Review, Teardown: Lightway 17W Dimmable LED Bulb (43287 & 43288)

It seems that LED lighting has become mainstream, with CFLs slowly starting to disappear from shelves. It’s a good thing, both from an environmental perspective and from an economic perspective, as it seems that LED globes have fallen in price dramatically and their designs are less inconsistent than before.

Recently, on Saturday 13th May, Aldi (in Australia) had a Special Buys sale of LED Dimmable Bulbs for AU$6.99 a piece. The listing had claimed that 7W, 13W and 17W bulbs would be available at the same price, with the LEDs used coming from Epistar (a Taiwanese company, known for making lots of mainstream LEDs). This proved somewhat attractive for a number of reasons – the first is that LED globes with a rating of above 13W are not easily available in Australia, and so realistically most LED retrofit globes top out at about 1100lm (with a few exceptions). The second was the price, which was quite reasonable especially for a 17W globe.

I am assuming that the price came about because it was the “average weighted price” for the proportion of bulbs ordered – after all, a 7W globe should be much cheaper than a 17W globe to produce, so in order to have this price, there wouldn’t be many 17W globes available. After visiting Villawood Aldi at opening time on the Saturday, I’d have to say this was somewhat confirmed, with a lot of digging required for me to score three 17W bulbs of each base type (BC and ES), and only less than a few remaining after I had checked out. Given the choice, I would go for the 17W every time, so I suppose others may have to settle for the others as a consolation prize. Strangely enough, on that morning, I was the only one interested in the globes – go figure.

So lets take a look at the bulbs and whether I got my money’s worth.

The Units

The units were branded Lightway, and claimed to be “powered by Muller Licht”. This latter brand is one that appears only on Aldi products, so it’s likely that this just means this is their own brand stuff, but it’s a way to jazz it up. It was available in both Bayonet Cap and Edison Screw bases, so I purchased both just in case there were any differences (which there appear to be none). The LED bulbs are dimmable and feature a 1500 lumen output for 17W input (88lm/W efficacy, which is good but not the best). Additional features include a claimed 25,000 hour lifetime, a CRI of >90 (which is surprisingly good), >=100,000 switching cycles and most importantly, a three year warranty. The unit has regulatory approval marking for sales within Australia, and is Made in China. Helpfully, dimensions of the globe versus a regular GLS globe are provided on the rear, although the packaging is a very inconvenient and inefficient plastic bubble.

While the luminous efficacy is a little lower than the class-leading products (now close to 115-120lm/W), it seems the increased lifetime claim and CRI may well be part of the reason. In order to achieve an increased lifetime, one way to achieve this is to claim a lower lumen rating so that it achieves 70% of the claimed lumen rating at the lifetime claimed – the downside is that the calculated luminous efficacy will be lower. The higher CRI chips may well be less efficient, as a more varied blend of phosphors is needed to achieve the better CRI rating. As a result, it seems to be a trade-off towards light quality over absolute efficacy figures, which should hopefully satisfy those who may not have used LED lighting in the past due to its poorer colour rendition.

The globes are very similar to every other LED globe, although scaled up just a little bit more due to its size. Again, there is a long heatsink neck, plastic coated, and a diffusing dome at the end. The globes measured approximately 127.2 grams on my scales, which I feel is on the light side.

The markings on the collar seem quite similar to those from some other brands (e.g. IKEA) although a little bit more smudged. It’s an LED bulb, so not much to see from the outside.

Teardown

Surprisingly, opening this globe proved to be fairly easy. The front diffuser separated from the base with less force than other globes, possibly due to the way the silicone was applied.

Inside, we are greeted by an MCPCB containing the LEDs, with an IC and resistor on the board as well. This is an uncommon arrangement. Wires from the current driver are soldered onto the MCPCB (likely to save on the cost of a connector).

A closer look shows that this unit comprises of 66 x 2835 SMD LEDs. Looking at a contrast-enhanced version of the image, the MCPCB sections connecting the LEDs are irregularly shaped, although as large as possible to help move as much heat away from the chips as possible. The unit appears to have 22 series connected sets of 3 LED packages in parallel, hence the 66 number.

The IC mounted on the board is a Winsemi WS9931 Controller for LED Current Ripple Removing Circuit. I’ve not seen one of these used in the past, however, it claims to reduce low-frequency current ripple by being in-series with the LEDs and adaptively changing its resistance as a linear current regulator to regulate the current across the LEDs. It operates for currents from 100 to 300mA (typical, 240mA), at temperatures up to 120 degrees C with shutdown occurring at 135 degrees C. The datasheet does specify that there is a 350ms start-up blanking time required to build up the internal reference, followed by a gradual removal of ripple from there on. The inclusion of this IC probably helps remove some flicker, although at the cost of efficiency as the chip is dissipating energy as heat (hence, the lower luminous efficacy) and rapid start-up. It may have enabled them to use a less sophisticated, or even less-filtered power supply to do the job.

Scraping around the silicone bead, the text appears to say MK-A66DIM-66SMD2835-V0.1(3522P) (HY) 2016 10 12. A fairly recent design, it seems.

The LED MCPCB is affixed to the outer heatsink just by a friction fit to a lip in the heatsink cone. Contact only occurs at the edge, and the silicone bead does not reach the thermal interface (edge of the board) itself, so thus a metal-on-metal contact is what is being used. This isn’t an optimal situation, and really, some thermal interface compound is probably best – as is increasing the contact area through which heat is transferred and even the contact pressure. I suppose, that’s one thing they have to do to save costs.

The power supply/current driver itself is completely encapsulated in potting compound in the base, and could not be examined. Unfortunately, that means it was not possible to assess whether any lifetime/heat-sensitive components (e.g. capacitors) have been appropriately rated and specified. The potting compound itself is likely to help reduce any ballast noise by physically dampening it, as well as improve against any environmental impacts (e.g. moisture). However, whether it is good for heat dissipation is not known.

Testing

Subjective Opinion

On powering up, it was observed that each package had two LED dice. I suspect these are connected in series, which was confirmed through line voltage sweep measurements. This may be more reliable and manufacturable compared to having twice as many single-die chips, but it does expose one major Achilles heel of higher rating LED globes built with low-powered dice and that is LED chip wire bond reliability through thermal cycling. As a result, the actual circuit looks a bit like this – current imbalances could well exist in this arrangement and would serve to shorten the lifetime if severe enough.

                      Three packages in parallel
                          |--------------|
+ --- [RIPPLE FILTER] ------+-----+-----+           ------
            |               |     |     |             |
            |              +-+   +-+   +-+            |
            |              X X   X X   X X            |
            |              + +   +-+   +-+            |
            |               |     |     |           ONE LED
            |               |     |     |           PACKAGE
            |              +-+   +-+   +-+            =
            |              X X   X X   X X          TWO CHIPS
            |              + +   +-+   +-+            |
            |               |     |     |             |
            |               -------------           ------
            |               |     |     |
            |              +-+   +-+   +-+
            |              X X   X X   X X
            |              + +   +-+   +-+
            |               |     |     |
            |               |     |     |
            |              +-+   +-+   +-+
            |              X X   X X   X X
            |              + +   +-+   +-+
            |               |     |     |
            |               -------------
            |         [... continued for a 
            |     total of 22 series segments...]
            |               |     |     |
            |              +-+   +-+   +-+
            |              X X   X X   X X
            |              + +   +-+   +-+
            |               |     |     |
            |               |     |     |
            |              +-+   +-+   +-+
            |              X X   X X   X X
            |              + +   +-+   +-+
            |               |     |     |
- ---------------------------------------

Another thing that was observed was the non-instant startup nature of the globe. This globe does take about 1-2 seconds to ramp up to full brightness on a power on. This may be a combination of the power supply/current drivers’ behaviour with that of the ripple filtering chip. It’s not a major issue, but it does differ from some other LED globes which come on after a very short delay at full brightness, rather than ramping.

One key issue was the stability of the globe when faced with mains ripple signalling, used to control off-peak devices. At my test location, the 1042Hz signal was measured to be 12Vrms on top of the 240Vrms mains, and this was sufficient to upset the globe and cause it to flicker randomly like a light in a horror movie during a storm. Because such signalling events happen for about 3 minutes every half-hour in the evening, this behaviour can be quite distracting. This may be due to the dimming logic being unable to separate this interference from legitimate phase-control dimming signals.

The running temperature of this globe was also of concern. In open air with a base-down configuration, a temperature of 85 degrees C on the outer shell was measured. The internal components would be at a higher temperature than this. Such high temperatures will stress the components within and shorten their lifetime – electrolytic capacitors are likely to only give their “datasheet rating” when operated at such high temperatures. Such large temperature swings may also place stress on connections due to thermal cycling.

On the whole, light quality for this globe was quite pleasing, and noticeably improved over the “>80 CRI” globe it replaced. The light was more rich and felt slightly warmer – it’s hard to describe with words, but it does resemble incandescent lighting more than the LED globe which was previously used. Testing for CCT by using white balance correction on the camera resulted in a figure of 2800K, close to the claimed 2700K. The light quantity was pretty much as expected – roughly matching what I remember from a 1500lm Philips Tornado Spiral CFL previously operated in the same fitting.

No RF interference issues were observed while using the globe.

Power Parameters

On a warm-up from cold test, I left the bulb running for an hour instead of the normal 20 minutes due to the fact this globe does get quite hot even in open air (measured 85 degrees C surface temperature for an ambient of about 25 degrees C). The action of the ripple filter can be seen on power-on, where the power takes a few readings to reach operating level. Initial power exceeds 17W by almost a whole watt, but after warm-up, settles to about 16.7W. In all, it seems the rating is honest, and there is no major “throttle-back” in output due to overheating (if such capabilities are even built in).

The globe had no difficulties within the operating voltage range, showing an above 0.91 power factor which is a good result. Throughout the extended range, proper regulation was maintained down to 160V, which is not as wide a range as some other globes, but is more than necessary. This does imply that each LED package is two chips in series, thus the LED voltage drop would be 22 x 2 x 3.3v = 145.2V or thereabouts, thus failure to regulate at slightly above this voltage is expected due to a (likely) buck regulator topology.

Dimming Compatibility

The unit was tested with a leading edge and a trailing edge dimmer and performed satisfactorily. It did not start up quickly at the lowest settings, however, did eventually come to life. Dimming to very low levels was achieved on the leading edge dimmer, and ballast noise was not apparent. The better compatibility may be due to the 17W load exceeding the 10VA minimum loading by enough margin to ensure the TRIAC fires consistently. Dimming did, however, exacerbate flicker issues when mains off-peak ripple signalling occurred.

Conclusion

On the whole, I think these globes represent roughly good value for money. The light output quality is good and definitely seems higher in CRI than the “>80” globes I formerly used. The light quantity appears honest, and a definite boost over the 13W swapped out. I saw no RFI issues, and dimmer compatibility was good especially due to the higher loading which may be enough to overcome minimum load requirements for older dimmers. The ballast was quiet and survived running for about two weeks on a daily basis now.

The downsides seem to be a desire to cost-optimize the bulbs. This results in no thermal interface material between the LED MCPCB and the shell heatsink and now, even a lack of screws or thermal adhesive. Instead, the majority of the work seems to be done by a friction fit of the MCPCB into the channel of the outer cone shell heatsink at the edges alone. Whether this maintains good thermal conductivity over the lifetime of the globe remains to be seen. The electronics were fully potted and not subject to examination – the quality of the components within could not be assessed. There is a reasonable suspicion that the capacitors may not be the best available. Given the side of the bulb was measured hitting 80-85 degrees C operating in open air, it seems likely all elements of the globe are running at thermal limit to reduce the weight and cost of the globe.

The biggest and most obvious downside, however, is its susceptibility to 1042Hz off-peak power ripple signalling, despite having a dedicated ripple filter chip mounted on the MCPCB. Every time the signal is injected into the mains (roughly every 30 minutes for 3 minutes in the evenings), the globe flickers almost if possessed (think horror movies) as the dimming logic fails to cope with the interference. As a result, this can be quite distracting (or even distressing to some). Full brightness is also not attained quite instantly – there is about a 1-2 second delay as it ramps up as the ripple filter “tracks” the incoming ripple. If you can cope with it, I suppose that’s no major issue.

I suppose you do get what you pay for – so while the globe is cheap, I have my doubts as to whether it will achieve the claimed 25,000 hour lifetime. My best guess is maybe in the 10,000 to 15,000 hour ballpark, especially as operating in more enclosed environments will increase the operating temperature, which consequently shortens the lifetime of the components within. It does come with a three year warranty, so that is peace of mind, and if you did leave it on 24/7, that would account for 26,280 hours.

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Lighting and tagged , , , , . Bookmark the permalink.

Error: Comment is Missing!