pyPhotometry - New fluorescent detectors avalable at Doric

[félix Leroy]

Hi,

I am assembling a fiber-photometry setup to perform two-color recordings.

I previously used the complete Doric system with Newport Visible Femtowatt Photoreceiver Module detectors and was not satisfied with them for two reasons:

- high-gain (10^11) mode is only available in AC mode

- they do not couple directly to the filter cube therefore requiring an extra coupling step which induces light losses

Doric is now offering two new types of detectors than can couple directly to the filter cube:

- their own Doric Fluorescence Detector

- the Hamamatsu Photosensor Module H10722-20

The Hamamatsu seems more sensitive but is more expensive and requires an external power supply. The Doric detector specifications looks similar to the Newport one except that you can record in DC mode at high gain (2*10^11) and they couple directly to the filer cube.

My question is whether the Hamamatsu and Doric detector would work with the pyPhotometry system?

best regards,

félix

[Thomas Akam]

Hi Félix,

The two things that will determine whether a photodetector is compatible with pyPhotometry are the voltage range of its output signal and its bandwidth (the range of signal frequencies it passes). 

The pyPhotometry board can read signals between 0-3.3V.  There is a clamp diode which stops the input voltage going above 3.3V so it will not be damaged by signals going a bit higher, but signals going well below 0V might damage the board.  The output voltage range of the Doric Fluorescence detector is -5V to +5V, however I think that the voltage will only go negative in AC mode, as in DC mode 0V presumably corresponds to 0 light intensity.  The Hamamatsu Photosensor module's max output voltage is 4V.   The minimum output voltage is not specified on the datasheet. It seems likely that 0V corresponds to 0 light intensity, in which case the output voltage would not go negative, but it would be worth checking this before buying.  In summary I think that the output voltage range on both would be compatible but it would be worth checking that 0 volts indeed corresponds to 0 light intensity.

Regarding the bandwidth, the Doric fluorescence detectors stated bandwidth is 0 - 1KHz in DC mode which should be fine as the Newport Photorecieveres bandwidth is only 0-750Hz.  It would be worth asking Doric for a trace showing the response of the output in DC mode to a short light pulse to check how long it takes the signal to return to zero (as in Fig. 3b in the pyPhotometry manuscript), as this will determine the maximum frequency you can acquire without crosstalk in time division illumination mode.  The Hamamatsu photosensor has a much wider bandwidth (0-20KHz), and the datasheet shows the transisent response is ~40us compared to about 2.5ms for the Newport 2151, so you should be able to use time division illumination at much higher frequencies, perhaps ~1KHz compared with 130Hz using the Newport or Doric detector.

Regarding the signal quality you will get with these different devices, the noise level is important in addition to the gain - having a higher gain will not necessarily give you better signal to noise if there is also higher noise.  When I was quantifying the noise level for sinusoidal illumination and lock in amplification for figure 3F of the manuscript I did not find that switching to high gain AC mode made much difference as the signal was larger but the noise was also.  The Newport 2151 has a noise equivalent power (NEP) of 12 fW/sqrt(Hz) in DC mode which corresponds to an output noise voltage of 4mVrms over the 0-750Hz bandwidth.  The Doric detector states an NEP of 12W/sqrt(Hz) but this has to be a typo as it is 10^15 times larger than the Newport detector.  Assuming that Doric meant that the NEP was 12 fW/sqrt(Hz) this would be the same as the Newport.  It would be useful to check with Doric what the NEP and the noise RMS voltage is at different gain settings.

best,

Thomas

[félix Leroy]

Hi Thomas,

thanks for looking into that. I have asked all your questions to Doric.

If I understood you right the pyPhotometry board will only work in DC mode as it can only read positive values. Sending -5 V into it by trying to record maximum negative values in AC could actually be damaging for the board. Also the intensity of the response will clamp above 3.3V so we have to be careful to not push the excitation to much or we will saturate our response.

Doric gave me price comparisons for 2 detectors plus a filter cube:

$6500 USD with 2 Doric photodetectors

$6500 USD with 2 Hamamatsu PMTs (+$1950 USD of power supply unit for each PMT)

about $4500 USD with 2 Newport photodetectors

best regards,

félix

[félix Leroy]

Hi,

here are the answers from Doric.

-The output voltage range of the Doric Fluorescence detector is -5V to +5V. Does this mean that the voltage will only go negative in AC mode and that in DC mode 0V corresponds to 0 light intensity?

That is correct. In DC mode, the voltage will be between 0 and 5 V. In AC mode, it can go between –5 and 5 V.

 

- Do you have a trace of the response of the output in DC mode to a short light pulse.

Not yet, but it is a good thing to know and we will do the measurement this week (probably tomorrow or Wednesday). We will send you the results once we have it.

 

- Finally, the Doric detector states a noise equivalent power (NEP) of 12W/sqrt(Hz) but this has to be a typo as it is 10^15 times larger than the Newport detector.

Indeed! It should be 12 fW/sqrt(Hz)! Thank you very much for noticing. We will update all our documentation.

 

- Did you check what the NEP and the noise RMS voltage is at different gain settings?

Unfortunately not for the NEP. We should make another run of more detailed tests using high-end spectrum analyzer in the following weeks to get the absolute values. For the RMS, by comparing the noise from our electronics to the Newport at all gain for small signals, we always find a decrease of at least 30 % of the noise using our photodetector.

[Thomas Akam]

Hi Félix,

Many thanks for posting those answers.  Based on Doric's replies it looks like their fluorescence detector would work with pyPhotometry. We would need to see the output response to a short light pulse to know the maximum frequency you could acquire in time division illumination modes, but is is likely to be as good as or slightly better than the Newport based on the Bandwidth they report in the specs.  It sounds like the noise level of the Doric detector is slightly better and you would presumably get some performance gain from having the detector directly coupled to the minicube rather than via a fiber.

I have been trying to work out what the signal to noise performance of the Hamamatsu PMTs would be relative to e.g. the Newport.  The data sheet reports various different senstivity measures which I do not fully understand, however the only measure that has the same units as the gain measurements on the other detectors (V/W) is the '[Anode] Radiant Sensitivity Typ.' which is reported at 150V/nW which is 1.5e11 V/W - so comparable to the AC high gain mode on the Newport.  The only noise measurement reported is 'Ripple Noise (peak to peak) Max.' which is 0.5mV - much smaller than the ~4mVrms noise on the Newport, despite the much wider bandwidth of the Hamamatsu.  Based on this I think that the Hamamatsu's are likely to have much better signal to noise performance than either the Newport or the Doric, and has the further advantage of much higher bandwidth, allowing much faster sampling rates with time division illumination.

Looking at the figure below from the datasheet for the Hamamatsu PMT, it seems that it can be controlled using just a +/- 5V power supply and a variable resistor, which would be a lot cheaper than the official power supply.

best,

Thomas

[félix Leroy]

Hi,

Doric replied to me.

- Do you have a trace of the response of the output in DC mode to a short light pulse.

Attached are the results of the comparison for a 5 ms light pulse as measured by the Newport and the Doric photodetector. Both were set at the same gain, and in DC mode. The decay time in our case is about 0.8 ms, whereas it is around 1.3 for the Newport when accounting for the undershoot.

So, in conclusion the Doric one goes back to baseline faster than the Newport one.

[Azahara Oliva]

Hi Thomas,

We have the system running and can get some signals but would like to know what do you think. Sometimes we get strong peaks that we don't know yet how to interpret (don't seem to correlate with muscle artifacts though...). You can see it in figure1 (2 seconds of recording). We are monitoring electrical signals at the same time and do not find any correlation with those peaks. Any idea?

thanks,

aza

[Thomas Akam]

Hi Azahara,

Those look like electrical noise not a physiological signal to me.  The physiological signals should be smooth and relatively slow, as the decay time of the fluorescence is several hundred milliseconds minimum.

T

[félix Leroy]

Hi Thomas,

We repeated our testing on 3 mice co-injected with GCaMP6f and mCherry.

When in 2-color continuous acquisition mode the green and red channels follow variation of the blue light intensity. However the red channel doesn't change when we vary the green light intensity. I do see green light coming out the fiber tip already (but when the blue is present it masks the green one).

We set up the blue light intensity in order to get the green signal mid-range in this mode (Doric photodetector are set to DC mode and the gain is at x100, otherwise it saturate).

Attached is an example of our recording in 2-color continuous acquisition mode with the entire session and a close up view.

best regards

félix

[Azahara Oliva]

Hi Thomas, 

Here is a closer view of the recording in two colors continuous mode.

And here an example recording in 2 colors time division mode in the same animal. In this channel the green channel remain stable more or less.

 

thanks,

aza

[Thomas Akam]

Hi Felix and Aza,

To understand what is going on in your data it is useful to think about the multiple possible sources that may be generating the fluorescent signal.  These are:

1.  Fluorescence from the flurophores you are expressing in the tissue (GCaMP and mCherry).

2.  Autofluorescence from normal brain tissue.

3.  Autuofluorescence from the optic fiber between the minicube and the sample.

Our experience is that the 465nm LED excites substantial autofluorescence from the optic fiber in both the green and the red channels, whereas the 560nm LED excites minimal autofluorescence from the optic fiber.  You can test this by pointing the optic fiber that normally goes to the subject at a black light absorbent target and varying the power of the 465 and 560nm LEDs.  It is worth doing this in both continous mode and time division mode to get a good understanding of what is going on.  On our setups, varying the 465nm LED power in continous acquisition mode causes large signal changes for both the green and red channels, due to autofluorscence excited from the fiber.  Varying the 465nm LED power in time division mode causes changes in the signal level for the green channel, but not for the red channel, because the red channel samples are being read when the 465nm LED is off.  In either mode we see little variation in either channel when we change the 560nm LED power because this led does not excite much autofluorescence from the fiber.  Of course, if the fiber is pointed at a fluorescent sample which absorbs at 560nm and emits in the red channel (e.g. mCherry or TdTomato), changes in the 560nm LED power will drive large signal changes in the red chanel.

Given that you see essentially zero signal in the red channel when you are in time division mode, this suggests that there is minimal mCherry expression within range of the fiber (I assume you have the 460nm LED on max power - what light power do you have for each color at the fiber tip).  Your observations about how the signal changes when you vary the LEDs powers could be consistent with your signal primarily being autofluorescence from the fibers.  It is really hard to know what is causing those large slow shifts you see in the green channel - they do not look very physiological though I don't know what neuronal population you are trying to record from.  They could be movement artifacts or due to the rotary joint if you are using one, though you would normally not expect  ~20% signal changes.  Have you done any histology yet to verify your fiber placements and virus experssion?

best,

Thomas