Since most low pressure sodium lamps have been discontinued, I am wondering if an underdriven high pressure sodium lamp would make a good source of monochromatic yellow light that is similar to that of low pressure sodium lamps.

Underdriven HPS lamps have less lumen efficacy than HPS lamps that operates at full output since the arctube is smaller.

If it's only the monochromatic light color of SOX that you are looking to reproduce regardless of efficiency, then yes, underdriving HPS lamps is the way to go. However, you need to go to a really low power level in order to get a reasonably monochromatic light. For a 150 W HPS lamp, you have to drive it at 10 W to get to that point (see SPD graph below, Philips SON-T Plus PIA 150W), above that you start seeing other lines from sodium emerging in the spectrum.

In this regime the light output is 706 lm only, the electrical specs are 33 V and 0.32 A, and the color properties are 1753 K CCT and -41.9 Ra8 CRI.

... and -41.9 Ra8 CRI.

For monochromatic light the true CRI is just plain 0. It can not be negative.
The nehative number you got is what fell out of some simplified equations (part of the spectrum analyser or its SW), but that value is out of their validity range, so means nothing at all.
The programmers just failed to implement there validity range check, so let the SW report a nonsense number instead of the correct "<xx" (displaying it is below the minimum where the used equations are still valid) or "N.A." or something along those lines.

For monochromatic light the true CRI is just plain 0. It can not be negative.

The way the CRI as defined by the CIE is calculated (i.e., the standard color quality parameter used in the lighting industry) it can indeed have a negative value for monochromatic light. So, the average value of -44 Ra8 that is typical for SOX lamps is indeed correct and indicates an extremely poor color rendering quality of its emitted light.

Are those amber LED filament lamps not too far away from monochrome?

The "Amber" LED filament lamps aren't produces an amber colour in reality. They have 2700K LED filaments and have a yellow or gold colouring on the glass so they produces 2000-2500K colour: https://www.lighting-gallery.net/gallery/displayimage.php?pos=-194742

The way the CRI as defined by the CIE is calculated (i.e., the standard color quality parameter used in the lighting industry) it can indeed have a negative value for monochromatic light. So, the average value of -44 Ra8 that is typical for SOX lamps is indeed correct and indicates an extremely poor color rendering quality of its emitted light.

Yes, the equation, the CIE specifies to be used to calculate the CRI, does indeed give a negative CRI for monochromatic light, but at the same time it is there explicitlely stated the CRI is defined as zero for monochromatic lights, without any equations. That equation is just part of the specified process of how to get the CRI value and it is definitely not the only part, nor the first one to be used.
The fact the CRI represents how well the light shows various surface colors remains the fundamental aspect, it was never removed from there, even when the rating of the test colors was moved from subjective evaluation to calculating exact spectra correlations. But it is not a single equation, but a set of equations, each intended for different range, so valid for different range. The actual calculation procedure specifies mathematically how to transition from one range to the other, sometimes this is described as another smooth function. But all that assumes all the calculation are done on a spectrum where it does make sense, assuming e.g. no one would ever attempt to do all the math hassle on a clearly monochromatic light, where the CRI is per definition zero. Yes, automated spectrum analyzer is a dumb computer so it does that calculation blindly, spits some number out, but it does not mean it is correct.
So when a light does not render any olor at all, so has CRI0, there is no way some other light could be ever worse than that.

[...]but at the same time it is there explicitlely stated the CRI is defined as zero for monochromatic lights, without any equations. That equation is just part of the specified process of how to get the CRI value and it is definitely not the only part, nor the first one to be used.[...]

Interesting... I must have missed that in my training. And all the engineers I worked with at Philips must have missed that too, also all the authors of books and papers on lamps and lighting. That's most peculiar.

Do you have a link to that specific part? and to the CIE specification that CRI values must never be negative?

CIE 13.3-1995 gives equations but does not actually say negative results should be rounded to 0 in connection to them. However, it is explicitly states in the summary that the index is not applicable to monocromatic light sources and even specifically mentions low pressure sodium.

So in my opinion, the concept of CRI simply does not apply to monocromatic light.

The "Amber" LED filament lamps aren't produces an amber colour in reality. They have 2700K LED filaments and have a yellow or gold colouring on the glass so they produces 2000-2500K colour: https://www.lighting-gallery.net/gallery/displayimage.php?pos=-194742

Apologies, when I said amber, I meant amber as in the colour, not the decorative style ones, like these
https://www.lighting-gallery.net/gallery/displayimage.php?pos=-173683

Thank you f36t8 for the information about CIE's recommendation. I also checked the training material from Philips and nowhere is it stated that the CRI value should be rounded to zero. In fact, in the Correspondence course #3 (Light and radiation) it is indicated p.33 that "for values below 25, the colour rendering index has no practical meaning". So, if we go by Medved's logic we should also round the Ra8 value of the light emitted by HPS and MV lamps to zero...

In the labs we always kept and used the CRI values as they were calculated. That was also the standard practice of the photometry department. When ceramic metal halide lamps were developed, an important parameter which was checked was the red rendering index, i.e. R9, and it was most often negative. We considered that the more negative this CRI value was, the less red was in the emitted light (i.e. lower saturation in the red). This was also consistent with our comparative tests with color charts and test objects (colored objects and clothing). So, the negative CRI had a practical purpose after all, but its "meaning" is certainly not universal and depends on circumstances and applications.

To return to the topic: So based on that information (and general experience of watching HPS lamps run up), a very large HPS tube like 400/600/1000 W would be needed to get a large light output at low pressures. I wonder if the cathodes would still run hot enough to operate with thermionic emission at these temperatures? If not, the life span might be too short to be useful.

I find the topic intersting, I even asked myself the same question in the past. At the moment I don't have any such lamp that I would like to sacrifice for the experiment.

Thanks guys for the fascinating information. I always wondered why the CRI of LPS was shown as a negative number 

Thank you f36t8 for the information about CIE's recommendation. I also checked the training material from Philips and nowhere is it stated that the CRI value should be rounded to zero. In fact, in the Correspondence course #3 (Light and radiation) it is indicated p.33 that "for values below 25, the colour rendering index has no practical meaning". So, if we go by Medved's logic we should also round the Ra8 value of the light emitted by HPS and MV lamps to zero...

Well, I admit the "CRI=0" is indeed not correct (but just a kind of "computer-form friendly" shortcut), the correct way is indeed the "Not applicable" (or alike).

In the labs we always kept and used the CRI values as they were calculated. That was also the standard practice of the photometry department. When ceramic metal halide lamps were developed, an important parameter which was checked was the red rendering index, i.e. R9, and it was most often negative. We considered that the more negative this CRI value was, the less red was in the emitted light (i.e. lower saturation in the red). This was also consistent with our comparative tests with color charts and test objects (colored objects and clothing). So, the negative CRI had a practical purpose after all, but its "meaning" is certainly not universal and depends on circumstances and applications.

To log what the instrument or calculation spat out is indeed tge correct way to maintain the test documentation, because it allows exact verification (even when it has no meaning, the same result means the inputs were most likely the same if someone repeats the measurement, so proving a random error is very unlikely; non equal result means there was something wrong with some measurement and other results may have been affected, so it must be verified). But it should be then labeled as "what the CRI equation led to" or so. But into the datasheet is should turn as "not applicable" or even better not mentionning the CRI at all.

And for when it has meaning: If it allows to distinguish color samples (e.g. test surfaces from the same group) purely based on color, there is some positive nonzero color rendering. Even when it is below the CRI rating range (so it still can not be expressed as the "CRI" per the newer standard).
But I agree, for practical reason could be handy to compare even the negative numbers, but then you should make sure all involved are interpretting it the same way, as it is not anymore bound by any universally accepted standard (so within an enclosed development team within one organization it may work).



Admin edit to fix broken quotes