Review, Teardown: Atom Lighting AT9033/WH/WW 10W LED Downlight

The review challenge submissions keep rolling in. This post looks at the Atom Lighting AT9033/WH/WW 10W Warm White LED Downlight, another product for the Australian market, also anonymously donated. Lets put it through its paces and see what we can learn from it.

Unboxing and Features

Atom Lighting is a brand I have heard of before, although I had never had my hands on any of their products. Their green logo is suitable, especially in light of their “a greener future – one atom at a time” slogan. The unit is packaged in a colour print thin cardboard box, with a short list of features at the top including a colour temperature of 3000K, 800 lumen output (80lm/W efficacy), 90 degree beam angle and 90mm cut-out. It claims compatibility with “selected dimmers”. It claims an IP20 rating, making it suitable only for indoor installations, and a three year warranty which seems pretty standard.

The side of the box covers basic measurements of the unit, which has an 87mm throat (recommended 90-93mm cut-out) and 58mm height. It claims to be double insulated, for indoor usage, and carries the necessary regulatory compliance mark (RCM) for the Australian market. It also carries a fire-rating for mounting on flammable surfaces, it seems, with the F in inverted triangle, although it doesn’t carry any IC-mark at all.

Supplied inside is a combined installation sheet and warranty statement on the opposite side. Within the installation sheet, there is a suggested method of installation diagram which doesn’t clearly depict this particular unit. The key takeaway is the need for 50mm separation from thermal insulation, 25mm from combustible material, and 50mm from the rear of the unit to combustible material above.

It shows other variants of the unit with 4000K and 5000K CCT achieving 88lm/W, with this unit at 3000K achieving 80lm/W efficacy.

It seems that the lack of IC-mark is corroborated by the phrase “always ensure there is adequate free air ventilation around the luminaire at all times.” Unfortunately, this means extra hassle in installation, and significant potential for heat loss around the fittings into the roof, and thus, increased heating energy bills. Accordingly, there is no gasket supplied to seal around the downlight, thus dust marks may show due to air movement around the downlight recess due to breathing of the ceiling space. The unit is available in a silver finish instead of white.

Curiously, looking through the datasheet, just like the SAL Wave, this unit seems a little light on specifications. This unit doesn’t have a lifetime expectancy figure, nor does it have any CRI information. These figures are normally provided on most products, with more professional grade products providing even further information on colour matching (SDCM), inrush current, etc. It does however, specify its maximum ambient temperature as 50 degrees Celsius.

True to its name, the downlight unit features a green driver “cap”, which is curved and distinctive. The curve does make it tricky to set it on its back especially during installation. It seems to have a wider voltage range than usual, claiming 200-240V. The SAA approval number – SAA142576 and batch number are laser etched into the cap. The flexible cable assembly comes from Ningbo Qiaopu Electric Co. Ltd, with the plug sporting an ESO approval number (ESO120303) and the cable sporting an NSW Fair Trading number (NSW18298).

The unit utilizes the familiar rat-trap style spring-loaded clip mounting system, with this clip being exposed metal. The body itself appears to be coated in some plastic, as is common with such construction.

As with the other units, the front is a relatively plain looking, simple, unbranded aesthetic. The central diffuser element is practically flush with the front, and the trim is relatively thin and low-profile for a nice look. The bezel is, subjectively speaking, a little wider than the competitors.

As with some of the others, occasional minor imperfections have been spotted – in this case, it’s another instance of a small excess of glue residue at the edge. Nothing major to worry about.

Teardown

The first step in taking it apart was to remove the cable retainer, which proved to be a very effective device featuring a double row of jaws to make sure the cable wasn’t going anywhere. Once this was removed, the green cap easily unclipped to reveal the driver PCB.

The unit itself is possibly somewhat dated judging by the PCB date code of Week 37, 2015. The PCB is a double-sided PCB with solder resist and silk screen. It has adequate clearance around the primary terminals, to ensure that the wires do not contact components inside. Because it claims to be a double-insulated design, the PCB does not feature any clear isolation between primary and secondary. The input is protected by a fusible resistor inside heatshrink, and transient surge protection is provided by a MOV. A primary side filtering capacitor can be seen that is not safety rated, although it is unlikely to represent any hazard. Inductors are also present for RFI suppression requirements.

The heart of this unit is the Power Integrations Lytswitch-4 LYT4321E Primary-Side Constant-Current Controller with PFC. The controller offers similar features to other products on the market including a tight output regulation (<5%), TRIAC dimmable with leading and trailing edge dimmers to less than 5% output, startup <0.25s at full brightness, power factor >0.9 and up to 92% efficiency. One specific feature of this controller is the lack of requirement for a primary side electrolytic capacitor, which can be a point of failure, however, it appears that this design still has them (middle of the image). It appears to be dated Week 35 of 2015, corroborating the period in which the unit was manufactured.

This controller has a very unconventional package, which is bonded to a sheet of aluminium plate with a copper heat-spreader for a heatsink. Because of the possibility for heat buildup, having the excess LED array wire in contact with this heatsink seems unwise, although the choice of silicone insulation on those wires probably mitigates any risk of melting of insulation.

The capacitors used are made by Samxon, which is a brand of Man Yue Technology Holdings, a Hong Kong based company. The brand has been established for over a decade, and I have personally had no problems with their products, although there appear to be some isolated negative experiences with the unrelated GF series. The particular capacitors used in this product are the GT series, rated at 105 degrees C with a lifetime of 4,000 – 10,000 hours. This rating is appropriate for the expected conditions (e.g. at a temperature of 75 degrees C, expected lifetime is 32,000+ hours).

Looking at the underside, the PCB is marked UP-FDB09300-PCB REV.A2. The manufacturer could not be conclusively determined based on this information. The underside shows very good soldering in general, with adequate solder and mostly consistent mounting of SMD components. Terminal blocks with long slotted legs are well soldered with smooth shiny appearance. LED array cables were properly through-hole soldered, rather than tack soldered as with some of the lower-cost competitors. Some legs could have been trimmed shorter, however. Similar to the SAL Wave, there is a cut in the PCB near the primary terminals for layout reasons, so that the output from the fuse mounted in the corner can be routed back into the main board area with less risk of flashover.

Like the other products, the inside of the unit has a conical white plastic sheet diffuser to help better utilize the produced light. However, the Atom takes it one step further, by putting a similar plastic sheet on the MCPCB as well, even though it is shiny and coloured white. This shows that they are probably doing their best to eke out every last bit of performance from the LEDs.

Three screws secure the MCPCB to the heatsink at a middle radius. Unlike other products, the cable entry is from the side, which seems like a decent design choice. The MCPCB is marked with DL41-3-10W-12C2-2835LED-V01 and dated 20/08/2014 with some code 4.3401.09.02400. The board is made of 24 x 2835 SMD LEDs. A closer examination of the LEDs seem to show they may be slightly low on solder paste on the MCPCB, as their orientations are quite skewed in some cases, and some copper island remains exposed. This could potentially be a reliability issue with potentially reduced joint strength, especially if there wasn’t sufficient heat to reflow the solder paste correctly.

At this point, the manufacturer was determined to be Upshine Lighting, with this product being their DL41 series. They also seem somehow related to NBG LED Lighting.

Contrast enhancing the image, we can see that the MCPCB connection shows that it follows a similar connection pattern to the SAL Wave, having pairs of LEDs connected in series. This is potentially a bad design choice depending on how well matched the LEDs are, as it can result in current imbalance, causing unequal stress on paralleled partners, and premature LED failure. Its failure mode is possibly slightly more graceful though. The tradeoffs were explained in the SAL Wave article.

Further to that, the trace pattern is sub-optimal, as the traces are not covering the majority of the surface, as you would if you wanted to enhance the thermal transfer through the LED contacts to the MCPCB substrate. This can lead to higher temperatures, and uneven temperatures causing more current imbalance.

Further removing the MCPCB shows that a suitable amount of thermal grease was indeed applied to the board, enhancing the thermal contact with the heatsink, although its contact area appears to be biased towards one corner. This is probably reflective of inadequate or inconsistent torquing of the hold-down screws, and represents a sub-optimal condition which may cause LEDs on one side of the board to run slightly hotter.

Performance Testing

This section will detail some of the results of testing the downlight.

Subjective Opinion and Output Quality

The light started up rapidly, under a second under all circumstances which was impressive. The opalescent diffuser was quite adequate, resulting in a very smooth distribution of light.

The output colour temperature, as tested by DSLR camera and RAW file correction was measured as 2800K, slightly below the claimed 3000K rating. The output colour rendering was not significantly better or worse than its competitors, and seemed adequate for residential purposes. No flicker was observed, and no acoustic noise was observed while operating at full power. The light output appeared consistent with its claims.

Warm-Up Power Trend

This was tested with the unit powered on from cold, with measurements made with a Tektronix PA1000 Power Analyzer and line voltage regulated to 230v (+/- 1%).

Over a period of 20 minutes warm-up, the power consumption trend was the opposite of many of the tested products, instead, initially running below the rated power and increasing over time as the LED array warmed up. The trend also appeared slightly influenced by fluctuations in the line voltage. It appears to have reached just over 10.1W at the 20 minute mark, thus confirming its nameplate rating.

Power Consumption and Power Factor vs. Voltage

This was tested using a Variac to vary the line voltage, with results recorded with the Tektronix PA1000 Power Analyzer.

The downlight achieved a slightly positive slope in its regulated output above 165v, slightly increasing power consumption as the voltage increased. This is still well within the expected limits. Through the design range, the power factor was excellent, achieving >0.95 which makes it a good result for especially large commercial installations where power factor may be an important parameter.

Below 165v, the downlight could not maintain its full output, but still continued to shine down below 30v, resulting in an extremely wide operational voltage band. This means that the light is unlikely to be affected by brownouts, and even in severe power dips, may continue to offer some light.

Inrush Current

Inrush current was tested using the inrush current mode of the PA1000. The unit under test was toggled on and off ten times in succession, and the maximum positive and negative peak currents were recorded.

The maximum positive current was 2.248A, and the maximum negative current was recorded as -2.449A. This seems to imply good inrush current control, allowing for 12 units per circuit (very conservatively, as the duration of the inrush is not reported by the analyzer).

Ballast Output and Efficiency

Output testing was performed using the PA1000 taking sequential measurements of voltage and current for an approximate figure.

The output voltage was determined to be 36.80v with a 1.018 crest factor, making it essentially DC. No frequency was reported by the analyzer. The output current was 226.2mA with a 1.178 crest factor, implying some level of ripple, but not significant enough to cause flicker.

The output power was 8.32416W when cold, resulting in a calculated efficiency of 85.3%, which is reasonably good in general, but figures approaching 90% are more the norm.

Dimming

Only the Clipsal Universal Dimmer is recommended by Atom Lighting for use with the product, however, the datasheet for the IC used seems to imply wide dimming compatibility. The unit was tested with the usual suspects – namely the three dimmers that I have on hand.

Using the IKEA leading edge dimmer, the unit impressively consistently started up in under one second at all dimming levels. There was a slight acoustic noise from the dimmer and the unit, but not particularly noticeable. Dimming appeared to range from about 15% to 100% using this arrangement, with no noticeable flicker under ordinary circumstances. Rapid changing of dimming levels sometimes resulted in an “overshoot” and a step change in the output level within one second of stabilizing. The unit did appear to have some flicker when dimmed and subjected to excessive mains ripple signalling (a chance discovery).

When used with the troublesome Nixon Universal dimmer, the unit exhibited slightly less noise with a little shimmer when at ~75% of full output. Other than that, it was possible to get a wide range, well down to 1% output. At the minimum level, it took almost three seconds to start up.

With the DETA Trailing Edge Universal Dimmer, it was quiet and dimming was stable. The range was only down to about 15% minimum, and start-up times were slightly longer owing to the delay in this dimming unit.

It seems the Power Integrations Lytswitch-4 IC has very good dimmer compatibility, resulting in less troublesome dimming even with obscure and inexpensive dimmers as compared with some of their competitors.

Conclusion

The Atom Lighting AT9033 appears to be a product of Upshine Lighting, and seems a little dated judging from its internals.

It appears to perform well, with the Power Integrations Lytswitch-4 giving it very good power credentials (wide dimmer compatibility, quite reasonable efficiency, fast start-up, no acoustic noise at full power operation, wide operating voltage range, good power factor). The driver unit appears to be well built, but the use of Samxon capacitors with appropriate rating may divide some users depending on their prior experience with the brand. It achieves an 80lm/W efficacy, which is acceptable but not class-leading especially when 100+lm/W is where most products are now targeting or have already surpassed.

The unit seems to be let down when it comes to the mechanical construction where discrepancies in screw torquing for the MCPCB seem to have led to uneven contact, and the use of pairs of LEDs in parallel, put into series, which can cause current imbalance and possibly premature failure of LED chips leading to a cascaded failure condition. The MCPCB itself did not have the optimal copper foil trace design that would have optimized thermal heat flow from LED packages to the heatsink. The physical construction also showed signs of a small excess of glue on the diffuser.

Further drawbacks are seen when it comes to installation, as this unit is not IC rated and requires clearance from insulation which leads to additional work installation, heat loss through the downlight in the ceiling resulting in additional heating and cooling bills, and the potential for dust accumulation around the downlight which is unsealed as the roof space “breathes”. Another concern that remains is a lack of commitment on specifying lifetime and CRI in their datasheets.

As with all products tested, each of the products seem to have their own advantages and drawbacks. I hope this review has been informative and allows you to understand the important design concepts behind LED lighting design and assess which product best suits your needs.