Nichia NF2W757GR-V3 [GRV3]

[Agro]

@clemence

1. You’re welcome. I also really curious about this VS35SP36. I did an undocumented quick test, that’s why I said I know this board will work reliably for up to 40 watt continuously. Designed only for 36 watt of power, this test result was a good surprise :partying_face:

@clemence

2. It’s not relevant but I did it anyways. Done!
   

Ouch, the difference is so large that it’s hard to use this as a reference.

@Clemence

You owe me now.

I do:)

@clemence

Nope it’s not the equivalent at all, the characteristic are totally different. Like our conversation on the other day, ceramic MCPCB behaves differently due to it’s very low thermal resistance. Most of the heat travel down and only small percentage spread radially. In “normal” epoxy/fiber glass insulated MCPCB, the lateral thermal resistance is much higher, thus most of the heat go radially/sideways. Think of it as a graphite heat spreader which has faster radial than lateral heat transfer. That’s why epoxy insulated MCPCB needs large area and thick copper layer for good heat dissipation.
Ceramic MCPCB on the other hand works best with as thin as possible copper trace. Anything between the heat source and heatsink adds up the total thermal resistance.
So, in my opinion, it’s near impossible the same epoxy insulated MCPCB to perform as above with all dimensions kept the same. To my experience, ceramic MCPCB is more likely to perform 90-95% as above graph.

Thank you, this explains the poor performance of 36-LED MCPCB compared to the 1-LED one.

But this is where flashlights differ from rigs like yours.
Here the ceramic insulated MCPCB may indeed transfer the heat mostly laterally to the water.
In a flashlight, MCPCB transfers heat to the shelf and shelf transfers it to the body.
The shelf is cooled at the sides, so the heat must travel sideways all the way out. Through copper traces? Through MCPCB? Through shelf?
I think through all these layers. So in a flashlight that would be different. Traces would contribute to the total thermal transfer and thick ones would be a win for ceramic MCPCB as well. Though probably a marginal one.

@clemence

3. As far as I remember the price should be the same.

Interesting. Clearly 6V is a better value option…

@clemence

4. As mentioned in the OP and per your request, yes, it’s 36P. No, thermal runaway visibly detected even down to 0,8333mA/LED (30mA total input). Actually, insulated MCPCB are very resistant to thermal runaway phenomenon due to it’s heat spreading characteristic. Ceramic MCPCB should be worse in this regard.

Again, in a flashlight that would be different. Central LEDs would be hotter than the side ones. That would make them more suspectible to runaway.

But the result is good.

@clemence

5. As I explained above, insulated MCPCB needs large surface area to perform well. With only 0,4mm gaps between the LEDs, there’s not enough surface area to dissipate heat at higher current. I rechecked the damage and yes, it’s a burnt silicone not dusts as I suspected. All burnt dots located at exactly the same location, above the cathode - around the gold wires. Surprisingly the module performs as good as the last time, no measurable decrease. I could replicate the same output at the same forward voltage. Maybe because the LEDs on VS35SP36 although overheated, never exceeded the rated max 400mA. Usually when the die(s) is damaged, the voltage at given current is lower with reduced output.

Yes, I’m also surprised to see that with phosphor damage the output is the same. Even with the die intact there should be a loss.
4 LEDs overheated, at least one did it at 10.8A. The others survived to much higher total current. I wonder…why these 4 and not the others? Maybe they were on a way to thermally run away but the cooling was good enough to prevent them from taking all the current on themselves? As of now I don’t have a better hypothesis….

@clemence

6. I made a scale up diagram of the three MCPCB construction:

That’s very interesting!
But I can’t figure out which MCPCB is which. In your store I see that VR10S1-757 is copper. In the OP there’s a side of VS35SP36 visible and it looks copper as well. Ceramic is Alu. But I see 1 copper MCPCB as well as 2 alu ones on the diagram….

@clemence

7. Everything under the hood are the same. It’s total loss water cooling system. I did all the test in 29C ish air temp. The water should be slightly lower than that.

Yeah, I think that the effect of horizontal heat spreading on the insulator thermal resistance explains the lower performance when using 36 LEDs.

@clemence

8. The temp and voltage difference are significant, especially with ceramic MCPCB which had lower temp and higher voltage. You can check the difference in below chart

This is a nice chart, but I meant the difference between VS35SP36 and VR10S1-757. If thermals are a limit, there should be a clear difference in Vf between them.

@clemence

9. No, I measured the output in separate light box using Maukka’s calibrated lights. I have four sets of them: R70, R9050, Rfa00 6500K Optisolis, and Rfe00 2700K Optisolis. I used calibration factor between R70 and R9050 since these GRV3 are R8000. The output for both single LEDs and 36pcs LED averaged almost dead spot on to Nichia’s spec. Both two single LEDs were first calibrated at 150mA while the 36pcs were calibrated at 360mA total input to get the 50mA output value. The three values were the control value to calculate the base point for all charts.

This is a meaningless side note, but to be honest I don’t understand that…you surely clamp the MCPCB to the water loop in the output test as well, right? Are the clamps outside of the tube opening?