chain design leaving to dynode voltage droop or local
overload of electrodes (cathode, dynodes &/or anode) and exceeding
manufacturer limits (which will burn the coatings). In fact, the pmt
itself is generally more linear than the current-to-voltage amplifier
and A/D converter -provided you do not exceed design limits.
A few numbers may help make this clear. If you want < 0.1% non-linearity
total tube current cannot exceed 0.1% of dynode chain current which is
typically 1mA (it may be somewhat less than this as gain is only linear
on a log log plot only and is different for each tube). This implies a
maximum acceptable tube current of ~1uA for our application design
point. Assuming a maximum design tube gain of 10^7 (actual tubes can go
higher than this), this gives a detected photon flux = 0.1pA which is
about 6 x 10^5 photoelectrons per second. As you reduce operating
voltage, the tube gain goes down and the tube current falls in parallel.
So if the manufacturers have correctly set the A/D converter system gain
up so that the detector is linear to 0.1% at maximum gain then this will
NOT change with pmt voltage (unless you can exceed the maximum design
voltage/gain for the system). This also means that regardless of the
scene brightness, the maximum non-linearity is 0.1% provided no pixels
are saturated. At a dwell time of 1us per pixel this maximum gain
implies only 0.6 counts per pixel (/us) for this design example (I
suggest there is NO point scaling this value to 4096...).
The bottom line is that detector linearity will not change with pmt
votage if you always have the same A/D converter gain. Post
multiplication of the A/D value does nothing to the scene information.
In dim images where high gain is used, 8 bit recording is perfectly
adequate ('cos less than 100 photoelectrons per pixel) and gives smaller
file sizes. If the scene has a dynamic range of 4000 peak average photon
count per pixel would then have to be >1000 (from almost no saturated
pixels). This represents a huge photon flux at a 1us dwell time but
might be achieved with slower scans. (It is less clear whether it would
be better to scan fast and average).
If there is serious non-linearity in the detector of the Olympus, it can
only be due to bad electrical design (and it needs to be fixed by
redesign) and changing pmt voltage will not alter it. If the light is
too concentrated on the photocathode so that local saturation occurs
then that would need fixing too. I suspect that neither of these
problems are actually present and the original poster is in error. All
3 confocal systems (non-Olympus) I have seen inside have had non-zener
controlled k-d1 dynode steps implying they are being used far from
currents where dynode droop (and linearity) becomes a problem.
My advice is to ignore the digital gain and set pmt voltage to give the
desired gain -and remember to set the black level correctly or your
image really will be non-linear.
My 20c
Regards Mark Cannell
Julian Smith III wrote:
> Thanks for the very clear explanation of the difference!
> How does one determine where one's PMT's become non-linear?
> Julian
>
> Florian Eich wrote:
>> Dear Listers,
>> to clarify the HV vs. Gain on FluoView FV1000 topic and to follow
>> Stanislav Vithas request for someone from Olympus to answer:
>>
>> 1. Adjust HV of the PMT. This regulates how many counts are generated
>> by a photon hitting the PMT. When you have a weak sample with a good
>> S/N you can raise the HV to a value where the brightest Point on your
>> image generates 4095 counts, therefore using the complete dynamic
>> range. (Of course also with a strong sample it is desirable to use
>> the complete dynamic range for your sample, it simple generates a
>> better data-quality).
>> But be aware: If you raise the HV too much you reach the non-linear
>> range of the PMT, leading to an unequal amplification of areas of
>> different brightness.
>> 2. The Gain is an alternative way to facilitate the full dynamic
>> range of the AD-
>> Converter. To give an example: By using only the HV your sample has
>> intensity values from 0-1000 counts, rising the HV to top end would
>> take you into the nonlinear range of the PMT. If this signal is
>> converted by the ADC, you do not use the whole dynamic range of 4096.
>> If you add a Gain of e.g. 4, you get everything multiplied by 4 – the
>> PMT noise as well as the actual analog signal. The PMT noise will be
>> higher, but now you can use a much larger dynamic range – up to 4095
>> without sacrificing the linearity of the image. Another advantage is
>> that the quantization noise will be smaller in comparison to the signal.
>> Plainly, if you have a weak sample that can not be imaged with the
>> full dynamic range in the linear range using the HV, you can use
>> Gain to amplify the image and to keep the HV in the linear range. If
>> you don´t care about using the full dynamic range, there is no need
>> to use the Gain.
>>
>> Best regards
>> Florian Eich
>> Olympus Life Science Europa GmbH
>>
>>
>
>
|