The lighting industry has been undergoing a revolution thanks to LED technology, bringing with it ever higher efficiencies and better lifetime which are becoming more important both from an energy cost and environmental perspective. Home owners with older halogen based downlights often desire to change them with more energy efficient options, and sometimes, the plug-in retrofit options are not satisfactory resulting in short lifetimes, flickering, buzzing, unstable dimming or insufficient brightness. Likewise, new home builders are unlikely to opt for such older technology when faced with the dilemma of what fixtures to fit.
The solution to these problems are LED integrated fixtures, which are optimized for the characteristics of LEDs from the outset. Such fixtures are beginning to be inexpensive, and boast longer lifetimes than their retrofit counterparts. They come in a variety of configurations, with various insulation clearance requirements. These are designed to be permanently installed by a qualified electrician and run for their lifetime without any maintenance, after which, they are completely replaced.
HPM, one of the most recognized brands in the Australian electrical industry, have recently introduced the DLI series of LED integrated downlight. In this posting, we review and teardown the HPM DLI9002 7W Cool White LED downlight, featuring an 820 lumen output with 80+ CRI, IP44 rating, IC-F abutted and covered insulation contact rating and a 3-year warranty. It boasts one of the highest luminous efficacy values on the market, reaching 117 lumens per watt, and is a bespoke product of HPM. This makes the unit a very attractive proposition for replacement of existing downlights and new installations alike.
The unit comes in a small full-colour cardboard box with an anti-tamper security seal. The box clearly shows the product and its salient features, making it easy to compare the product at a glance.
A full run-down of the features is provided on one of the sides, along with installation depth and cut-out requirements. I find this very handy that it is clearly stated on the outside of the box, so prospective purchasers can make a decision without having to look up datasheets.
The 70mm depth allows installation in small ceiling cavities, along with the abutted and covered insulation contact rating, means easy installation with less worries as to insulation clearance and fire risk. Its IP44 rating makes it suitable for undercover outdoor installation, which is also relatively unique. The driver integrated within the device is suitable with “most quality dimmers”, indicating support for trailing-edge phase control, and the light comes pre-fitted with a 1m flex cord for simple installation, especially where the circuit contains GPO points already. It also claims to have 2,500v surge protection, which seems interesting, as normally surge protection ratings are provided in joules absorbed rather than surge voltage. The test waveform for this rating is not specified.
The other side more clearly states the specifications, which includes an 820 lumen output (typical) with 80+ CRI. The light sports an impressive 35,000 hour rated lifetime, and 117 lumens per watt luminous efficacy. Assuming a 24/7 application, the lifetime corresponds to around 4-years of continuous use. In an average residential use of 5 hours per day, it corresponds to over 19 years of service, making it very suitable for a maintenance-free lighting solution.
Enclosed is the downlight unit itself, with a plastic outer body with the driver atop the LED array and heatsink. This arrangement has an air gap, which reduces the temperature of the LED driver to ensure a longer lifetime – a good arrangement. The downlight is marked with the regulatory compliance mark, and is double-insulated. The flex comes pre-fitted to the downlight, with a plastic “clip in” cable retainer that needs to be carefully pried open if you wish to hard-wire it into a circuit.
The light fits recessed within a 90mm cut-out thanks to two “rat-trap” style clips which makes installation quick and simple. Also included is a white rubber gasket, which helps to seal out dust and draughts, which is a nice touch. The unit is quite light, and with its plastic exterior, its heatsinking is not obvious. Read along to the teardown to find out more.
From the front side, the luminaire sports a clean, uncomplicated appearance with no branding. This is aesthetically neat and desirable. The unit has a matte white surrounding trim with a recessed opalescent diffuser that creates a claimed 85 degree beam. When powered on, the light emission is slightly non-uniform with a slight “halo” effect, but in use, this is not apparent.
A very close and detailed look shows some slight silicone residue adhesive around the diffuser. While not particularly obvious and eye catching, it is an area where the build quality could be improved.
To round out the package, an instruction manual with full specifications and warranty conditions leaflet are also provided.
Some of the information gleaned includes the reduction in lifetime due to insulation abutting and covering the fixture, which reduces the lifetime to 20,000 hours versus 35,000 hours as installed with default clearances. The unit is not suitable for use with loose-fill insulation and requires clearance from structural members of >5mm and combustible elements of >30mm. The unit is designed for operation up to 45 degrees C ambient temperatures.
The leaflet also details the recommended compatible dimmers, maximum number of units and their dimming range, which is very useful information. For reference, at the time of publication, this list includes:
|Dimmer||Maximum Units||Dimming Range|
|Legrand EM400A3P, AR400A3P||12||1 – 100%|
|Legrand EM400A2P||12||9 – 97%|
|Legrand EM400T, AR400T||12||7 – 100%|
|HPM 400T||12||7 – 100%|
|Legrand EM250T||8||7 – 100%|
|HPM 250T||8||7 – 100%|
|HPM 700T||21||13 – 100%|
|HPM 1000T||30||16 – 100%|
We start our analysis of the insides of the unit by first taking apart the driver. The driver comes apart by removing the cable retainer and prying along the edge of the plastic housing. The mains cable can be seen pre-fitted to the terminals with bootlaced ends. A slight stripping of the screw head is seen on the live terminal.
The driver is relatively simple, featuring a minimum of components. On the input side, a fusible resistor is fitted to maintain safety, with inductors, capacitors and (what appears to be) a MOV to provide transient filtering of the mains and RFI suppression. A single channel output is wired out to the LEDs in the enclosure underneath.
The switching circuit appears to utilize a single high-voltage Silan Microelectronics SVF7N65F F-cell N-channel MOSFET, with the output of the switching circuit filtered by capacitors and a toroidal inductor, which should reduce any flicker and variations in current.
The capacitors used are Aishi RS-series capacitors, which are generally reliable in my experience with teardowns of other electronic ballasts. The RS-series is rated for a 4,000 to 10,000 hour lifetime at 105 degrees C, which implies a lifetime of 16,000 hours to 40,000 hours at 85 degrees C assuming full voltage/ripple loading. As a result, it can be seen that the choice of capacitors (a failure prone component) is quite appropriate for the intended lifetime.
On the underside, we can see a bridge rectifier on the left, and the switching controller, a Monolitihic Power Systems MP4056 TRIAC Dimmable, non-isolated offline LED controller with active PFC with integrated over current and voltage protection. Interestingly, according to the controller datasheet, it also supports dimming with leading-edge dimmers contrary to the supplied information.
Because the driver topology is non-isolated, the output is considered “at mains potential”. As a result, the driver doesn’t have a clear primary to secondary isolation, as the output is not intended to be accessible because of the double-insulated nature of the product.
Eagle eyed viewers may have spotted a potential issue with the PCB above –
A close look at the mains terminal connector blocks show that they are only supported by the two solder points underneath the PCB. However, these points are not particularly well soldered – the top two clearly show voids around the terminal legs, which may be due to improper soldering temperature or surface finish resulting in poor wetting of the solder. The bottom two joints show repair work, on the left especially with stray solder and a different shape, although some voiding is evident on the left. The right bottom joint has a slight solder blob on the left side which indicates uneven cooling.
This is a potential concern despite working well initially, as it could be a cause for failure in the future once subjected to numerous thermal cycles, vibration and cable disturbances.
The cause of these problems may be due to the torque applied to the terminal connector block during assembly, which may cause the terminals to break way from the solder joints. A better terminal block design with more support may be appropriate – and I would not recommend hard-wiring this particular unit to avoid potential damage to the terminal block.
This unit is actually the second unit, which was a replacement for the first unit which had intermittent contact issues, where the terminal block actually “falls out” of the PCB when pressure is applied. The PCB also shows signs of a split, which could imply a little too much force applied to the PCB during assembly. As this pattern is seen on both samples, it seems the design might need a little rework to avoid this. That being said, repairing such joints is not particularly difficult.
To access the LEDs requires destructively prying at the front diffuser until the adhesive lets go, and the cover pops out of its slot. The resulting design has 28 x 2835 type SMD LEDs, in a series string on the MCPCB. This design is a good design because it avoids parallel LEDs which can have uneven current sharing causing premature failures, however, a single open circuit failure will render the whole array inoperative. This is probably a good thing, as failure modes on parallel arrays can lead to changes in colour, and seizure inducing flickering.
While not immediately obvious, the MCPCB is actually mounted on a metallic heatsink shaped into a bowl. The edge of the bowl is covered with a white plastic strip to improve the diffusion and light extraction for better efficiency.
Removing the LED array screws reveals the array is thermally connected to the heatsink with a thin layer of thermally conductive grease. The amount applied is perfect, as a thin layer provides the best thermal conductivity and many other solutions have too much.
On the whole, the design has many merits. It sports an advanced integrated circuit based driver with power factor correction and the necessary components to filter RFI and mains transients, appropriately specified capacitors, highly efficient LEDs in a single series string arrangement, and good thermal bonding arrangement. The only downside was the terminal block arrangement, which could be improved to improve reliability and safety.
In this segment, we review some of the results of testing the device. Keep in mind that I am a hobbyist, and not an accredited test lab, so the results are provided in good faith that it would be useful, but their accuracy should not be considered as absolute.
Power consumption was tested using a variac-adjusted input voltage set to 230v (+/- 1%) to power the fixture. Measurements were made using a Tektronix PA1000 Power Analyzer.
Over a 20 minute run-up from cold, the fixture consumes somewhere in the ballpark of 7.08W when warm, initially drawing about 7.28W. This characteristic is very common of LED lights, as the voltage drop of the LED changes as a function of temperature as it warms up. The result confirms the rated 7W nameplate consumption.
Voltage Operating Range and Power Factor
The operating voltage range and power factor was likewise determined using a Tektronix PA1000 Power Analyzer with a variac used to vary the voltage.
The nameplate rating voltage range is coloured in green, whereas the tolerance voltage range is coloured in yellow. Testing showed that the unit operated satisfactorily in a wide range of voltages outside even the design region. Active regulation was seen, resulting in the consumed power remaining within 7W to 7.25W within the extended operating range. The unit ceased to regulate correctly below an input voltage of 90v AC, meaning even the deepest of brownouts are unlikely to yield any appreciable dip in brightness. This is similar to other LED products on the market, and implies that the array requires a fair amount of voltage to drive, as expected from the teardown.
It survived operation at 278v AC, and exhibited very good (>0.9) power factor within the intended range. While this is not as important for residential consumers, it reduces reactive power demand which reduces transmission power wastage.
Driver Output and Efficiency
The driver output was determined using the Tektronix PA1000 Power Analyzer when the unit was operated off an isolated supply for safety, as the driver is not isolated. Due to the high expected output voltage, flicker and current tests using an oscilloscope were not performed.
The output voltage of the driver was 80.50v DC, with a voltage varying from 79.42 to 81.34v, indicating a very “shallow” variation in output voltage. The negative lead was cut to intercept the current to the LEDs, which showed a reading of 79.15mA, ranging from 65.26mA to 91.26mA, indicating a wider range of currents. However, as the current did not fall to zero, flicker should not be apparent.
Based on these full-drive results, the actual power delivered to the LED array is 6.371575W, for a driver efficiency of 89.99% which is very respectable. The upside of high driver efficiency is also less produced heat, and longer lifetime.
Inrush Current Test
Inrush current testing was performed using the Tektronix PA1000 Power Analyzer in inrush current mode. The unit was power cycled 10 times at random within two minutes with no time for cooling of soft-start components (if any), and the peak positive and negative currents were recorded.
In the test, the peak positive inrush current was recorded as 2.024A, and the peak negative inrush current was recorded as -2.071A. This indicates a relatively tame inrush current of approximately 2A. This implies that it should be safe to install about 24 units in a single 16A circuit owing to the circuit breaker time characteristics, and more may indeed be possible.
As the only spectrometer I have access to is an old, beaten-up Ocean Optics USB2000, it seems that it is starting to fail badly especially in regards to blue sensitivity, and as a result, I was not able to get reliable results even after compensation with sunlight measurements. The reported results are likely to be quite far off from the truth, and should only be compared with other results on this site. The results are not absolute values and should not be used for comparison with other results.
The spectrum result seems to show quite a lack of blue, likely because of sensitivity problems with the spectrometer, even after compensation with sunlight based readings. As a result, the measured CCT using this means underestimates the CCT. The CRI also seems low on this particular test, but is likely again due to the test equipment. Compared to other ~80 CRI globes, the CRI readings seem to be a bit on the low side, and in my experience in viewing photographs under the light, the CRI is probably a little closer to 80 rather than 85 of more expensive LEDs.
Colour temperature measurement by using a camera and white balance correction in Lightroom yielded a colour temperature measurement of 3900K, which is consistent with the advertised figures.
On application of full power, the light starts up with an imperceptible delay. Under low dimming levels, the light does take about a second to ramp up to a stable brightness, but all in all, it is par for the course.
As I don’t have anything so great as an integrating sphere, the lumen output could not be determined. Subjectively speaking, it appeared to be consistent with the package claims.
Unfortunately, as I don’t have any of the dimmers in the compatible list, I tried evaluating the unit on its own on a Nixon universal dimmer and an IKEA Dimma leading-edge dimmer (contrary to instructions).
In my experience with the universal dimmer, the dimming range of the HPM DLI9002 was impressive, reaching a very dim level at the minimum level and reliably starting up even at low dimming levels. At low dimming levels, the light was stable and flicker was not visually apparent. No audible noise was experienced. However, at high dimming levels, some instability in the form of flickering was exhibited, and this is likely due to not meeting the minimum load requirement of the dimmer and not a fault of the device itself.
Operating it contrary to instructions on a leading-edge dimmer did not result in any failure, although some acoustic noise was experienced (as expected). Dimming range was as expected, and instability at full brightness was also experienced. I would not recommend attempting this configuration as it will void your warranty and could theoretically result in dimmer failure.
It is recommended to use the listed dimmers and comply with the minimum load requirements to ensure such instability is not experienced.
HPM’s DLI9002 provides a very compelling package of features in a bespoke design. It is a relatively compact unit which can be installed in virtually any space thanks to its abutted and covered insulation contact rating and IP44 rating. It features a long rated lifetime of 35,000 hours, with a high luminous efficacy of 117 lumens/watt, making it more efficient than most LED products on the market. Its 820 lumen output is comparable to a 52w incandescent globe. It has a simple, clean design aesthetic and features simple installation. It even boasts a wide operating voltage, with great power factor results, no flicker, and a driver with just about 90% efficiency. With low inrush and operating currents, and a phase control dimming compatible driver, with compatible dimmers it achieves a wide 1 – 100% dimming range. Its internal thermal design shows excellent thermal contact, separation between driver and LED, appropriately rated capacitors and single-series LED array which prevents premature failure through current imbalance. Best of all, it is backed by a three year warranty, while remaining similarly priced with competitors.
While there is much praise for the unit, I did have one major concern in regards to the terminal block soldering which appeared to be sub-par and prone to connection damage in the course of re-wiring. I also saw a small amount of adhesive residue, which may not be apparent in normal situations, but could be improved. The CRI also appeared to be on the low side of 80, which may not meet all users’ expectations.
Update: After acquiring a DETA branded 150W “Universal” trailing-edge dimmer with a minimum requirement of load of 4W, Model 6031, it was found that full dimming range with smooth linear response and quiet operation was achievable despite it not being one of the recommended dimmers. Start-up is delayed by almost a second due to slow-ramp up of the dimmer and the stabilization of the driver circuitry. Just goes to show that sometimes it takes the right equipment to get the job done.