Nichia NF2W757GR-V3 [GRV3]

clemence · 2019-07-28T10:05:57.000Z

Nichia sent me hundreds of several different LED types for my past city garden light project prototyping. I found some very interesting “BLF-worthy” LEDs. One of them is currently the only medium power Nichia LED which is OEM rated up to 2,4 watt/LED. For 3030 medium power LED, it’s like an XHP50 to XML. It’s Nichia’s most powerful medium power LED. There are two variants available: NF2W757G-V3 (3V) and NF2W757GR-V3 (6V)

I’m only interested in NF2W757G*R*-V3 (GRV3) because it’s OEM rated up to 2,4watt, 2nd most efficient in Nichia MP LED, and available all the way from 2000K up to 6500K in R70, R8000, and R9050.

If you opt for 3V version, forget NF2W757G-V3, better go with NF2W757G-V3F1 (lower voltage, brighter). There’s NF2W757H with even higher efficiency butt… unfortunately, only a tad higher than GV3 at below 35mA. Above 35mA, GV3, GRV3, and GV3F1 smoked it by significant lead.

Prepared some test victims:
The biggest board (VS35SP36) is from my early attempt, you can see some solder balls and slightly misaligned LED. I don’t think that would cause measurable output difference. I don’t want to kill another 36pcs (only few left) nor doing rework on them. Because another rework will just put more stress on the LEDs, possibly reducing the output. They’re all NF2W757GRT-V3, P20, sm305, R8000.

VS35SP36 wired in 36P (6V)

Total loss water cooling. Had to shielded from the light, that was blinding!

My RD DPS5015 couldn’t go higher than 15,09 A at below wattage. I only have 180watt DC PSU at 19,65V max voltage. Wiring the module to 3S12P would require close to 21V input.
The LED module maxed out just a tad lower at 14,4A.

Here’s the result:
It’s actually unfair to compare the tiny 10mm non DTP copper board to 21mm ceramic non DTP board but that’s all I had. You might think GRV3 over current is not that great or perhaps why the 10mm board performed somewhat bad. The output of 10mm started to look less than adequate at 200mA. Remember this is a 6V LED, that means it outputs roughly twice the 3V LED. So if you use 3V LED, 10mm copper non DTP should perform equally good to ceramic up to 500mA (in single LED application).
The voltage swing is quite high from 5,68V at 50mA up to 9,01V at 850mA

I included the plot of ceramic MCPCB ouput prediction using data from VR21P4 above multiplied by 36. In real world we can expect about 90-95% of the plotted output. RD’s DPS5015 couldn’t get steady reading below 50mA (pulsating output). The efficiency should be higher than 240lm/watt at 30mA total current.

4 LEDs damaged. Still unsure if it’s test related or simply burnt saw dust. My workshop was rather dusty during the test. Saw the smoke started at 10,8A but the output was climbing steadily up to 14,4A. At 8800lm, even a slight oily surface can start the burn.

Conclusions:

- See what can we get from 36pcs $0,1 LED + $1 MCPCB… Of course any high power LED such as XPL2, 319B, etc…will outperform these LED. But we have to spend $$$ to get the same or higher efficiency than this. At $4,6/module it’s the cheapest 9000lm CRI80 I’ve ever seen.

To get the most out of these GRV3, a nanoceramic MCPCB is needed. But at lower power level this VS35SP36 is more than capable to give 5000lm of efficient lighting. In general lighting it would be hard to design a fixture with passive cooling for more than 50 watt of LED power. In my opinion, 2000lm per module with this GRV3 is the sweet spot for efficiency and heat management. 180lm/watt is very good.

[Clemence]

staticx57 · 2019-07-28T17:55:47.000Z

Following!

Corvette · 2019-07-28T18:11:24.000Z

Very interesting LED, subscribed

SKV89 · 2019-07-28T20:03:36.000Z

Very interesting! What about the tint?

id30209 · 2019-07-28T20:35:24.000Z

Subscribed

clemence · 2019-07-28T20:51:41.000Z

OP updated!

hasddie293 · 2019-07-28T22:29:36.000Z

This is a very good newss my next pick for my warmer lights… this should be high cri…hope so…

djozz · 2019-07-28T22:34:12.000Z

Thanks for the test! These are cheaper than Optisolis and more efficient. So if ultrahigh CRI is not needed (these still come in R9050) these are wonderful.

clemence · 2019-08-06T15:59:20.000Z

UPDATE 190729; 14:51 GMT+8:

OP updated. Switched to orange color for easier viewing. I also added two comparison plots of theoretical ceramic board output. I use 95% output because it’s unlikely that 36pcs of closely packed LEDs would output the same as a single LED.

.
.

[Clemence]

Agro · 2019-07-29T08:03:57.000Z

Thanks Clemence for the interesting test.

Could you multiply the VR10S-757 chart by 36? That would be the equivalent of VS35P36, right?
You could use the difference between this chart and the VS35P36 to make a more accurate prediction of the lossess caused by tight packing.

I would be much more interested in 3V LEDs due to better driver availability as well as lower driver price but these look cool nevertheless.
Is the price the same for both 3V and 6V? That would make 6V a much better option as long as there are enough of them to fund a better driver. Or to have cells in 2s.
How did you connect the LEDs on VS35P36? 1s36p, right?

I see that 36 LEDs max out at much lower current than a single LED. Single went nicely to about 500 mA and maxed out at 650 mA with the cheap MCPCB. 36 go nicely to about 300 mA and max out at 400 mA. A part of the reason may be the burnt LEDs…but maybe also just thermals? And since the construction is the same from the die down to MCPCB copper (it is, right?) I would guess the main difference in cooling is actually below MCPCB. Maybe the PCB-block interface, maybe the water loop itself. Anyway - a difference in temperature should manifest in a different Vf. The charts are not precise enough to see whether the difference is there or not….is it? It surely doesn’t look large…

On a separate topic, I wonder how do black clamps next to the LED affect the measured output…

clemence · 2019-07-29T15:08:23.000Z

[quote=SKV89] Very interesting! What about the tint? :D [/quote]

Lend me your spectrometer! The sm305 R8000 is so-so by our tint snobism standard but I like the R9050 version. Nowhere near as good as Optisolis for sure, but tint wise it's OK. The sm405 R9050 is similar to LH351D in my TO50R. The sm305 R9050 has more deep reed tone compared to the sm305 R8000.

[quote=djozz] Thanks for the test! These are cheaper than Optisolis and more efficient. So if ultrahigh CRI is not needed (these still come in R9050) these are wonderful. [/quote]

Djozz, there's a bad news for you...Only E17A and E21A has R9050 2000K and 2200K. I checked the entire catalog and for 757 family, anything below 2700K only available up to R8000, sad day...

[Clemence]

SKV89 · 2019-07-29T16:27:39.000Z

If it is similar to LH351D that means it is above the BBL +DUV. At least your board works with Optisolis which is generally on the DUV or just slightly above but some tint mixing will get it down to anyone’s preferred tint. Too bad E21A is not compatible otherwise 32 E21A can produce some monstrous ultra HI CRI output with good tint.

clemence · 2019-07-29T16:30:23.000Z

SKV89:

If it is similar to LH351D that means it is above the BBL +DUV. At least your board works with Optisolis which is generally on the DUV or just slightly above but some tint mixing will get it down to anyone’s preferred tint. Too bad E21A is not compatible otherwise 32 E21A can produce some monstrous ultra HI CRI output with good tint.

…
And monstrously expensive too

clemence · 2019-08-06T15:58:18.000Z

Agro:

Thanks Clemence for the interesting test.

Could you multiply the VR10S-757 chart by 36? That would be the equivalent of VS35P36, right?
You could use the difference between this chart and the VS35P36 to make a more accurate prediction of the lossess caused by tight packing.

I would be much more interested in 3V LEDs due to better driver availability as well as lower driver price but these look cool nevertheless.
Is the price the same for both 3V and 6V? That would make 6V a much better option as long as there are enough of them to fund a better driver. Or to have cells in 2s.
How did you connect the LEDs on VS35P36? 1s36p, right?

I see that 36 LEDs max out at much lower current than a single LED. Single went nicely to about 500 mA and maxed out at 650 mA with the cheap MCPCB. 36 go nicely to about 300 mA and max out at 400 mA. A part of the reason may be the burnt LEDs…but maybe also just thermals? And since the construction is the same from the die down to MCPCB copper (it is, right?) I would guess the main difference in cooling is actually below MCPCB. Maybe the PCB-block interface, maybe the water loop itself. Anyway - a difference in temperature should manifest in a different Vf. The charts are not precise enough to see whether the difference is there or not….is it? It surely doesn’t look large…

On a separate topic, I wonder how do black clamps next to the LED affect the measured output…

In the spirit of Argo’s questionnaire style….

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:

2. It’s not relevant but I did it anyways. Done! You owe me now.

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.

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

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.

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.

6. I made a scale up diagram of the three MCPCB construction:
(Click to enlarge)

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.

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

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.

[Clemence]

Agro · 2019-07-30T12:56:41.000Z

clemence:

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:

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:

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

Interesting. Clearly 6V is a better value option…

clemence:

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:

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:

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:

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:

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:

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?

clemence · 2019-07-30T18:48:24.000Z

LEFT TO RIGHT: Ceramic MCPCB/VS21P4 - aluminum MCPCB/VS35SP36 - copper MCPCB/VS10S1-757

Solder layer thickness is approximately as thick as the masking thicknes.

Agro:

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?

I was explaining the measurement were two steps. First, in the light box to get the correct base point. Second, to get the whole relative output measurement. The second measurement then adjusted to the graph using the first measurement as a datum point.

[Clemence]

clemence · 2019-08-03T16:17:06.000Z

Thermal images above showed how minimal heat spreading on nano ceramic based MCPCB, the thermal imaging profile looks very similar to copper DTP. On insulated MCPCB, we usually see shallow thermal gradient from hotspot.

[Clemence]

Agro · 2019-08-05T06:48:01.000Z

Yes, but that’s again a setup that cools from the bottom rather than from the sides.

Agro · 2019-08-07T08:19:58.000Z

Clemence…you have a nice business relationship with Nichia.
And Nichia makes nice LEDs.
However, since MP is so much about lm/$, have you considered Chinese LEDs to lower the cost further?

clemence · 2019-08-07T09:18:25.000Z

Agro:

Clemence…you have a nice business relationship with Nichia.
And Nichia makes nice LEDs.
However, since MP is so much about lm/$, have you considered Chinese LEDs to lower the cost further?

No, cost is not my primary concern in this business.

[Clemence]