The LED downlight reviews continue, with another anonymous donation for the review challenge. This time, we’re looking at the Sunny Australia Lighting (better known as SAL) Wave S9065 DL (Daylight 6000K) 10W LED downlight. SAL is a major Australian distributor of lighting products, and this particular product is said to be one of the most popular amongst LED downlights. Lets see how it fares when subjected to a full teardown analysis.
Unboxing and Features
The unit is packed in a glossy compact cardboard box, featuring a pictorial image of the product and some of its most important features summarized as icons.
One side features a bullet point listing of specifications, although quite vague and generalized. This particular unit has a 6000K “daylight” colour temperature, and offers 900 lumens with a 10W power consumption, for a 90 lumens/Watt efficacy. Notably, the efficacy varies as a function of colour temperature (due to LED efficiency constraints), thus their 4000K “cool white” achieves an output of 850lm (85lm/W) and their 3000K “warm white” achieves 800lm (80lm/W). This is a little better than some others, but far from the 110+lm/W achieved by recent products such as the HPM DLI9002. It has an IP44 “weatherproof” rating suitable for installation in protected outdoor environments. It claims to be trailing-edge dimmable, comes with 1.2m of prefitted flexible cable, and is double-insulated.
It carries the insulation contact IC-mark, approved for abutting and covering the luminaire. The unit carries the regulatory compliance mark for Australia and is Made in China.
The unit requires a 92mm cut-out for installation and has a 59mm height. It also has a 90 degree beam angle and >80 CRI (although the exact figure is not stated).
Inside the unit, there is a single-sheet warranty statement and an installation instruction sheet. A spectral graph is provided, which is somewhat unusual, but the unit appears to lack specifications surrounding lifetime where no figure is available. A three-year warranty is offered, but in the fine print, it claims this warranty is only extended for residential operation up to ten hours a day, else just one year warranty for commercial application. While most residential consumers are unlikely to have such high duty cycles, it appears they are only willing to stand by their product for ~10,950 hours at the most. Considering many other manufacturers don’t have this particular limitation in their warranty statements, this seems somewhat surprising.
Interestingly, their website appears to have an updated installation instruction sheet which claims compliance with IC-F mark (abutted and covered) and claim that side clearance to building element of zero millimeters. Other manufacturers usually claim some nominal distance rather than zero, but I suppose it’s unlikely to indicate any significant differences in heat output, especially given the (relatively) low power consumption. Clearance to combustible elements or elements above fitting are not specified. Of course, the impact to lifetime from covering the luminaire is not stated.
The unit itself shares some visual similarities with the DETA unit reviewed earlier in having a coloured screw-down cable retainer and moulded logo, but I suspect it is more a likelihood that the DETA unit had imitated the look of the SAL unit instead. The body of the unit has the connection order (neutral on the left, which is a little uncommon) and acceptable wire gauge near the terminal blocks which is convenient.
The rest of the details are laser etched onto the side of the luminaire, with the IC mark label applied to the outside. As with other units, a spring-loaded “rat trap” style mounting is used. The lower section of the body does feel more like metal, with a thinner almost-paint-like coating on the aluminium which is probably better for heat dissipation than a thicker layer of plastic in some other products. The rear of the unit differs from the DETA in that it literally has a wave in the casing, which makes it a bit difficult to stand on its back.
The driver and the LED are connected in this unit, without any air gap or clear separation. This may mean that the driver is operating at higher temperature, being influenced by the waste heat from the LED array, potentially shortening its lifetime.
The unit does not include a gasket, which means that the recess hole is not “sealed”. This can lead to some heat loss through air movement around the edges, and the breathing of the roof-space through these holes could lead to some accumulation of dust around the fittings.
The included flex and plug appear to be from Yunbiao Electronics, with appropriate electrical safety numbers moulded in – ESO130205 on the plug and ESO120686 on the cable.
The front of the unit appears no different to the majority of the units on the market with a simple plain white surround and almost-flush opalescent diffuser. An extra cost option of a satin nickel ring is available.
Now, we arrive at the “fun part” of taking it apart and looking inside. First, it was noted that the cable retention mechanism works well, unlike that of the DETA, and positively grips the included flex cable. Once removing the screw for the cable grip, another screw needed to be removed to allow the rear cover to be removed. The rear cover is a simple plastic cap, with nothing special within it unlike the DETA.
The main PCB is a relatively high-quality double-sided PCB with solder resist and silkscreening. It has marked upon it KADA-059401 2015-05-20 YM1505068 which implies that the product is produced specially by Kada LED for SAL.
The board itself features good clearance around the terminal blocks, so that the mains wires won’t “poke” into the components like the DETA. For safety, we can see the use of a fully enclosed 500mA slow-blow fuse, and mains-safety rated mains capacitor for transient suppression and inductors for RFI supression. There appears to be no primary side MOV for surge protection, which could be a negative especially in areas where there are high incidences of surges. While switching converters in general are not too sensitive to power quality issues, having a MOV would offer peace of mind especially for fixed installation where easy consumer replacement may not be an option and labour is expensive.
No real “isolation” boundary can be seen between primary and secondary, despite the routed channels in the PCB. This is because the routed channels are more for the placement of L1 (inductor) and the wires connecting it due to the way the board is laid out. This is not a big issue, as the unit claims to be double-insulated, thus nobody is expected to be in contact with the output.
The heart of the unit (U1) is the iWatt (now Dialog Semiconductor) iW3614 which appears to be an older variant of the iW3688 used in the DETA. This particular controller appears to be slightly less featureful, offering compatibility with leading and trailing edge dimmers (no support for digital dimmers), but otherwise offering the same features of wide dimming range (1-100%), typical 85% efficiency without dimmer, tight LED current regulation (5%), and a variety of safety protections that operate on a pulse by pulse basis. However, this controller is somewhat less flexible in its configurations, using an isolated flyback converter topology. As a result, it seems that this controller could have had an isolated secondary, provided the PCB was laid out with appropriate clearances.
The transistors, of which there are two, appear to be both Silan Microelectronics SVF4N65K 4A high voltage (650v) N-channel F-cell MOSFETs.
The capacitors, however, are somewhat disappointing. All the capacitors on the board are Pchicon branded, from Yiyang Pencheng Technology Development Co. Ltd. This particular brand appears not to be particularly well known, and in past experience, lower cost capacitors from unknown companies are less likely to have high reliability compared with their higher quality Japanese counterparts.
Regardless, this unit has three such electrolytic capacitors – a primary side 6.8uF 400v CD11GAL-series capacitor, and two secondary side capacitors, a 47uF 50v and 100uF 63v RF-series capacitor. All capacitors are rated for 105 degrees Celsius operation. The CD11GAL series is rated for a 6,000h load life, with the RF series rated for 3,000-7,000h load life. With this in mind, the capacitors are within the expected range for a ~24,000+ hour lifetime, although I would still be cautious of the quality of these capacitors given their unknown reputation.
The underside of the PCB has various surface mount diodes, resistors and capacitors, as well as a bridge rectifier. A QC label is also neatly placed on the underside. The soldering quality is highly consistent for the components, and the terminals are long slotted types with lots of soldering area for support. However, it does seem that the amount of solder may be a little on the low side with the solder not fully flowing through the vias, although there are no major voids that can be seen.
The LED array is connected by two silicone-rubber covered wires, which are tack soldered onto the bottom. One contact has a slight excess of solder, but otherwise, the quality of the connection is very acceptable.
Removing the front diffuser exposes the LED array. The unit follows a similar structure to others seen before, featuring a loose white plastic conical insert to improve light output. This particular cone was not perfectly concentric, suggesting a small dimensional discrepancy, and this “uneven” diffuser may cause slight aberrations in beam distribution.
Removing the diffuser exposes the “heatsink” which is shaped like a dog-bowl. The MCPCB is labelled AL-4C-10W-20S-2835-280mA, which gives us some information, although it is potentially a little misleading. The connections to the MCPCB are well soldered to the board. Sure enough, there are 20 LEDs on the board, and they appear to be 2835 SMD type (commonly 0.2W each, some up to 0.5W each), but the arrangement is a little different than would be suggested by 20S.
A closer look at the patterning of the MCPCB exposes the fact that it is a design where pairs of LEDs are put into series. This arrangement is a little different than just distinct single-series or multiple parallel strings. For explanation, I will have to resort to some circuit diagrams which can get a little technical.
This arrangement is a fairly simple all-series configuration. This configuration has been used by the products reviewed in the past, and basically has one path throughout the diode array, thus each and every LED is subject to the same current. This results in a mostly even heating and stress of the LEDs, equal brightness from each LED, and generally less problems although should any LED fail open circuit due to a failure of the LED or its connections, the whole array fails in one go. This is not a particularly big disadvantage, as we will see later. I personally find this configuration preferable in the vast majority of cases, although for various manufacturing constraints (e.g. limiting output voltage), it may not be the configuration employed.
This is another possible configuration of two parallel strings. Something similar to this has been seen in prior reviews of retrofit LED bulbs and basically consists of the whole array of LEDs divided into a number of independent series string, with strings connected in parallel. This arrangement is also fairly common, but has problems with current balance between the two strings. As “no two components are the same” and the temperature each LED experiences is not the same, a potential for current share imbalance exists because the voltage drop of each string may not be equalized. Unless the LEDs are well matched or run significantly below their rated current, such an arrangement may lead to less than optimal lifetime, especially as LEDs age and their characteristics drift. In this arrangement, normally a single current driver provides a fixed current of about twice the current handled by a single string, and it is assumed that both strings will share the current equally enough that no one string is overtaxed. Where the current is not balanced, one string will likely heat up more as it takes the lions share of current, resulting in higher stress and premature failure, especially as warmer LEDs will reduce their voltage drop and increase its current share, thus resulting in a potential runaway feedback loop.
In reality, this may work initially, but over time failures can result. Where one LED fails open circuit, half the array suddenly becomes disconnected, resulting in the provided current being forced through the remaining half. This can result in the current driver flickering and flashing if it exceeds its maximum output voltage, but more likely, will result in dimmer output, a shift to a more blue output due to overdriving the LED, significantly increased heating in the remaining functioning LEDs leading to rapid failure of the remaining LEDs. This arrangement is not optimal, but is used quite frequently.
The arrangement used in the SAL Wave looks like this. I have mentioned that it is a possible configuration in the past, but I did not come across a product with such a configuration until now. This arrangement has pairs of LEDs in parallel, put into a single series string. It maintains the reduced output voltage requirement of the above configuration, but has a few slight differences in the way it operates.
Because it is essentially a series arrangement of parallel elements, at each pair of diodes, the “sharing” of the current can be different, rather than having half the array at one current, with the remainder in the other. This could potentially lead to more imbalance events across the array and potential stress on individual LEDs, because with the two series strings in parallel arrangement, there is a statistical probability that the average of a large population will converge towards the mean – in other words, by having the sharing occur on a ten-diode voltage drop basis, there’s a better chance that if there are any particularly low or high voltage drops, that it would average out across a population of ten rather than of two.
However, for this particular arrangement, its failure mode could be more graceful. Any one LED failure only takes out that one LED and forces its one partner to run at a higher current. Where the current level is somewhat tolerable to the remaining partner LED, the array will continue to function with a small reduction in output rather than catastrophic loss of half of the array. However, regardless, any LED chip failure is the beginning of a downward spiral towards total failure as any partially operative array is unlikely to be a stable configuration, and thus, the failure behaviour is probably not as important as the potential for individual LEDs to be overstressed.
How well such an arrangement fares in practical terms depends on how well matched the LEDs are, both at the beginning of life and throughout the lifetime of the unit. It’s almost impossible to predict this with any certainty, at least, with the tools available to me at the present time.
The pattern also has excess area to improve heat transfer between the LEDs and the MCPCB, and ultimately the heatsink. One LED on the rightmost side appears to have some damage from solder splash onto the upper left hand corner, while not “fatal”, seems to imply the process of soldering wires to the MCPCB may affect the LEDs.
The MCPCB is held onto the heatsink with three screws about two-thirds the way towards the outer edge. Further disassembly revealed that the MCPCB is screwed onto the heatsink with no thermal interface material at all. This is likely to result in higher thermal resistance than otherwise necessary, meaning that the LED array operates at a higher temperature and will have a shorter lifetime than absolutely necessary as a result as the heat is not passing through to the heatsink and out of the luminaire as efficiently as it could.
Applying some thermal grease at the factory would improve the thermal resistance by filling in voids between the metal surfaces, which provides further paths for heat to travel. This is common practice with other products, and isn’t particularly expensive. It’s a wonder why this particular unit was assembled without it.
Performance Testing Results
In the following section, I will discuss some of the obtained test results and subjective impressions of the SAL Wave downlight.
Impressions and Output Quality
The output from the light was not quite evenly diffused, with a small “patterning” discernible from the outside. The colour temperature advertised was 6000K, although when tested with a camera and RAW-based white balance correction, the figure obtained was 5150K, a little less than expected. The output colour rendering quality appeared slightly, but not significantly, better due to the higher colour temperature.
When started up directly on the mains, a short delay of less than one second was evident, with the output stable and flicker free to the eyes. Output intensity seemed to be at the level expected for the claimed lumens. Acoustically, the downlight had no appreciable noise.
The physical build quality of the downlight seemed quite solid, and the thinner “coating” on the exterior of the heatsink seems to provide a better thermal conductivity and improves heat dissipation. The curved design of the rear does make placing the unit with the driver side down a little more difficult (e.g. during installation to prevent scratching the front surface).
A 20 minute warm-up test was conducted with line voltage regulated to 230v +/- 1% using the Tektronix PA1000 Power Analyzer.
The power trend shows a decreasing power consumption as the lamp warms, as is normal with LED products, reaching under 10W at 9.9W at 20 minutes.
Power and Power Factor vs Voltage
The power and power factor was recorded with the Tektronix PA1000 while varying the input voltage using a Variac.
Within a wide range of voltages (down to 130V), the downlight maintained a nearly constant power consumption just below 10W. Power factor was reasonably high, at 0.85 at the high end of the acceptable voltage range, and reaching 0.92 at the low end of the range. This isn’t the best result, however, is quite acceptable especially as power factor is more of a concern for larger commercial installations. No harm occurred with 277V applied to the downlight.
Below 130V, the driver appeared to operate in a somewhat less regulated regime, with increased power consumption to a peak of close to 12W, followed by a nearly linear decrease down to 18V when the light extinguished. This is an extremely wide voltage operating range, beyond that which is necessary for proper operation in Australia, and also means that it is likely that the light will maintain full brightness even in brown-out conditions.
The Tektronix PA1000’s Inrush Current mode was used to gauge the inrush current with the mains switched ten times in succession.
Peak positive current was recorded as 17.285A, and peak negative current at -17.234A. This is a relatively high reading, but unfortunately, the PA1000 does not provide the inrush duration, thus its full significance can’t be accurately established. However, it appears that the lamp may not have as effective soft-start mechanisms as other products on the market. This may be particularly important for dimmer applications where high inrush can cause stress on the unit and where many units are to be installed under the control of a single switch or on a single circuit.
Output Voltage/Current & Efficiency
Current driver output parameters were measured using the PA1000.
The output voltage was determined to be 31.30v with a 1.003 crest factor, with output current at 264.0mA with a crest factor of 1.009. This implies an extremely smooth output from the current driver with very little current variation. The output power is 8.2632W, which represents a conversion efficiency of approximately 83.5%, which is similar to that expected based on the datasheet. This is not a particularly high figure compared to others, where figures closer to 90% are becoming normal.
Aside from specifying trailing-edge dimming from the manufacturer, no list of compatible dimmers was available. Dimming was tested with a variety of not-so-popular dimmers which I have in my possession.
Using the IKEA Dimma lead, which is a leading-edge dimmer (not strictly compatible according to the manufacturer, but compatible according to the controller datasheet), it was found that dimming only operated over a very limited range (~60-100%) with significant noise from the dimmer (possibly due to high inrush currents at turn-on) and flashing at low levels.
On a Nixon Universal dimmer, it was found that the dimming range was good, although highly non-linear. There were some instabilities in light level at the lowest level, and an occasional “flash” of brighter light when turning down from higher brightness. Turning on at the lowest brightness resulted in about a five second process where the light will flash several times before settling to a more stable output. This might be due to the minimum 10VA load requirement, although this light does theoretically meet that (or almost).
Using a DETA Universal 150VA dimmer, it was found that a stable, smooth and quiet dimming operation could be achieved with full range of control. However, it seemed that due to the “soft start” of the dimmer mechanism itself, the time to power on had lengthened to about two seconds. I suppose this illustrates the need to match the dimmer to the load and purchase more modern dimmers with lower minimum load requirements suited for LED applications.
The SAL Wave S9065 DL while superficially sharing some similarities with the DETA, is a completely different beast, with a fairly sturdy build quality and simple aesthetic. It offers trailing-edge dimming compatibility, IC abutted and covered (IC-F according to the latest data) rating, with IP44 rating suitable for protected outdoor installation, and comes pre-fitted with a 1.2m flexible cord. The daylight edition offers 90 lumens/watt, down to 80 lumens/watt for warm white, which is respectable, although not class-leading.
When closely analyzed, it appears to be a product of Kada LED, with a quality double-sided PCB and good soldering quality. The design dating from 2015 features an older iWatt/Dialog Semiconductor controller. As a result, it wasn’t able to reach as high of an efficiency and power factor as some of its competitors. Its dimmer compatibility seemed slightly worse as a result.
The unit uses Pchicon capacitors, a relatively unknown brand which deserves some caution despite its mostly appropriate manufacturer claimed lifetime ratings. The unit was also found to lack surge transient protective MOVs and thermal interface material between its MCPCB containing the LEDs and the heatsink which can compromise durability and lifetime. The arrangement of the LEDs in pairs connected in series also appears unique to the product, although it potentially carries a higher risk of overstressing individual LEDs due to current imbalance. Slight mis-shaping of the internal reflective card and damage to an LED unit from solder splash was also observed.
At this stage, it seems every LED product has its own strengths and caveats, and the SAL Wave S9065 is no different. While there were some positive elements, there were also areas where cost-cutting seemed to be apparent and could affect the lifetime of the unit. It’s not a particularly terrible unit, but it’s not a particularly great one either. One potential concern surrounds the fact that no lifetime figures are quoted for the unit, and the three-year warranty comes with a 10 hour/day limitation, otherwise it is only warranted for one year.
After completing the review, I couldn’t resist the urge to be a handy-person and improve the product by applying some thermal grease to the MCPCB, which (subjectively) seems to have made the heatsink somewhat warmer in operation, indicating improved thermal conductivity. Just me, being me, I suppose :).