I just received the manual for the unit I'm inspired by.
They used interchangable detectors, but the gratings were factory instal led. They were running from 200 nm to 4.5 microns in five bands depending on in stalled grating, and which detector option you ordered. So basically each grating covered a band. By fortunate co-incidence, the Thorlabs grating has identical specs to the IR module for the band we're interested in.
I've found a US company that makes a Si detector bonded to your choic e of infrared detector, in a TO Can. Samples are 130$ each with web orderin g, otherwise the broad diodes I've found for 500 nm to 1800 nm come from o ur favorite difficult to deal with detector / lamp supplier at the House of the Rising Sun. They must be very fond of the broad range photodiode, b ecause they are not inexpensive.
So one diode choices can cover say 600 to 1800, 330 to 1800 or 330 to 220
- The dual diodes come with a response curve that is diminished but not ext inguished at the overlap which is around 1100 nm.
Only problem with the fast spinning grating is AC powered sources or pulsed /flash sources, but there are techniques to deal with it in software and wi th counter/timers.
I've used the US Digital product in a design at the University, that sc ored me a publication in RSI, thanks to a Prof who thinks all technicians should be first author at least once in their life. I've got another public ation with him, coming out as soon as the reviewer heals from the virus. Th at resulted in two provisional patents related to nanofiber measurement th at are open to licensing. They have a related business that makes interpol ating decoder ICs for optical encoders which I use quite a few of.
I have to see how bad the 1600 nm grating blaze falls off in the visible, b ut with updated detector technology it looks like 5 nm resolution with a si mple encoder and either 500 to 1800 or say 400 to 2200. With a dual set of gratings this could be quite an amazing instrument for next to nothing in h ardware.
The original digitizing option ran on a Apple II of all things, and capture d a high res scan in 1.2 seconds with an asynchronous, free running, motor. It then did the wavelength correction with a lookup table. They had an opt ion for pulsed source profiling by building up an average of scans. Also a software option for linearizing the detector, if needed. So real time disp lay was the scope, and data collection was the Apple.
All the thumb switches did was set either a marker or the starting wavele ngth edge for the scope display, depending on the evolution of the unit. H ence the low chip count in the unit I was using.
For a basic user, correction is not needed, the error is not even noticeabl e for some one on the production floor. Spend an hour with a cal lamp , the n laser print a graticle on a transparency, cut to fit the scope screen and call it a day.
OK, it has some quirks, but the mechanics are easy, the cost to manufacture is low , its flexible, and useful on the bench.
With a modern microcontroller, you could flip in a grating, turn on a calib ration lamp, let the controller find the peaks, run the simple regression, and switch ranges in probably 30 seconds.
Steve