Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
On Sat, 17 May 2025 12:30:38 -0700, Don Yx3ZbRTwkboNHcYYefd34pvzEKKty4RSFiiA4v0BSw_gbiEQH-khaK5lIXJ36q2q2xqW39hJj34hYp1MPTT9w4wb0RRE01F52nClp8J-VhECWQ18IWoopERU1Pl8khD8T_UPIBnauk.iFb6dsfIy8kEJvdCzNVyv8buyH2ji-Budd1i9iTh3IE&dib_tag=se&keywords=spectrometer+handheld&qid=1747511016&sprefix=
<blockedofcourse@foo.invalid> wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
The people who make spectrometer-type instruments seem to be in a
battle for ever finer resolution.
I want a spectrometer that spans 400 to 1600 nm, or at least 800 to
1600. I want to know if a 1310 nm laser is about 1310 and not by
accident 1550 or something.
I was thinking about making such an instrument. A few filters and a
few photodiodes might work, with some overlap for interpolation.
A rotating, graded filter and one wideband detector could work.
Or a grating and a couple of detectors, with software to resolve
ambiguities.
Maybe just three detectors with different wavelength peaks.
We did buy a couple of fiber WDM splitters, which can, for instance,
tell us if a laser is 880 or 1300 or some such.
Are there toy-level visual spectrometers?
https://www.amazon.com/EISCO-Premium-Quantitative-Spectroscope-Accuracy/dp/B00B84DGDA/ref=sr_1_3?crid=21PO5QTTGGA06&dib=eyJ2IjoiMSJ9.IwA9B16820dPfj5ct0JEivvGqDD0YV5wFHFcG9c1Xss1BCoKEJvHFm_dYkhhHHK8lICo1KuioeQ85usmShFPtgMSSa0gzI2E-_
Cool. I just ordered one.
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
On Sat, 17 May 2025 12:30:38 -0700, Don Yx3ZbRTwkboNHcYYefd34pvzEKKty4RSFiiA4v0BSw_gbiEQH-khaK5lIXJ36q2q2xqW39hJj34hYp1MPTT9w4wb0RRE01F52nClp8J-VhECWQ18IWoopERU1Pl8khD8T_UPIBnauk.iFb6dsfIy8kEJvdCzNVyv8buyH2ji-Budd1i9iTh3IE&dib_tag=se&keywords=spectrometer+handheld&qid=1747511016&sprefix=
<blockedofcourse@foo.invalid> wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
The people who make spectrometer-type instruments seem to be in a
battle for ever finer resolution.
I want a spectrometer that spans 400 to 1600 nm, or at least 800 to
1600. I want to know if a 1310 nm laser is about 1310 and not by
accident 1550 or something.
I was thinking about making such an instrument. A few filters and a
few photodiodes might work, with some overlap for interpolation.
A rotating, graded filter and one wideband detector could work.
Or a grating and a couple of detectors, with software to resolve
ambiguities.
Maybe just three detectors with different wavelength peaks.
We did buy a couple of fiber WDM splitters, which can, for instance,
tell us if a laser is 880 or 1300 or some such.
Are there toy-level visual spectrometers?
https://www.amazon.com/EISCO-Premium-Quantitative-Spectroscope-Accuracy/dp/B00B84DGDA/ref=sr_1_3?crid=21PO5QTTGGA06&dib=eyJ2IjoiMSJ9.IwA9B16820dPfj5ct0JEivvGqDD0YV5wFHFcG9c1Xss1BCoKEJvHFm_dYkhhHHK8lICo1KuioeQ85usmShFPtgMSSa0gzI2E-_
Cool. I just ordered one.
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Spinning or oscillating prism?
On 17/05/2025 20:30, Don Y wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Short answer is you can't - at least without making some *very* questionable assumptions. It is even worse now with narrowband LEDs.
If you are allowed to make the assumption of a radiant perfect black body (something that doesn't exist) then it is much easier.
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
True colorimeters were designed to match visible colours pretty much exactly under *any* lighting conditions (extremely tough problem). The first that actually worked well enough was the Imperial Match Predictor which ISTR was an
analogue computer made in the UK by ICI strictly for internal use only. I don't
think any documentation survives.
There was a US made spectrometer which formed a part of it whose manufacturers
name escapes me for the moment. Got it Hardy Spectrophotometer:
https://collection.sciencemuseumgroup.org.uk/objects/co11842/ge-hardy-spectrophotometer-c-1940
That model isn't quite the right one but it is close.
Now any suitable paint test chart and a mobile phone will do the job.
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
If you are serious about doing this right then a 2D CCD sensor and a prism hires grating combo at right angles will allow you to quantify the entire visible spectrum at ultra high resolution. Be careful though Perkin-Elmer (and
others) have some very good lock out patents on this trick (may be about to expire).
A few people can see longer wavelengths than most with an extra type of cone cell. They were sought after in WWII (pre thermal IR band imaging) because they
could see the difference between live foliage still growing and cut down dying
foliage used as gun emplacement camouflage.
Denatured chlorophyll looks much darker to them.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
Measure the intensity at all wavelengths in a single shot.
PE OES instrument in the early 1990's was the first with this.
(I forget the model number) I was seriously impressed with it.
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Spinning or oscillating prism?
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
On 5/17/2025 2:03 PM, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a
prism hires grating combo at right angles will allow you to quantify
the entire visible spectrum at ultra high resolution. Be careful
though Perkin-Elmer (and others) have some very good lock out patents
on this trick (may be about to expire).
Again, not looking to make an "instrument". The phone idea may work
if the CCDs don't freak out with high intensity sources.
If you are serious about doing this right then a 2D CCD sensor and a
prism hires grating combo at right angles will allow you to quantify the entire visible spectrum at ultra high resolution.
On 5/17/2025 2:03 PM, Martin Brown wrote:
On 17/05/2025 20:30, Don Y wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Short answer is you can't - at least without making some *very* questionable >> assumptions. It is even worse now with narrowband LEDs.
If you are allowed to make the assumption of a radiant perfect black body
(something that doesn't exist) then it is much easier.
I'm not looking for a laboratory grade instrument. (hence the
"corner cutting" caveat).
Rather, "how does the light falling on THIS body compare to the
light on this OTHER body" (using the same measuring instrument)
I've worked with true colorimeters (dual wavelength) in the past.
But, they were optimized to look for specific wavelengths.
True colorimeters were designed to match visible colours pretty much exactly >> under *any* lighting conditions (extremely tough problem). The first that
actually worked well enough was the Imperial Match Predictor which ISTR was an
analogue computer made in the UK by ICI strictly for internal use only. I don't
think any documentation survives.
Ours controlled the color temperature of an incandescent lamp
"seen" through a pair of filters. Then, compared the detected
signal from the sample under test (inserted between the emitter
and detector) in the same short time interval, looking for a
particular color shift (analyzing blood assays)
Again, you don't care WHAT "color" it is, just how the chemistry
altered the color within a band of expected results.
But, that system KNEW what to expect (expectations were dependent
on the actual assay being run)
There was a US made spectrometer which formed a part of it whose manufacturers
name escapes me for the moment. Got it Hardy Spectrophotometer:
https://collection.sciencemuseumgroup.org.uk/objects/co11842/ge-hardy-spectrophotometer-c-1940
That model isn't quite the right one but it is close.
Now any suitable paint test chart and a mobile phone will do the job.
How durable are the CCDs used in phones? Especially to high intensity light sources?
I calibrate the light emitted by my monitors with a device,
but it controls the light source to do so.
If you are serious about doing this right then a 2D CCD sensor and a prism >> hires grating combo at right angles will allow you to quantify the entire
visible spectrum at ultra high resolution. Be careful though Perkin-Elmer (and
others) have some very good lock out patents on this trick (may be about to >> expire).
Again, not looking to make an "instrument". The phone idea may work
if the CCDs don't freak out with high intensity sources.
A few people can see longer wavelengths than most with an extra type of cone >> cell. They were sought after in WWII (pre thermal IR band imaging) because they
could see the difference between live foliage still growing and cut down dying
foliage used as gun emplacement camouflage.
Also folks who are truly colorblind. Camouflage looks different than
natural foliage when you are just looking at the values without the
hues to distract.
Denatured chlorophyll looks much darker to them.
With no knowledge of the actual (visible) spectrum impinging on
a sensor (and a bit of time to integrate results), how can I
do this short of swapping individual filters in front of the
sensor(s)?
Measure the intensity at all wavelengths in a single shot.
Or, leverage the fact that the spectrum won't be changing in
the short term (for some value of "short") and cycle a set
of filters (rotating disc?) between the detector and source.
Again, if you aren't looking for repeatability instrument to
instrument, this may be good enough to answer the question above.
PE OES instrument in the early 1990's was the first with this.
(I forget the model number) I was seriously impressed with it.
On 18/05/2025 03:29, Don Y wrote:
On 5/17/2025 2:03 PM, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a prism >>> hires grating combo at right angles will allow you to quantify the entire >>> visible spectrum at ultra high resolution. Be careful though Perkin-Elmer >>> (and others) have some very good lock out patents on this trick (may be
about to expire).
Again, not looking to make an "instrument". The phone idea may work
if the CCDs don't freak out with high intensity sources.
CCDs are almost indestructible unless you point them at the sun. Even then they
handle it much better than a human eye. Webcams are probably a lot cheaper though. If you find one of the paint firm's colour matching apps and test chart
it may already do what you want or close enough.
If you just want to color match then your phone camera is dandy. There are apps used by printers and film lighting cameramen to do just that. ISTR chromlink ?
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a prism >> hires grating combo at right angles will allow you to quantify the entire
visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some had weird
formats like 1024x64 pixels so I assume they were really made for spectroscopy
On 5/18/2025 5:37 AM, Martin Brown wrote:
On 18/05/2025 03:29, Don Y wrote:
On 5/17/2025 2:03 PM, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a
prism hires grating combo at right angles will allow you to quantify
the entire visible spectrum at ultra high resolution. Be careful
though Perkin-Elmer (and others) have some very good lock out
patents on this trick (may be about to expire).
Again, not looking to make an "instrument". The phone idea may work
if the CCDs don't freak out with high intensity sources.
CCDs are almost indestructible unless you point them at the sun. Even
then they handle it much better than a human eye. Webcams are probably
a lot cheaper though. If you find one of the paint firm's colour
matching apps and test chart it may already do what you want or close
enough.
The way it was described to me (how does the light falling on this body compare
to the light on some other body) suggests it was expected to receive
radiant
light directly (not reflected light of of two bodies that have different reflectance characteristics)
But, I don't know how intense the light would be.
I was asked because of my past experience with the colorimeter shining
light
directly onto the detector, through filters. (I've seen products that can tell you what color an object is, etc., using reflectance)
On 5/17/2025 2:03 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Spinning or oscillating prism?
That might be better than a varied filter. But, probably require finer control (or sensing) of its current orientation.
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a
prism hires grating combo at right angles will allow you to quantify
the entire visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some
had weird formats like 1024x64 pixels so I assume they were really
made for spectroscopy
As mentioned elsewhere, how do they fare when light is shining directly
on the
sensor? How do you keep it from saturating -- dark lens to attenuate
the signal?
The way it was described to me (how does the light falling on this body compare
to the light on some other body) suggests it was expected to receive radiant >> light directly (not reflected light of of two bodies that have different
reflectance characteristics)
But, I don't know how intense the light would be.
As ever the devil is always in the details. Identical colours but with different surface finishes can look incredibly different. Vantablack is very much like looking into the void it is quite literally blacker than black!
Any other "black" looks grey next to it.
I was asked because of my past experience with the colorimeter shining light >> directly onto the detector, through filters. (I've seen products that can >> tell you what color an object is, etc., using reflectance)
You can trick almost any sensor. Human eye can be quite easily misled by didymium glass which is a narrowband Na-D blocking filter used to see into a bright yellow sodium flame when glassblowing.
Side effect is to produce cartoon like out of gamut colours when the brain tries to compute colours from the cones. Its apparent colour varies radically with the source of illumination.
The same property is shared with the natural gemstone Alexandrite.
https://en.wikipedia.org/wiki/Chrysoberyl
Such materials are rare and highly prized for their strange behaviour.
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a
prism hires grating combo at right angles will allow you to quantify
the entire visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some
had weird formats like 1024x64 pixels so I assume they were really
made for spectroscopy
As mentioned elsewhere, how do they fare when light is shining directly
on the
sensor? How do you keep it from saturating -- dark lens to attenuate
the signal?
Don Y <blockedofcourse@foo.invalid> wrote:
On 5/17/2025 2:03 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
Spinning or oscillating prism?
That might be better than a varied filter. But, probably require finer
control (or sensing) of its current orientation.
If it is spinning steadily, all you need is a synchronising pulse at
some point once per revolution and a wide spectrum photocell with an
optical slit and a lens. Software can work out the wavelength from the rotational speed and the known characteristics of the prism. The
resolution can be as coarse or as fine as you like and algorithms can
work out the visual perception of line spectra (if that is what you
need).
The same hardware could be used for an expensive high-resolution device
or a cheap and cheerful version - the software and the time to reach a
steady reading (longer integration period for lower 'noise') being the
only real differences.
On 5/18/2025 2:15 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
On 5/17/2025 2:03 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution >>>> and accuracy?
Spinning or oscillating prism?
That might be better than a varied filter. But, probably require finer
control (or sensing) of its current orientation.
If it is spinning steadily, all you need is a synchronising pulse at
some point once per revolution and a wide spectrum photocell with an optical slit and a lens. Software can work out the wavelength from the rotational speed and the known characteristics of the prism. The
Of course. But, if spinning faster than your integration interval,
I suspect any jitter in your angular resolution might be difficult
to factor out of the mix.
This would, instead, suggest a slower rotation so the prism feeds
the detector a single wavelength for a longer (continuous) period.
That means the time to get a sampling of the spectrum is multiplied
by the integration interval. If, instead, you could get "quick peeks"
at each wavelength "quickly", and the more precise integration "later",
you have more data to work with, sooner.
[This is the approach I have historically taken with data acquisition
as it lets me trade response time for resolution, dynamically]
The same hardware could be used for an expensive high-resolution device
or a cheap and cheerful version - the software and the time to reach a
"cheerful"?
On 5/18/2025 2:15 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
On 5/17/2025 2:03 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution >>>>> and accuracy?
Spinning or oscillating prism?
That might be better than a varied filter. But, probably require finer >>> control (or sensing) of its current orientation.
If it is spinning steadily, all you need is a synchronising pulse at
some point once per revolution and a wide spectrum photocell with an
optical slit and a lens. Software can work out the wavelength from the
rotational speed and the known characteristics of the prism. The
Of course. But, if spinning faster than your integration interval,
I suspect any jitter in your angular resolution might be difficult
to factor out of the mix.
This would, instead, suggest a slower rotation so the prism feeds
the detector a single wavelength for a longer (continuous) period.
That means the time to get a sampling of the spectrum is multiplied
by the integration interval. If, instead, you could get "quick peeks"
at each wavelength "quickly", and the more precise integration "later",
you have more data to work with, sooner.
[This is the approach I have historically taken with data acquisition
as it lets me trade response time for resolution, dynamically]
resolution can be as coarse or as fine as you like and algorithms can
work out the visual perception of line spectra (if that is what you
need).
The same hardware could be used for an expensive high-resolution device
or a cheap and cheerful version - the software and the time to reach a
"cheerful"?
steady reading (longer integration period for lower 'noise') being the
only real differences.
That might be better than a varied filter. But, probably require finer >>>> control (or sensing) of its current orientation.
If it is spinning steadily, all you need is a synchronising pulse at
some point once per revolution and a wide spectrum photocell with an
optical slit and a lens. Software can work out the wavelength from the
rotational speed and the known characteristics of the prism. The
Of course. But, if spinning faster than your integration interval,
I suspect any jitter in your angular resolution might be difficult
to factor out of the mix.
Mount the prism on a a flywheel and spin it rapidly. The only jitter
might come from errors in the timing pulse (or knackered bearings!).
One way of obtaining a jitter-free timing pulse would be to reflect a
known pattern of light off the faces of the prism into the photocell;
use the software to recognise it and make corrections for any long-term
speed drift.
This would, instead, suggest a slower rotation so the prism feeds
the detector a single wavelength for a longer (continuous) period.
That means the time to get a sampling of the spectrum is multiplied
by the integration interval. If, instead, you could get "quick peeks"
at each wavelength "quickly", and the more precise integration "later",
you have more data to work with, sooner.
If it spins faster you can simply integrate multiple 'passes' for as
long as you want until the noise is negligible. The frequency response
of the photocell and head amplifier is likely to be far wider than any mechanical system needs, so the physical narrowness of the slit and the distance from the prism will set the resolution limit. .A narrow and
distant slit will give higher resolution at the expense of a worse S/N
ratio, which can be overcome with a longer integration time.
[This is the approach I have historically taken with data acquisition
as it lets me trade response time for resolution, dynamically]
Yes, it has many advantages.
The same hardware could be used for an expensive high-resolution device
or a cheap and cheerful version - the software and the time to reach a
"cheerful"?
"Cheap and cheerful" is a slang [UK English] expression meaning a quick rough estimate or goods that aren't intended for serious long-term use.
Sampling jitter within a window corresponds to spectral resolution;
the more jitter, the wider the range of wavelengths potentially
involved in the sample (over time). As sampling the detector
is a discrete time event (the interval between samples being the
width of the window), how frequently you do this further defines
the spectral resolution.
"Cheap and cheerful" is a slang [UK English] expression meaning a quick rough estimate or goods that aren't intended for serious long-term use.
"Quick and dirty", "spit and baling wire", "bubble gum and shoe strings", "good enough for government work", "mickey-mouse", "jury-rigged", "jerry-built", etc.
On 20/05/2025 18:43, Don Y wrote:
On 5/18/2025 2:15 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
On 5/17/2025 2:03 PM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution >>>>> and accuracy?
Spinning or oscillating prism?
That might be better than a varied filter. But, probably require finer >>> control (or sensing) of its current orientation.
If it is spinning steadily, all you need is a synchronising pulse at
some point once per revolution and a wide spectrum photocell with an
optical slit and a lens. Software can work out the wavelength from the >> rotational speed and the known characteristics of the prism. The
Of course. But, if spinning faster than your integration interval,
I suspect any jitter in your angular resolution might be difficult
to factor out of the mix.
This would, instead, suggest a slower rotation so the prism feeds
the detector a single wavelength for a longer (continuous) period.
I really wouldn't consider anything with moving parts. You can get
reasonable grade replica grating for low resolution spectroscopy from
the likes of Edmund scientific (intended for school labs).
https://www.edmundoptics.co.uk/c/gratings/621/#27766=27766_s%3AClear%2BPol yester%2BFilm
Or if you aren't too fussy about quality the photo filters sold to put rainbow stars on disco lights or on eBay. Astronomy magazines often have adverts for slightly better than average gratings for eyepieces.
That means the time to get a sampling of the spectrum is multiplied
by the integration interval. If, instead, you could get "quick peeks"
at each wavelength "quickly", and the more precise integration "later",
you have more data to work with, sooner.
[This is the approach I have historically taken with data acquisition
as it lets me trade response time for resolution, dynamically]
resolution can be as coarse or as fine as you like and algorithms can
work out the visual perception of line spectra (if that is what you
need).
The same hardware could be used for an expensive high-resolution device
or a cheap and cheerful version - the software and the time to reach a
"cheerful"?
UK alliterative saying.
I guess it doesn't translate into USAian too well.
steady reading (longer integration period for lower 'noise') being the
only real differences.
Figure out how much resolution you need before starting out.
A shovelware DVD at glancing incidence can resolve the absorption lines
in the suns spectrum if you do it just right. You look down onto the
disk with the sun at a very shallow angle to the surface. Don't look at
the reflection of the sun - only at the very dispersed spectrum.
The spectrum obtained with this simple kit is impressively high
resolution. It will also have various funny organic dye lines in with a modern writeable one.
You really want aluminised media for this trick.
Don Y <blockedofcourse@foo.invalid> wrote:
[...]
Sampling jitter within a window corresponds to spectral resolution;
the more jitter, the wider the range of wavelengths potentially
involved in the sample (over time). As sampling the detector
is a discrete time event (the interval between samples being the
width of the window), how frequently you do this further defines
the spectral resolution.
I was assuming very fast sampling so that the presentation of each line
was captured by many samples, that way the software could sort it out
over a large number of repeated passes. Keep the hardware simple and
let the software deal with the errors if it can be given enough data to
start with.
Less developed software, lower sampling rate, slower ADC and less memory
in the cheaper version. (And a garish box with "Professional" on it, to
let customers know that this is the cheap and nasty version.)
[...]
"Cheap and cheerful" is a slang [UK English] expression meaning a quick >>> rough estimate or goods that aren't intended for serious long-term use.
"Quick and dirty", "spit and baling wire", "bubble gum and shoe strings",
"good enough for government work", "mickey-mouse", "jury-rigged",
"jerry-built", etc.
That's the sort of thing. An interesting demonstration toy, rather than
a laboratory instrument.
Don Y <blockedofcourse@foo.invalid> wrote:
Sampling jitter within a window corresponds to spectral resolution;
the more jitter, the wider the range of wavelengths potentially
involved in the sample (over time). As sampling the detector
is a discrete time event (the interval between samples being the
width of the window), how frequently you do this further defines
the spectral resolution.
I was assuming very fast sampling so that the presentation of each line
was captured by many samples, that way the software could sort it out
over a large number of repeated passes. Keep the hardware simple and
let the software deal with the errors if it can be given enough data to
start with.
Less developed software, lower sampling rate, slower ADC and less memory
in the cheaper version. (And a garish box with "Professional" on it, to
let customers know that this is the cheap and nasty version.)
[...]
"Cheap and cheerful" is a slang [UK English] expression meaning a quick >>> rough estimate or goods that aren't intended for serious long-term use.
"Quick and dirty", "spit and baling wire", "bubble gum and shoe strings",
"good enough for government work", "mickey-mouse", "jury-rigged",
"jerry-built", etc.
That's the sort of thing. An interesting demonstration toy, rather than
a laboratory instrument.
Don Y <blockedofcourse@foo.invalid> wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
How broad and how much resolution?
There are sensors, eg: https://ams-osram.com/products/sensor-solutions/ambient-light-color-spectral-proximity-sensors
versions of which can be found in cheap dev boards: https://shop.pimoroni.com/products/as7343-breakout?variant=41694602526803
I'm sure I remember reading recently of a consumer grade multispectral
camera part with a moderate resolution (something like 8x8 or 32x32) but I can't find a reference to it now. But it seems there's a phone launching with such a camera soon (according to rumours): https://www.gizmochina.com/2025/05/13/huawei-nova-14-series-to-launch-in-may-with-harmonyos-5-and-an-ultra-model-specs-here/
On 5/18/2025 4:22 AM, Theo wrote:If the light levels are very high then you can get LEDs with emission
Don Y <blockedofcourse@foo.invalid> wrote:
Not quite, but, close enough...
How can I determine the spectrum of incident light on a sensor,
in general? Then, how many corners can I cut to sacrifice resolution
and accuracy?
How broad and how much resolution?
I've not been read into those details. It's unclear if they are *known*
or being withheld (as they know I'm not interested in any more work; why disclose information that you don't have to?)
There are sensors, eg:
https://ams-osram.com/products/sensor-solutions/ambient-light-color-spectral-proximity-sensors
Thanks, I will pass that along.
versions of which can be found in cheap dev boards:
https://shop.pimoroni.com/products/as7343-breakout?variant=41694602526803
I'm sure I remember reading recently of a consumer grade multispectral
camera part with a moderate resolution (something like 8x8 or 32x32)
but I
can't find a reference to it now. But it seems there's a phone launching >> with such a camera soon (according to rumours):
https://www.gizmochina.com/2025/05/13/huawei-nova-14-series-to-launch-in-may-with-harmonyos-5-and-an-ultra-model-specs-here/
That would target a different application (IMO). As presented to me, they're
just looking for characterizing the light falling on a *spot*. If
different (e.g., bandwidth sensitive) detectors were employed AT that
spot, I would assume they would have to be treated as a single point
(despite there being some obvious separations involved in their
manufacture)
I.e., almost like a photographer's "light meter" but with the interest
being on the spectral content and not the overall intensity.
[Whether this is true or not, it has influenced how *I* have thought
about the problem -- in terms of function, size, portability, power requirements, etc. Assumptions are always the bane of a good design...
:< ]
On 5/21/2025 6:19 AM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
Sampling jitter within a window corresponds to spectral resolution;
the more jitter, the wider the range of wavelengths potentially
involved in the sample (over time). As sampling the detector
is a discrete time event (the interval between samples being the
width of the window), how frequently you do this further defines
the spectral resolution.
I was assuming very fast sampling so that the presentation of each line
was captured by many samples, that way the software could sort it out
over a large number of repeated passes. Keep the hardware simple and
let the software deal with the errors if it can be given enough data to start with.
Are you expecting to frequently sample the entire spectrum in each
"pass" ("revolution")? Or, walk the sampling window up/down the spectrum
in stages?
I.e., how much time are you expecting to spend PROCESSING the sampled
data vs. acquiring more data?
E.g., as presented to me, there was no need for calibration against
a reference standard, "flat" response across the spectrum, etc.
A "laboratory grade" device likely WOULD impose such specifications.
And, bear an associated (likely high) cost.
Don Y <blockedofcourse@foo.invalid> wrote:
On 5/21/2025 6:19 AM, Liz Tuddenham wrote:
Don Y <blockedofcourse@foo.invalid> wrote:
Sampling jitter within a window corresponds to spectral resolution;
the more jitter, the wider the range of wavelengths potentially
involved in the sample (over time). As sampling the detector
is a discrete time event (the interval between samples being the
width of the window), how frequently you do this further defines
the spectral resolution.
I was assuming very fast sampling so that the presentation of each line
was captured by many samples, that way the software could sort it out
over a large number of repeated passes. Keep the hardware simple and
let the software deal with the errors if it can be given enough data to
start with.
Are you expecting to frequently sample the entire spectrum in each
"pass" ("revolution")? Or, walk the sampling window up/down the spectrum
in stages?
I was expecting to sweep the whole spectrum at high speed many times,
then analyse the captured data. Television-type technology could easily
cope with that data rate from a single photocell.
A small gas-filled discharge tube, pulsed by an ignition transformer,
would suffice for non-critical calibration.
I was assuming very fast sampling so that the presentation of each line
was captured by many samples, that way the software could sort it out
over a large number of repeated passes. Keep the hardware simple and
let the software deal with the errors if it can be given enough data to
start with.
Are you expecting to frequently sample the entire spectrum in each
"pass" ("revolution")? Or, walk the sampling window up/down the spectrum
in stages?
I was expecting to sweep the whole spectrum at high speed many times,
then analyse the captured data. Television-type technology could easily
cope with that data rate from a single photocell.
I.e., how much time are you expecting to spend PROCESSING the sampled
data vs. acquiring more data?
The ratio can be varied by either the user or the designer of the
instrument. If greater accuracy is required, it will take longer to do
both the capture and the analysis.
[...]
E.g., as presented to me, there was no need for calibration against
a reference standard, "flat" response across the spectrum, etc.
A "laboratory grade" device likely WOULD impose such specifications.
And, bear an associated (likely high) cost.
Some sort of reference source could be used to generate a known spectrum every 'n' passes; this would also serve for synchronising purposes.
There would be no need to accurately control the rotational speed as
long as it was steady in the short term. The reference spectrum would calibrate the span and the end points; it could also calibrate the
spectral amplitude response of the photo-detector.
A small gas-filled discharge tube, pulsed by an ignition transformer,
would suffice for non-critical calibration.
On 22/05/2025 18:34, Don Y wrote:
I.e., almost like a photographer's "light meter" but with the interestIf the light levels are very high then you can get LEDs with emission profiles
being on the spectral content and not the overall intensity.
[Whether this is true or not, it has influenced how *I* have thought
about the problem -- in terms of function, size, portability, power
requirements, etc. Assumptions are always the bane of a good design... :< ]
of about 50nm width. They work as sensors in the opposite direction with some leakage for higher energy photons. Choose them wisely and calibrate against a reference white and you might have something that is both cheap accurate and durable.
On 18/05/2025 20:45, Don Y wrote:
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a prism >>>> hires grating combo at right angles will allow you to quantify the entire >>>> visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some had >>> weird formats like 1024x64 pixels so I assume they were really made for
spectroscopy
As mentioned elsewhere, how do they fare when light is shining directly on the
sensor? How do you keep it from saturating -- dark lens to attenuate the >> signal?
You vary the exposure to avoid spillover.
On 5/18/25 21:45, Don Y wrote:
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and a prism >>>> hires grating combo at right angles will allow you to quantify the entire >>>> visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some had >>> weird formats like 1024x64 pixels so I assume they were really made for
spectroscopy
As mentioned elsewhere, how do they fare when light is shining directly on the
sensor? How do you keep it from saturating -- dark lens to attenuate the >> signal?
or a shutter to limit the time light hits the sensor
On 5/19/2025 1:33 PM, Lasse Langwadt wrote:
On 5/18/25 21:45, Don Y wrote:
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and
a prism hires grating combo at right angles will allow you to
quantify the entire visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence, some
had weird formats like 1024x64 pixels so I assume they were really
made for spectroscopy
As mentioned elsewhere, how do they fare when light is shining
directly on the
sensor? How do you keep it from saturating -- dark lens to attenuate
the signal?
or a shutter to limit the time light hits the sensor
There's still a limit as to the peak intensity that a sensor can
tolerate. And, gating the light (instead of attenuating it)
means there is no signal when the source is gated off.
Fine if you are making a device with a button that says "measure now".
But, if you expect to be able to collect data at any time, you
want to be sure data is available.
As mentioned elsewhere, how do they fare when light is shining directly on the
sensor? How do you keep it from saturating -- dark lens to attenuate the >>>> signal?
or a shutter to limit the time light hits the sensor
There's still a limit as to the peak intensity that a sensor can
tolerate. And, gating the light (instead of attenuating it)
means there is no signal when the source is gated off.
Simplest solution is to limit the aperture that you let the light in through which is normally focussed onto a narrow slit anyway. You might get a bit of an
issue with readout smearing but it probably won't be too bad.
Please bear in mind that my experience with spectroscopy the problem was mostly
getting enough light to have *any* signal to noise.
Fine if you are making a device with a button that says "measure now".
But, if you expect to be able to collect data at any time, you
want to be sure data is available.
In extremis the measure now button could just move a spring loaded mechanical shutter that normally blocks the light path.
Unless the thing is imaging a nuclear blast, steel furnace or an arc lamp then
I don't think light intensity is likely to harm a modern CCD. There are hot mirror and anti-UV low pass filters to protect such equipment from hostile radiation.
On 24/05/2025 21:07, Don Y wrote:
On 5/19/2025 1:33 PM, Lasse Langwadt wrote:
On 5/18/25 21:45, Don Y wrote:
On 5/18/2025 6:13 AM, Lasse Langwadt wrote:
On 5/17/25 23:03, Martin Brown wrote:
If you are serious about doing this right then a 2D CCD sensor and >>>>>> a prism hires grating combo at right angles will allow you to
quantify the entire visible spectrum at ultra high resolution.
use a CD https://youtu.be/EoAZ-u6hn6g?si=Mv-DfJ5swtq2-j1X&t=98 :)
eons ago we used some CCDs as detectors for X-ray fluorescence,
some had weird formats like 1024x64 pixels so I assume they were
really made for spectroscopy
As mentioned elsewhere, how do they fare when light is shining
directly on the
sensor? How do you keep it from saturating -- dark lens to
attenuate the signal?
or a shutter to limit the time light hits the sensor
There's still a limit as to the peak intensity that a sensor can
tolerate. And, gating the light (instead of attenuating it)
means there is no signal when the source is gated off.
Simplest solution is to limit the aperture that you let the light in
through which is normally focussed onto a narrow slit anyway. You might
get a bit of an issue with readout smearing but it probably won't be too
bad.
Please bear in mind that my experience with spectroscopy the problem was mostly getting enough light to have *any* signal to noise.
Fine if you are making a device with a button that says "measure now".
But, if you expect to be able to collect data at any time, you
want to be sure data is available.
In extremis the measure now button could just move a spring loaded
mechanical shutter that normally blocks the light path.
Unless the thing is imaging a nuclear blast, steel furnace or an arc
lamp then I don't think light intensity is likely to harm a modern CCD.
There are hot mirror and anti-UV low pass filters to protect such
equipment from hostile radiation.
| Sysop: | Keyop |
|---|---|
| Location: | Huddersfield, West Yorkshire, UK |
| Users: | 546 |
| Nodes: | 16 (2 / 14) |
| Uptime: | 09:59:01 |
| Calls: | 10,389 |
| Calls today: | 4 |
| Files: | 14,061 |
| Messages: | 6,416,850 |
| Posted today: | 1 |