> On Monday, June 2, 2014 9:14:51 AM UTC-7, Phil Hobbs wrote:
>> On 06/02/2014 10:49 AM, George Herold wrote:
>>
>>> On Sunday, June 1, 2014 2:43:28 PM UTC-4, Phil Hobbs wrote:
>
>>>> One good method for that is to use a multimode fibre bundle on the
>>>> Rx side--you spread one end out into a line to form the exit slit,
>>>> and bunch the other end into a disc to match the photodiode shape.
>>
>>>> Then you focus it down onto the PD to reduce the required area.
>
>>> Yeah, I think you mentioned this trick. Do you loose much light
>>> getting it into the fibers?
>
>> You lose about 4% at each face to Fresnel reflection, and the fill
>> factor for 200/240 um fibres is about 0.8.
>
>>>> You can typically reduce the PD area by a factor of 100 to 1000
>>>> compared with a square PD the size of the exit slit.
>
> I've seen this recommendation before (and it makes sense) but I've also
> seen long, thin photosensors (the CCD in most flatbed scanners). Why
> doesn't someone make slit-shape photodiodes?
You can get them made if you have the volume, or can afford the NRE. Of
course surface leakage would tend to go like the perimeter rather than
the area, but in big enough quantity that would be cheaper than a fibre
bundle.
>
> Light enters low-index media most effectively. Does that mean a plastic light fiber
> would be preferred? Cladding has high refrective index, would reflect
> much of the illumination.
POF is generally much larger than glass, but it would work OK over the
spectral range where it's transparent.
>
> One recommendation was to use epoxy to hold the fibers in position,
> which would be a light-directing nightmare (can one dismiss the light leaking
> into the epoxy?). Come to that, why not just use an acrylic lightpipe
> with a slot-to-spot shape?
Basically because it modifies the angular spectrum of the light in an
undesirable way. Rearranging fibres gives you a pretty well circular
light cone coming out.
>
> The use of index-matching cement at the photodiode was not raised
> (perhaps because that contaminates the photodiode chemically?).
You can coat the fibre ends if you care about the 4%.
>
> Finally, if one wants an efficient coupling of diode to slit, what's wrong with
> using a V-assembly with a mirror opposite the photodiode? The
> multiple reflections in such a structure would result in most of the
> slit's light hitting the diode, and any reflected light would get a few
> more passes at the diode on its way out.
The light emerging from the slit is fairly far from collimated, though.
Using fibres allows you to slice and dice your etendue (area-solid
angle product) in a fine-grained way that is a big win in this application.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Reply by whit3rd●June 2, 20142014-06-02
On Monday, June 2, 2014 9:14:51 AM UTC-7, Phil Hobbs wrote:
> On 06/02/2014 10:49 AM, George Herold wrote:
>
> > On Sunday, June 1, 2014 2:43:28 PM UTC-4, Phil Hobbs wrote:
> >> One good method for that is to use a multimode fibre bundle on the
> >> Rx side--you spread one end out into a line to form the exit slit,
> >> and bunch the other end into a disc to match the photodiode shape.
>
> >> Then you focus it down onto the PD to reduce the required area.
> > Yeah, I think you mentioned this trick. Do you loose much light
> > getting it into the fibers?
> You lose about 4% at each face to Fresnel reflection, and the fill
> factor for 200/240 um fibres is about 0.8.
> >> You can typically reduce the PD area by a factor of 100 to 1000
> >> compared with a square PD the size of the exit slit.
I've seen this recommendation before (and it makes sense) but I've also
seen long, thin photosensors (the CCD in most flatbed scanners). Why
doesn't someone make slit-shape photodiodes?
Light enters low-index media most effectively. Does that mean a plastic light fiber
would be preferred? Cladding has high refrective index, would reflect
much of the illumination.
One recommendation was to use epoxy to hold the fibers in position,
which would be a light-directing nightmare (can one dismiss the light leaking
into the epoxy?). Come to that, why not just use an acrylic lightpipe
with a slot-to-spot shape?
The use of index-matching cement at the photodiode was not raised
(perhaps because that contaminates the photodiode chemically?).
Finally, if one wants an efficient coupling of diode to slit, what's wrong with
using a V-assembly with a mirror opposite the photodiode? The
multiple reflections in such a structure would result in most of the
slit's light hitting the diode, and any reflected light would get a few
more passes at the diode on its way out.
Reply by Phil Hobbs●June 2, 20142014-06-02
On 06/02/2014 10:49 AM, George Herold wrote:
> On Sunday, June 1, 2014 2:43:28 PM UTC-4, Phil Hobbs wrote:
>> On 6/1/2014 11:45 AM, George Herold wrote:
>>
>>> I might have tried eating the PD capacitance and running it at
>>> zero bias...
>>
>>
>>
>> One good method for that is to use a multimode fibre bundle on the
>> Rx side--you spread one end out into a line to form the exit slit,
>> and bunch the other end into a disc to match the photodiode shape.
>> Then you focus it down onto the PD to reduce the required area.
>>
> Yeah, I think you mentioned this trick. Do you loose much light
> getting it into the fibers? (I'll guess ~ 10% at each face, and then
> some other filling factor due to the round fibers and long square
> slit... I guess you have to match the fiber to the slit width
> too...)
You lose about 4% at each face to Fresnel reflection, and the fill
factor for 200/240 um fibres is about 0.8. (You don't care about the
cladding on the slit edges, and the little bit of scalloping of the
effective slit area doesn't reduce the resolution much.)
>
> And what do you gain in a low light situation? (I guess the smaller
> PD means less reverse bias leakage... so you can do the measurements
> faster and average a bit.
Right. 100x less area means 100x less crap in the data, and more than
that if you're in the omega*eN*Cd limit.
>> If the spectrometer runs at f/4 (0.125 NA), which is common, you
>> can match that with catalogue fibres. Then you can save a factor
>> of 16 in detector area just by going to 0.5 NA on the detector, and
>> another
>>
>> (maybe bigger) factor due to illuminating the whole detector area.
>>
>> You can typically reduce the PD area by a factor of 100 to 1000
>> compared with a square PD the size of the exit slit.
>>
>> (I did a SWIR spectrometer last year that did that--worked great
>> except for some X-axis instability problems that I'm working on
>> now.)
>
> Hmm, lets see Z axis is the light direction, and I'll guess y is up
> and down, (Along the long slit axis) so x is the short slit
> dimension?
Yes. If you have a wavelength uncertainty of 0.01 nm and a 1 AU/100 nm
slope, you get 1e-4 AU of absorption uncertainty.
>
> You'll want accurate angular alignment of your fiber bundle and slit,
> as well as getting it in the right x-position.
That isn't hard at all at f/4 with multimode fibre. The fibre bundle
_is_ the exit slit, so you get that for free, and eyeball accuracy is
better than good enough for the angle.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Reply by George Herold●June 2, 20142014-06-02
On Sunday, June 1, 2014 9:38:53 PM UTC-4, whit3rd wrote:
> On Sunday, June 1, 2014 8:45:12 AM UTC-7, George Herold wrote:
>
> > On Saturday, May 31, 2014 8:28:46 PM UTC-4, whit3rd wrote:
>
>
>
> > > One, in particular, was for fluorescence spectroscopy on microscopic
>
> > > samples. It had a ~1 cm photodiode to cover the monochromator's exit
>
> > > slit, and took a minute or so to scan the spectrum with a stepping motor
>
> > > moving the grating. The output was picoamps. So was leakage. Capacitance
> > > was high, seconds-to-minutes response was what worked best
> > > with a chart recorder.
>
> > Ouch, pico amps with a PD. That's hard. I assume the wavelenghts were
> > too long for a PMT? (Maybe a job for a APD?)
> >
> > I might have tried eating the PD capacitance and running it at zero bias...
> Good guess on wavelength; we did eventually borrow a red-sensitive PMT, and chopper
> for the illuminator, and phase-lock amp - the results were ... excellent.
> Two days later, we gave those all back, though.
>
> It was running at zero bias (we thought the signal cable, a plain old RG-58 BNC ,
> would have been sensitive otherwise).
Oh, a PD on the end of coax, only physicists are allowed to do that :^)
George H.
>An off-the-shelf picoammeter was the
>
> preamp we used, and it probably had low bandwidth (10 Hz?) which helped the noise situation.
> The diode(and cable) capacitance just didn't hurt us at all at the
> scan speeds we were using.
Reply by George Herold●June 2, 20142014-06-02
On Sunday, June 1, 2014 2:43:28 PM UTC-4, Phil Hobbs wrote:
> On 6/1/2014 11:45 AM, George Herold wrote:
>
> > I might have tried eating the PD capacitance and running it at zero bias...
>
>
>
> One good method for that is to use a multimode fibre bundle on the Rx
> side--you spread one end out into a line to form the exit slit, and
> bunch the other end into a disc to match the photodiode shape. Then you
> focus it down onto the PD to reduce the required area.
>
Yeah, I think you mentioned this trick. Do you loose much light getting it into the fibers? (I'll guess ~ 10% at each face, and then some other filling factor
due to the round fibers and long square slit... I guess you have to match the fiber to the slit width too...)
And what do you gain in a low light situation?
(I guess the smaller PD means less reverse bias leakage... so you can do
the measurements faster and average a bit.
> If the spectrometer runs at f/4 (0.125 NA), which is common, you can
> match that with catalogue fibres. Then you can save a factor of 16 in
> detector area just by going to 0.5 NA on the detector, and another
>
> (maybe bigger) factor due to illuminating the whole detector area.
>
> You can typically reduce the PD area by a factor of 100 to 1000 compared
> with a square PD the size of the exit slit.
>
> (I did a SWIR spectrometer last year that did that--worked great except
> for some X-axis instability problems that I'm working on now.)
Hmm, lets see Z axis is the light direction, and I'll guess y is up and down,
(Along the long slit axis)
so x is the short slit dimension?
You'll want accurate angular alignment of your fiber bundle and slit, as well as getting it in the right x-position.
George H.
>
>
> Cheers
>
>
>
> Phil Hobbs
>
>
>
> --
>
> Dr Philip C D Hobbs
>
> Principal Consultant
>
> ElectroOptical Innovations LLC
>
> Optics, Electro-optics, Photonics, Analog Electronics
>
>
>
> 160 North State Road #203
>
> Briarcliff Manor NY 10510
>
>
>
> hobbs at electrooptical dot net
>
> http://electrooptical.net
Reply by whit3rd●June 1, 20142014-06-01
On Sunday, June 1, 2014 8:45:12 AM UTC-7, George Herold wrote:
> On Saturday, May 31, 2014 8:28:46 PM UTC-4, whit3rd wrote:
> > One, in particular, was for fluorescence spectroscopy on microscopic
> > samples. It had a ~1 cm photodiode to cover the monochromator's exit
> > slit, and took a minute or so to scan the spectrum with a stepping motor
> > moving the grating. The output was picoamps. So was leakage. Capacitance
> > was high, seconds-to-minutes response was what worked best
> > with a chart recorder.
> Ouch, pico amps with a PD. That's hard. I assume the wavelenghts were
> too long for a PMT? (Maybe a job for a APD?)
>
> I might have tried eating the PD capacitance and running it at zero bias...
Good guess on wavelength; we did eventually borrow a red-sensitive PMT, and chopper
for the illuminator, and phase-lock amp - the results were ... excellent.
Two days later, we gave those all back, though.
It was running at zero bias (we thought the signal cable, a plain old RG-58 BNC ,
would have been sensitive otherwise). An off-the-shelf picoammeter was the
preamp we used, and it probably had low bandwidth (10 Hz?) which helped the noise situation.
The diode(and cable) capacitance just didn't hurt us at all at the
scan speeds we were using.
Reply by Phil Hobbs●June 1, 20142014-06-01
On 6/1/2014 11:45 AM, George Herold wrote:
> I might have tried eating the PD capacitance and running it at zero bias...
One good method for that is to use a multimode fibre bundle on the Rx
side--you spread one end out into a line to form the exit slit, and
bunch the other end into a disc to match the photodiode shape. Then you
focus it down onto the PD to reduce the required area.
If the spectrometer runs at f/4 (0.125 NA), which is common, you can
match that with catalogue fibres. Then you can save a factor of 16 in
detector area just by going to 0.5 NA on the detector, and another
(maybe bigger) factor due to illuminating the whole detector area.
You can typically reduce the PD area by a factor of 100 to 1000 compared
with a square PD the size of the exit slit.
(I did a SWIR spectrometer last year that did that--worked great except
for some X-axis instability problems that I'm working on now.)
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Reply by George Herold●June 1, 20142014-06-01
On Saturday, May 31, 2014 8:28:46 PM UTC-4, whit3rd wrote:
> On Saturday, May 31, 2014 4:41:21 PM UTC-7, Phil Hobbs wrote:
>
<snip>
> One, in particular, was for fluorescence spectroscopy on microscopic
> samples. It had a ~1 cm photodiode to cover the monochromator's exit
> slit, and took a minute or so to scan the spectrum with a stepping motor
> moving the grating. The output was picoamps. So was leakage. Capacitance
> was high, seconds-to-minutes response was what worked best
> with a chart recorder.
Ouch, pico amps with a PD. That's hard. I assume the wavelenghts were
too long for a PMT? (Maybe a job for a APD?)
I might have tried eating the PD capacitance and running it at zero bias...
at least there is no leakage current then.
George H.
>
>
>
> It mismatches by a few orders of magnitude the numbers you've plugged in.
Reply by Phil Hobbs●May 31, 20142014-05-31
On 5/31/2014 8:28 PM, whit3rd wrote:
> On Saturday, May 31, 2014 4:41:21 PM UTC-7, Phil Hobbs wrote:
>> On 5/31/2014 7:21 PM, whit3rd wrote:
>
> [about error current in a photodiode]
>>> dC/dt * Vbias + C * d/dt(Vbias)
>
>> There's no error term in the external circuit from dC/dt.
>
> Of course there is! The charge on the capacitor doesn't
> move through the depletion region, it goes in the external wiring.
So if you have an unconnected photodiode, what happens when its
temperature changes? You still have a depletion region, and its width
still changes with temperature.
>
>> The temperature of the photodiode changes on a timescale of minutes to
>> hours. If you have a capacitance of 10 pF at 25 C...
>> The current error resulting from your mechanism would be
>
>> delta I = (dT/dt)*dC/dT*V_bias
>
>> = (175 K)/(30s) * (1 pF/100 K) * 10 V = 0.6 pA.
>
>> How many sensitive photoreceivers have you actually designed?
>
> One, in particular, was for fluorescence spectroscopy on microscopic
> samples. It had a ~1 cm photodiode to cover the monochromator's exit
> slit, and took a minute or so to scan the spectrum with a stepping motor
> moving the grating. The output was picoamps. So was leakage. Capacitance
> was high, seconds-to-minutes response was what worked best
> with a chart recorder.
>
> It mismatches by a few orders of magnitude the numbers you've plugged in.
Do tell.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
Reply by whit3rd●May 31, 20142014-05-31
On Saturday, May 31, 2014 4:41:21 PM UTC-7, Phil Hobbs wrote:
> On 5/31/2014 7:21 PM, whit3rd wrote:
[about error current in a photodiode]
> > dC/dt * Vbias + C * d/dt(Vbias)
> There's no error term in the external circuit from dC/dt.
Of course there is! The charge on the capacitor doesn't
move through the depletion region, it goes in the external wiring.
> The temperature of the photodiode changes on a timescale of minutes to
> hours. If you have a capacitance of 10 pF at 25 C...
> The current error resulting from your mechanism would be
> delta I = (dT/dt)*dC/dT*V_bias
> = (175 K)/(30s) * (1 pF/100 K) * 10 V = 0.6 pA.
> How many sensitive photoreceivers have you actually designed?
One, in particular, was for fluorescence spectroscopy on microscopic
samples. It had a ~1 cm photodiode to cover the monochromator's exit
slit, and took a minute or so to scan the spectrum with a stepping motor
moving the grating. The output was picoamps. So was leakage. Capacitance
was high, seconds-to-minutes response was what worked best
with a chart recorder.
It mismatches by a few orders of magnitude the numbers you've plugged in.