Jump to content
UltravioletPhotography

Mystery IR filter - transmission spectra and images


JMC

Recommended Posts

I won't change anything. Jonathon warned us... my fault. :-)

Jonathan, what is different between your spectrometer and Ulf's? I thought you both had the same brand and model, no?

I still don't see why some of the longpass filters show as longpass and others don't. Makes no sense.

 

Same brand, different model and more importantly, different measurement range.

The range depends on what type of internal reflection grating is used.

My spectrometer can see ca 200nm - ca 1000nm with my extended range grating.

With current wavelength calibration it shows values from 202nm to 1024nm.

 

I will explain why we get these confusing results in a later post.

There is a logical explanation to the false results.

I'll be back!

Link to comment

Andy, I don't see this as a malfunction of the device, but more of a limitation which crops up under certain circumstances. What I find worrying, is that that limitation isn't obvious. I looked at the data and thought because it was a nice smooth graph with relatively low noise it was good data. As for why happens sometimes and not others, I think I have an understanding of why it is occurring, but I'm not clear on exactly what is happening within the spectrometer to cause it. A few more tests should hopefully help figure it out.

 

Steve, that Kenko graph really concerns me - it is so similar to what I was seeing. You may not be far from the truth with what you're saying. Perhaps someone there has made the same mistake I have, and not realised its wrong.

Link to comment

Are you seeing the same error on both sides?

If you looks at a germicidal mercury lamp do you get similar error at the 254nm and 185nm lines. Given 185nm may just be too low.

Link to comment
Not tried it David. All I typically seem to get down at the bottom end of the measurement range is noise, but I don't have a germicidal light to test.
Link to comment

To shed som light on this problem (pun intended), i have to go back to basics about array spectrometers and how to measure filter transmission.

 

I hope that what I write below is possible to understand, even if my thoughts are running around in circles when I am writing.

Sorry for any strange phrasing.

Please ask for clarifications, if needed.

 

Background theory

First of all I'd like to show that there is much bouncing around of the light inside an array spectrometer.

Ocean Optics show an example of that here: https://oceanoptics....-bench-options/

 

Much light is reaching the detector's sensor elements, well separated into their different wavelength components, as intended.

However some of it is lost and it is bouncing around freely in the optical system, as stray light.

Some of the stray light eventually ends up hitting the detector.

This is the weakness of this type of spectrometer.

 

As the stray light can be of any wavelength, entering the spectrometer, any detector element is detecting photons from a combination of the intended wavelength for that element, plus whatever kind of stray light present.

Often the intended part overpowers the stray part, but not always.

The stray light becomes more bothering in the longer wavelengths area because the detector's sensitivity is decreasing for longer wavelengths.

Any stray light hitting the detector-elements in the upper end can more easily disturb the measurement.

 

The sensor array used in my spectrometer have this sensitivity curve with less sensitivity for longer wavelengths:

post-150-0-23092200-1569760082.png

 

The final measurement response is a combination of the light source 's spectrum and the spectrometer's optical response.

It is often improved as many light sources have much NIR, making the sensitivity less bad, but the stray-light problem is still present.


Transmission measurements.

 

Calibration

To make a transmission measurement a sequence of actions is done.

 

First the light path is measured, without the filter to be tested.

This is done to define what signal pattern is going to represent 100%.

During the calibration the detector-elements in the upper end are calibrated with a mixture of correctly grating-separated light and any present stray light.

The calibration still looks OK when checked.

 

Measurement

Measuring and getting fooled by a partly flawed calibration: :sad:

 

The raw data from the detector after a measurement with the filter in the light path is then normalised against the sensor-data from the calibration.

Here I omit any baseline calibrations that also is done to simplify the explanation.

 

The sensor data from a long pass filter combined with a wide band light-source can sometimes be flawed with stray light.

When it is measured, that measurement also looks OK, as the upper elements has been hit again by the same stray light as during the calibration.

An example of this can be seen in Jonathan's post #95 above with the B+W 091-filter

 

When a long pass filter with a longer cut-on wavelength is measured, less stray-light is reaching the upper elements.

Then we see the phenomenon Jonathan discovered with his mystery filter and the R72-filter, with an incorrect BP-characteristic.

The upper dip is caused by, that the stray light present during calibration is not there anymore, as the filter cut away the wavelengths causing the stray-light.

All the graphs in Jonathan's post #95 shows more and more false upper drop for filters with a higher cut-on wavelength.

The Heliopan 830 have the deepest drop.

I predict that if Jonathan makes a new measurement of the Heliopan 830 after calibrating with the Heliopan 780 in the light path, the Heliopan 830 will show a long pass curve above 780nm.


My investigation:

 

First a 100%-calibration and verification was done.

I used my light-source, Ocean Optics DH-2000-BAL with just the halogen lamp activated.

Then the output wavelength range is between 400nm and 2500nm.

 

Wide band measurement:

Then I measured Jonathan's R72.

A drop similar to the one Jonathan saw, but starting at 950nm instead is visible.

The difference show that something is strange and likely wrong with the measurements.

 

Narrow band measurement:

I wanted to focus on the transmission above 850nm and added a 850nm long-pass filter (Zomei) to the light path.

After re-calibration and verification of the setup, I measured the R72-filter again.

The measurement result now show a reasonably constant transmission, just as expected for a long pass filter.

Naturally, now the response below 850nm is not valid.

 

post-150-0-14496200-1569762197.png

Cyan graphs reaching ca 96%, measurements of the R72-filter

Red, black and other graphs at 100%, verification measurements taken after each filter measurement.

 

More measurements to see the trends:

I repeated the narrow band measurements with different long pass filters limiting the light source's bandwidth.

I used 680nm and 950nm filters.

I also measured my Zomei 720nm filter. This filter is naturally not as good as the filter from Hoya. It lacks the AR-coating the R72-filter has and the cut on wavelength is more off.

I am really impressed with the performance of the R72 filter. The cut on is at 719.7nm while the Zomei cut on at 711nm. I do not think this si at all visible in the images.

 

Here are all the graphs:

post-150-0-65880000-1569764521.png

The grey graphs shows an intermediate NIR decay, caused by stray light from the wider light source range with the cut on limit at 680nm.

The purple and red graphs reaching 90% are from the Zomei 720nm-filter.

 

I must confess that when just looking at trends I rushed the measurements a bit and the amplitudes are likely to be ±1-2% off.

 

At the very beginning of the session I re-calibrated the wavelength scale too.

The wavelength errors above are well below ±1nm.

Link to comment
Okay, I think I understand the explanation. It sounds like you (both) can just adapt to the limitation and make your spectra in two pieces, one piece using the original setup, the other piece after recalibrating with a known LP filter. Then stitch the two piece together to make the whole spectrum.
Link to comment

Following on from Ulfs excellent explanation, a quick experiment.

 

This was done with my halogen light from the Ocean Optics DH-2000-BAL light source, and my Ocean Optics FX spectrometer. On the plot below are three transmission curves of the Heliopan 830 filter;

 

1. Calibration done with the full spectral range of the halogen light (red line).

2. Calibration done using the Heliopan 715 filter to filter the the halogen light (blue line).

3. Calibration done using the Heliopan 780 filter to filter the halogen bulb (green line).

 

post-148-0-75856600-1569784137.jpg

 

The legends of the graphs also include the collection times. These go from 1200us for the full spectrum lamp up to 6ms when the halogen light was filter using the Heliopan 780 filter (because there is less light getting to the sensor the collection times were increased to compensate for that). All scans were an average of 50 runs, and with a boxcar smoothing of 1.

 

As the incoming light is progressively filtered during the calibration phase with the Heliopan 715 or 780, the effect of the stray light becomes less and less, and the position where the Heliopan 830 filter transmission starts to drop off again moves to longer wavelengths.

 

Filtering the incoming light did not eliminate the problem but did reduce it. This still gives me an upper limit of around 820nm, above which I cannot trust the data.

Link to comment

That makes perfect sense to me. Excellent thinking through and testing.

Ok your going to need a second spectrometer optimized for 600 to 1100nm.

When I was considering buying a spectrometer, that was my original plan. One from 200nm to 800nm and a second from 600nm to 1200nm. Then use the over lap to blend. An IR optimized sensor for IR measurements would be best. A Foveon sensor would be very IR sensitive.

 

I stopped this plan as I am happy with the pretty images and less accurate 1000 lines/mm grating. I get a rough idea and I don't need exact numbers.

Link to comment

Thank You for testing my idea Jonathan!

 

It troubles me that you didn't reach all the way.

There might be some other stones to turn over to find yet another bug causing this if we are lucky.

 

However the dip might also be due to your spectrometers configuration.

 

OO have a few standard configurations and two fit what I think you have:

OCEAN FX-UV-VIS and OCEAN FX-UV-VIS -ES

Both these have specified functional range of 200-850nm

 

Can you please tell us if you have one of these or if it is a custom configurations-version?

Link to comment

Jonathan, if you have a 850nm filter it might improve the measurements range a bit further.

 

With my 850nm filter I got usable light for a test from ca 825nm and upwards.

Even before reaching the stated cut-on wavelength the transmission of the filter is high enough.

 

I am amazed of the sensitivity and speed of the CMOS-detector in your FX.

With full range halogen light my Flame needs 20ms per sweep.

Limited with a 850nm filter it needs 58ms.

 

I have the halogen lamp slightly dimmed to extend its life, but not to a very low level.

Link to comment
  • 1 year later...

Holy thread resurrection Batman.

 

Some data to finally put this one to bed (if it wasn't already put to bed). I borrowed an Ocean Insight (formerly Ocean Optics) STS-NIR miniature spectrometer, as I've been thinking of expanding my transmission measurement capabilities up to 1100nm. By the way, this thing is tiny - 42x40x24mm, here it is next to a UK 2p coin.

post-148-0-36413500-1616509893.jpg

 

This little spectrometer measures between 650nm and 1100nm and is a good match to my FX unit (250nm to 800nm).

 

Been doing a few test measurements with it, and here are the scans for Hoya R72, Heliopan 715, 780, 830 and 1000 IR filters, all now behaving as expected out to 1100nm.

post-148-0-32079300-1616587481.jpg

 

The plan is to use this to look at lens transmission too, and with the light and spectrometers I have access to, that means I should be able to measure from 280nm to 1100nm. More to come on that later.

 

EDIT 24th March 2021 - removed R72 data from the graph as it does not correspond to the known cut off data for the filter.

Link to comment
I have been following this thread with some curiosity, as I did a bit of spectroscopy myself, long ago. So the problem is that there is stray light in the reference beam that is not present in the sample beam? I struggle a bit to understand that--instrument design seeks to minimize issues such as that.
Link to comment

So there is still a twist in this story. The filter I have in front of me says "Hoya 52mm Infrared [R72] Japan", and I've just rerun it with both spectrometers, and both give the cutoff as show in Post #114 from earlier today. Which as Stefano pointed out is too low for that filter. I've also run a set of band pass filters from 650nm to 900nm and they all behave as expected.

 

So it is not a spectrometer issue.

 

Looking back over this thread, I got another R72 checked it and sent it to Ulf so we could compare our results. This is where it gets really weird. My original R72 and new one I bought to test and send to Ulf do not have the same cutoffs - the one I bought to send to Ulf cut off at a higher wavelength. Not sure why I never saw this before, I think I was too focused on what was happening at the other end of the spectrum to notice it.

 

I'll get some graphs sorted out to help explain all this, but all very odd.

 

Anyway, despite all that, they are not bandpass filters (which was the original reason I started this thread - which I'm beginning to wish I hadn't now).

Link to comment
Andy Perrin

Anyway, despite all that, they are not bandpass filters (which was the original reason I started this thread - which I'm beginning to wish I hadn't now).

Hah! More of a can of worms, right?

Link to comment

Nah just label it fake and be done with it.

At least you have a nice new IR spectrometer.

My Jaz spectrometer is funny. I bought it with two detectors. One with grating #2 and one with grating #4. The #4 gratings has a lp filter and was to be used for fluorescent detection.

 

Then when testing a xenon pulse module the one with grating #2 stopped working, it was still under warranty so returned for fix. He had one with grating #1, which for me was perfect as now can see the 185nm Mercury line. So I have that module but its 4x more sensitive than the one originally for fluorescence with #4 grating. So haven't used the IR #4 grating module yet for original purpose. But it is better to eliminate stray light.

 

Have fun looking into the IR world.

 

I now like to scan stuff with the light I intend to use to get idea for the signal variation.

Link to comment

Can of worms indeed.

 

My 'R72' came from someone who used to do forensics work, and there was no mention of if not being what it said it was, and the filter ring is an R72 one. Looking at the spectra it looks to be something like an 89B, but that is just speculation. I have bought myself another R72 to test - hopefully this one will be a good one.

 

Does make me wonder how many people don't have what they think they have?

 

For now I'll update the graph I shared yesterday, to remove the suspect filter.

Link to comment

" Does make me wonder how many people don't have what they think they have? "

 

Filters, Lenses & lights....that is why I bought a spectrometer... :cool:

Link to comment

" Does make me wonder how many people don't have what they think they have? "

 

Filters, Lenses & lights....that is why I bought a spectrometer... :cool:

 

Oddly you almost need one.

The advantage of buying a spectrometer that I see is after purchasing it everything I have purchased is what it says it is. To make you feel like you don't need it. But if you don't have one than its all questionable.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...