Conversation

Notices

1. Embed this notice
   Thomas 🔭🕹️ (thomasfuchs@hachyderm.io)'s status on Monday, 24-Nov-2025 11:14:45 JST Thomas 🔭🕹️

   in reply to

   • Graham Sutherland / Polynomial

   @gsuberland I've recently upgraded some lights around the house to the highest CRI I could find for 2700K (mostly Philips "ultra-definition") and honestly I can still see it's lacking when compared to incandescent

   I wish I wasn't so attuned to this

   • Embed this notice
     Graham Sutherland / Polynomial (gsuberland@chaos.social)'s status on Monday, 24-Nov-2025 11:14:47 JST Graham Sutherland / Polynomial

     in reply to

     (the "3500K" LED it picked there is actually closer to 3870K based on the published SPD)

   • Embed this notice
     Graham Sutherland / Polynomial (gsuberland@chaos.social)'s status on Monday, 24-Nov-2025 11:14:48 JST Graham Sutherland / Polynomial

     in reply to

     it's not really that surprising that the top-ranked by CRI ends up being a colder temperature than the top-ranked by CFI. CRI Ra focuses on just a few desaturated tones which means it tends to score 4500K+ LEDs better. getting a high CRI at 2700K is extremely difficult.

     CFI on the other hand covers a very wide range of tones, making it more representative of the perceptual fidelity you'll see. more red and orange tones are covered, so it's easiest to hit high scores with neutral CTs.

   • Embed this notice
     Graham Sutherland / Polynomial (gsuberland@chaos.social)'s status on Monday, 24-Nov-2025 11:14:49 JST Graham Sutherland / Polynomial

     in reply to

     the top-ranked LED by CRI Ra is the GW KAGMB9 KM 5000K, scoring CRI 97.2. it is advertised as CRI 90+. the TM-30-18 CFI for this LED is 92.8, which is very good. the R9 is 89.3.

     the top-ranked LED by CFI is the GW KACCBB GM 3500K, with a CFI of 95.3. the CRI is 96 and the R9 is 85.7.

     the CRI figure still tells us that it's a good LED in both cases, but can't really be relied upon for evaluating which of two LEDs is better once you're past CRI 90.

   • Embed this notice
     Graham Sutherland / Polynomial (gsuberland@chaos.social)'s status on Monday, 24-Nov-2025 11:14:51 JST Graham Sutherland / Polynomial

     in reply to

     the measurements are based solely on the published spectral data, so naturally there's going to be some variation and you've got to trust that the values they provide will be representative of the real product performance. but we can still do some interesting analysis.

   • Embed this notice
     Graham Sutherland / Polynomial (gsuberland@chaos.social)'s status on Monday, 24-Nov-2025 11:14:52 JST Graham Sutherland / Polynomial

     I wrote a script to pull down all of OSRAM's optical data packges, extract the spectral power distribution data, and run colour quality analyses on them. the reason for doing this was to take a look at the correlation between CRI Ra (the typical CRI number you'll see advertised), CRI R9 (the saturated red reproduction figure from outside the standard CRI Ra range), and the TM-30-18 CFI quality metric.

     CFI is a better measure, but the advertised CRI figures are still representative past CRI 90.