LED test / review - Nichia NV4WB35AMT sm653 R9080 E900 - very high light quality, but limited performance

LED Test / review EN

Nichia NV4WB35AMT sm653 R9080 E900


A multi-die LED from Nichia, the B35AM, has been available since 2019. Due to the lack of a dome, it is to some extent a competitor to the XHP35 HI or the newer XHP50.3 HI model, and it is also available in CRI 90 at up to 6500 K, which is also unusual by Nichia standards. This test clarifies to what extent this LED also performs at higher currents and whether it can really keep up with an XHP50.3 HI.

The emitter tested here, including the LED board, was made available to me for this test by BLF user @thefreeman. Many thanks for that!

Technical data

Tj 25 °C, If 1,400 mA
  • Type: multi-die, 2S2P
  • Bin: E900
  • Color group: sm653 (6500K)
  • CRI: min. 90 - R9 min. 80
  • Rated voltage: 5.93 V
  • Max. Forward current: 1,800 mA
  • Max. Peak current: 2,400 mA
  • Viewing angle: unknown °
  • Thermal resistance: min. 1.5 - max. 2.8 K/W
  • Max. Temperature Tj: max. 150°C

Datasheet can be downloaded here: [Datasheet (newest version, Nichia)] (https://led-ld.nichia.co.jp/api/data/spec/led/NV4WB35AMT-E(5768D)R70%20R8000%20R9080.pdf)

First appearance

The B35AM is a flat LED without a dome. The four recognizable LED chips are surrounded by white silicone and protrude only slightly. The gray substrate corresponds to that of other Nichia LEDs.

The large illuminated area compared to the size of the LED itself is striking. The edges around the illuminated area are very small so that the footprint area is used as efficiently as possible.

The B35AM is 3.65 x 3.65 mm in size. The format is therefore not compatible with standard dimensions, which can be a problem depending on the application and accessories required.

Due to the unusual dimensions, centering rings for XP or 3535 LEDs cannot be used directly. Thanks to the symmetrical housing, centering rings produced using a lathe can at least be used. I recommend custom-made centering aids using 3D printing or a lathe, or the use of other centering methods or TIR lenses.

The footprint is also unusual. Unlike most emitters in this performance class, there is no dedicated thermal pad. This is likely to limit the maximum performance. A board with the lowest possible thermal resistance of the contacts or a reduced operating current is therefore essential for this LED. This LED was tested on an LED board from ‘Eurekatronix’ (Clemence), which has the lowest possible thermal resistance under the contacts.

It should be noted here that this special LED board does not correspond to the standard dimensions of other 20mm boards and in particular the recesses for screw mounting had to be subsequently widened.


Thanks to the very closely spaced segments (LED chips) and the densely applied phosphor, almost no gaps are visible. Darker areas are only visible at the edges and in the middle, but these should not impair the optical properties.

The segments are connected as flip chips, there are no visible bonding wires.

The luminous area is 9.83 mm² in size. This value is very close to the 9.73 mm² specified by Nichia (3.12 x 3.12 mm illuminated area).

After operation at the maximum possible operating current, the luminous surface was minimally damaged (red circle in the following picture). Although the phosphor was not burnt, it is therefore not advisable to operate at the maximum possible current, even if this effect should not necessarily occur with every B35AM and is possibly also dependent on the CRI, because this is a factor for the thickness of the phosphor layer, among other things.

However, operation at 3-3.5 A should avoid this problem for longer periods of continuous operation.

Power and overcurrent capabilities

Within official parameters, as far as known:

  • at 1,800 mA (official maximum current): 1227 lm @ 5.86 V
  • Power at official maximum: 10.6 W
  • Efficiency at 1,800 mA: 116.4 lm/W

This LED is a good sample. Not only does it achieve the higher flux binning E1000 (1000 lm @ 1.4 A), but the Vf is also very low and below the typical values specified by Nichia.


  • Maximum reached at 5.0 A, at this point 2521 lm @ 6.24 V
  • Power at maximum 31.2 W
  • Sweet spot at about 3.4 A (2031 lm @ 6.09 V)
  • Power at sweet spot 20.7 W
  • Efficiency at maximum 80.8 lm/W
  • Efficiency in the sweet spot 98.2 lm/W

This LED behaves strangely when overcurrent is applied. Up to around 4.4 A, the curve behaves normally and rises as is already known from other LEDs, before suddenly flattening out and then ending as a plateau just 600 mA later (at 5.0 A). For me, this indicates a massive heat build-up, which occurs despite the LED board being as ideal as possible and low thermal resistance due to the lack of a thermal pad. It is likely that this point is reached much earlier with less high-quality LED boards.

The 719A also has this characteristic curve, but it is much less pronounced.

The B35AM is therefore not suitable for high-performance applications with thousands of lumens of luminous flux. If Nichia were to equip the B35AM with a thermal pad to eliminate the heat build-up, it would probably (estimated) achieve over 4500 lumens and 11-13 A max. current! (This is assuming no damage to the light surface).

Interestingly, the B35AM tested here is more efficient than the XHP50.3 HI, also in 90 CRI, but this could be also only due to the high CCT of 6500 K. The 719AC has no chance at all, which is due to its high thermal resistance because of the stacked LED chips.

However, the SFT-70-X is a surprise. Although it has 70 CRI at the highest CCT, at 7 A it is only 15-20 % more efficient than the XHP50.3, although the latter achieves the same luminous flux at a higher maximum current later on. However, it must be keep in mind that the XHP50.3 is not stable at the maximum possible operating current and quickly shows damage on the phosphor (see test).

In general, the B35AM has little chance against other emitters due to the lack of a thermal pad and the resulting low maximum output. As already mentioned, if Nichia were to massively improve the heat dissipation of the four LED chips, as would be possible with a dedicated thermal pad, it would in principle match or even slightly outperform the XHP50.3 in 90 CRI!

The comparison with a Getian GT-FC40 has been requested several times. I normally avoid such direct comparisons of different LED types (Vf classes), as this can quickly become misleading, as the following differences must be taken into account:

  • The GT-FC40 is a 12 V LED

  • The GT-FC40 has a huge thermal pad (7070 footprint)

  • The GT-FC40 has a considerably larger illuminated area with significantly more chips connected in parallel/serially (4S-4P)

As a result, the possible total output of the GT-FC40 is almost 5 (!) times higher at over 150 W, despite the same color rendering rating.


Values at 25 °C Tsp, at 85 °C Tsp values are 13 % lower

The luminance is low, a 519A with dome is only just surpassed, seen at maximum current. Due to the large illuminated area and the low maximum current possible, it is not worth mentioning. The LED is perfect for a good all-rounder, but it is not suitable for a thrower despite the lack of a dome.

The luminance was calculated for 5 A in order to prevent further possible damage to the illuminated surface.

Light quality and use in optics

The light pattern is perfect. There are no disturbing color distortions or rings. The latter can only occur with poorly calculated reflectors. This LED is perfect for all those who value an optimum light pattern!

Color and light quality

Despite the slightly greenish color localization according to the Nichia data sheet, the tint of this sample is good. The green tint is very low and the tint is also suitable for more demanding users without any problems. However, this LED is definitely of the colder variety, with almost 7400 K. However, this makes this LED unique again, as such a high CCT combined with extremely high color rendering is very rare and makes this LED suitable for daylight lighting (color temperature of an overcast sky).

Otherwise there are no surprises, the Nichia specifications for Ra (min. 90 ) and R9 (min. 80) are met without any problems, and the extremely high R9 is also in line with the SFT-40 3000 K and Getian FC-40.

  • Ra: 96
  • R9: 99
  • CCT: 7370 K
  • duv: 0.0019


The B35AM is an ambivalent LED. I like the extremely good light quality typical of Nichia in optics, without any artifacts, and the extremely high color rendering with a very cold color temperature, which opens up new possibilities, especially in the field of daylight lamps.

However, this emitter also gets in its own way.

The biggest drawback is the lack of a thermal pad and the unusual footprint, which limits the maximum possible output. There is also a risk of permanent damage to the LES if the emitter is operated at maximum power for long periods. The luminance is also relatively low as a result, but this is only a disadvantage for thrower applications.

All in all, a very good LED with excellent light quality if a good all-rounder is required, but only if light output is not the main priority.


  • Perfect light image in reflectors and lenses without artifacts
  • very high CRI
  • available in many different CCTs and Ra


  • Relatively low luminance


  • no dedicated thermal pad
  • unusual footprint, unusual dimensions
  • Limited maximum output

Thank you for reading the test. :slight_smile:

Greetings, Dominik

10 Thanks

Thanks for the test.

This confirms what I suspected using wonky comparisons between Clemence (B35A) and Djozz (FC40) tests, the B35A is a superior LED vs the FC40 at low to moderate~high power that we find in EDC lights.

6W nominal :
B35AM 1A : 740lm , 130lm/W
FC40 0.5A : 610lm : 113lm/W

12W nominal :
B35AM 2A : 1335lm , 113lm/W
FC40 1A : 1100lm : 98lm/W

18W nominal:
B35AM 3A : 1850lm, 102lm/W,
FC40 1.5A : 1700lm, 99lm/W (higher than at 1A, reading the exact values on the graphs is hard)

24W nominal :
B35AM 4A : 2270lm, 92lm/W
FC40 2A : 2200lm, 94lm/W

The B35A is measured about 1000K higher than the FC40 (6300K) but at these high CTTs it shouldn’t matter much, maybe a couple of % of efficacy.
So it has higher to equal efficacy up to 18W while being less floody and having better duv in general, especially at low CCT where the FC40 of often yellowish whereas the B35A goes under the BBL.

If only it had a thermal pad it would be perfect.

6 Thanks

Thanks @koef3 for the tests! Have been eagerly waiting for more concrete data on the B35AM. As @freeman laid out your excellent review confirms what we thought. It’s more efficient than the 719A but with the caveat of a suboptimal thermal interface for higher power. Zeroair’s review also confirms the efficiency gains compared to the 519A. In real world use it’s 6v architecture also guarantees that a more efficient boost driver is required for flashlights.

Will definitely be choosing this emitter in the S21E :wink:

Thanks for the tests. In terms of lumen output, is there any advantage of the S21E with B35AM over the S21E with 519A?

Zeroair reviews both leds in the convoy s21a and s21e. Comparing his results the B35AM is more efficient and uses less power for the same output compared to the 519A.

Looking at claimed numbers for the S21E, the 519A appears to have a 6A linear driver while the B35A has a 2.2A output boost driver. Djozz shows the 519A making 1370lm at 6A while this post has this B35A making a hair under 1500 at 2.2A.

Convoy, however is probably not using as high a flux bin as this; one product image shows their 4500K B35A being E800, so there’s probably no significant difference in maximum output compared to the 519A. The B35A is doing it with about 2.2A * 6.0V = 13.2W though, and the 519A is using 6.0A * 3.3V = 19.8W.

That’s what the LEDs are actually using anyway. On a full battery, that linear driver will pull 6.0A * 4.2V = 25.2W, making 5.4W of waste heat.

1 Thank

Zeroair’s S21A graphs suggest the mechanical switch boost driver performs differently from my S21E 719A. I should note the high-drain Samsung 50S battery gave me a few minutes more runtime than a low-drain 50E. A B35A should run a little longer due to its lower Vf.

Thanks. So likely, similar(ish) lumen output, but longer runtime with the B35A?

Yes, longer runtime, and not just a little bit. See Zeroair’s runtimes for the 519A. They’re uncooled and thermal throttled with a 3750 mAh battery. He gets 70 minutes on high, almost all of which is throttled to just under 50%. With a 5000 mAh battery, we could expect 93 minutes.

Here’s my 719A without cooling using a 5000 mAh cell. It, too runs at about 50%, but throttles in small steps and takes a lot longer to do so. It’s just under 200 minutes. The B35A, again should run longer and/or throttle less/slower than the 719A.

1 Thank

Great work! Very interesting and thorough review.

Anecdotally, I use a s21e b35am 5700k with a 50s. It looks fantastic for such a cold cct, has a teriffic beam. Takes a long time to heat up too since its such a low wattage. One of my most carried work lights.

Thanks, very good testing and provided data.
But led itself being nominally 6500k was mesured at 7400k, very cold blue.
And if not using this super heat conductive mcpcb, that only Eurekatronix produces, led was just burning at 3A in conoys and emissars.
Typical performance of 4500k 9080 was much lower than this 6500k.

B35AM Emisar were initially 3.6A, I think they’re at 3A (?) now due to this. Convoy MCPCBs are worse so Simon lowered the drive current to 2A (from 3A).
But yeah not having a reliable source of good MCPCB is a problem.

What do you use to measure the CRI, CCT, and DUV?

I wonder what the intended application of this LED is and why they would limit performance in this way. Maybe it’s a cost optimization? Market positioning is an possibility , but doesn’t really make much sense in this context to me.

Despite the great thermal conductivity, this board is an absolute nightmare to solder on. It was extremely difficult to reflow the B35 onto this board because of the (way too) big solder pads, and also soldering the connectors was extremely hard, because the heat is simply dissipated away from all the solder pads into the heatsink. The Noctigon DTP boards are way easier in handling and soldering.

Yes, I don’t understand that either. A dedicated thermal pad shouldn’t really be a cost issue, as Luminus and Cree LEDs (which are now really cheap) also have them.
Even at second glance, this strange footprint doesn’t make sense to me, the 3535 footprint would have done just as well, as the package hardly differs in dimensions by 0.2 mm.

I always use BiSn for these MCPCBs, it’s much easier (I should have joined some), also when soldering wires in place in the light, raising the MCPCB a bit to break thermal contact helps a lot.

That’s what I did in the end.
The problem is that the LED board normally has to be screwed onto the heat sink before I can connect the contacts…

Testing other solder is another approach tho.

Not necessarily, the screws can be not fully screwed down, on lights where there is very little play I use thin nickel strips with the end bended into a hook, slid under the MCPCB. (Not sure my explanation is good)

I do not connect the LED board like in a normal flashlight to keep the voltage loss as low as possible, so I am limited in my possibilities…

But I found a way :smiley:

Ah I thought it was about soldering in lights.
I’m curious why the wires can’t be soldered before mounting, for accurate Vf sensing one would use dedicated voltage wires separate from power wires, all four wires should be solderable before mounting, but I don’t know your setup.