This week, I was very lucky to receive a rather large and heavy box. (No, It wasn’t pizza …) It was “unlovingly” left outside my front door by the postman. I wonder what it could be?
It says Philips on the front … so I think I know what it is … lets check the product label …
Aha! It’s a Philips GreenPerform Highbay LED luminaire, model BY619P LED100/NW PSU WB. This unit is rated for 220-240V AC operation at 50/60Hz, and is rated at 110W and uses 18 x 6.2W (totalling 111.6W) LEDs. The NW indicates neutral white colour temperature, PSU indicates a basic power-supply unit current driver (no dimming) and WB indicates a wide-beam optic. It’s packaged with one unit per carton, and is Made in China.
It seems these units are used in warehouse and commercial lighting retrofits, and are intended to replace Halogen, High Pressure Sodium and Metal-Halide lamps, commonly seen with 250W, 400W and 500W ratings. This particular unit was donated by an anonymous person for research purposes, and was pretty much the only thing I played with over the weekend – many thanks!
The luminaire was packaged nicely in styrofoam ends, but taking the unit out proved a little more challenging than I first expected. Part of the reason was because of its size and weight – it weighs almost 6kg … and I haven’t been doing my exercise!
The design of the light is quite unique. One of the first things you will notice is the large size and round shape with numerous fins. Given that it’s pretty much all metal (it seems to be aluminium), it has significant thermal mass and forms a good heatsink to keep the LEDs cool which allows them to be more efficient and long-lived. The wide channels are not conducive to dust collection, and the lack of any failure-prone fans are a big feature.
The main LED panel can be seen through a thick and sturdy polycarbonate window. This is held in place by four large slotted screws with rubber grommets. The LED panel consists of 20 positions, but only 18 are used. The other two have been bridged over with a surface mount resistor/link. Each position houses a single LED chip with four emitters. There is also an internal plastic lens sheet that defines the radiation angle (and is likely to be changed depending on your choice of wide-beam, narrow-beam or high rack optic).
Another feature you will note is the IP65 ingress protection rating, which makes the light mostly waterproof. You can clean it off by using a hose and expect no problems at all. Despite this, the luminaire is specified for indoor usage only.
From the rear, we can see the heatsink fin design. The LED emitter metal core printed circuit board (MCPCB) seems to sit on a thick bed of aluminium which has thick “ridges” to transfer heat across the whole area. Fins extend into the ridges to carry the heat away.
A “bridge” shaped piece is installed over the circular portion and contains the current driver electronics. By isolating it from the LEDs themselves, the unit is less affected by the heat and should have a better lifetime. Clever design.
The product nameplate label and specifications are visible on the back. It also provides additional information that the power factor rating is a high 0.95, with a current consumption of 0.53A. This is good news for commercial businesses as they often are billed for reactive power and need to keep their power factors as close to unity (1) as possible. High power factor lamps reduce the demand on power factor correction banks.
The unit also states that it is fire-rated as well. The flexibility in the mounting of the unit also becomes apparent, with a threaded rear pole mount clearly visible. This can be locked into place with a securing screw from the side.
If that doesn’t appeal, an adapter unit is available to turn that rear mounting thread into a hook mount which probably suits most warehouse “chain style” hanging. Otherwise, there are threaded screwholes either side which accommodate a bracket mount. The unit can be mounted in any direction – which isn’t something that can be said about all LED lights due to the thermal design. The mounting kits, however, are to be separately ordered.
The branding, and the way the fins taper into the centre can also be seen in the shot. The unit itself is pretty attractive – it’s got a very low profile which makes it less conspicuous.
In order to meet the IP rating, special design criteria must be met. To keep the water out, the current driver cable enters through a Weyer (Shanghai) branded water-proof cable grommet.
Special rubberized cable has to be used as well, in this case, the leadset comes from Qiaopu (Ningbo), and is of type H05RN-F with Australian SAA Approval Number ESV110434. Good to know that it’s all been qualified and tested for the Australian market.
With all of the metal involved, and no double insulation rating, the earth pin on the luminaire is critical for the safety of the unit. Here we can see an earth jumper lead from the power supply unit to the main heatsink, secured by a screw, and locking washer for good contact – the way it’s supposed to be done. Good job!
The unit itself is also supplied with a poorly photocopied leaflet with some mounting instructions and specifications. The diagrams are almost impossible to read, but luckily, the sheet can be downloaded online. There is a QA Pass stamp at the bottom.
The information on the sheet seems to be slightly contradictory to expectations. For one, it claims an ambient operating temperature of 35 degrees, which is pretty low. In a hot warehouse, near the roof, temperatures are more likely to be about 50 degrees, which their other leaflet seems to claim is acceptable. The other thing is the cable thickness, and terminal connection block, which claims it is not included and to use 1.0-1.5mm^2, but the included leadset has been pre-wired and is 0.75mm^2. For such a low current device, it’s not going to really be a problem though.
This is the exciting bit – we take it apart and see what it’s made out of. First thing’s first – lets approach this one from the front and take out the four slotted screws holding the front polycarbonate cover on.
Looking at the screws, you can see the washer and grommet combination which serves two main purposes – sealing the water out, and apply pressure to the polycarbonate over a wider area so it doesn’t crack at the hole itself. But already, a quality issue with the screws has been spotted – the two left ones have very good threads and end in a nice taper. The two on the right don’t – one of them have severely short threads, and the other one doesn’t seem to have consistent threads resulting in the aluminium threads being embedded into the threads of the screw.
The inside of the unit houses a silicone gasket around the edges that keeps the water, dust and humidity out. Examining the seal closely, we can see metal flakes sitting in it, which suggests that the damage to the threads may have happened earlier during construction of the unit, leading to these flakes being liberated. While there isn’t a sufficient quantity to cause a short circuit (especially due to the way the unit is constructed), it’s a sign of potential quality problems. I had extreme difficulty reassembling the unit, even when filing the screw threads and cleaning them up a little, it wasn’t possible to restore a quality seal as one of the screws stubbornly refused to fasten tight enough.
The main reason metal flakes themselves won’t cause issues is because the LEDs themselves are covered by PMMA plastic optic sheets. Here we can see one corner where one LED was not fitted – this is the same for the opposing corner.
Removing six screws allows the optic sheets to be removed and examined. These six screws are fitted with plastic washers to help spread the load and insulate them from the LEDs in case of any failure as they’re screwed into the chassis at ground potential.
Removing the optic sheet (one of two), we come into a surprise. The sheet itself is only one part of it – there’s a white plastic card as well behind the sheet. This is probably to improve the optical performance, as the white sheet has a higher reflectance than the white MCPCB coating itself, this might allow for a slightly higher lumen rating and a better appearance.
Examining the clear PMMA optic sheet, we find in the moulding, the words Luxeon M. This tells us that the LEDs used in the solution are Philips Lumileds Luxeon M series LEDs. This gives us confidence, as the LEDs are their own products and the Luxeon branding has always been a name of quality. The datasheet provides several interesting tidbits:
Freedom from Binning – tightly controlled correlated colour temperature.
- Tested at Tj=85 degrees C.
- Over 1400 “hot” lumens (120lm/W) from 3x3mm LED area.
- 70, 80 and 90 CRI bins.
- 11.2V, 5.6V and 2.8V packages, tested at 700mA, 1400mA and 2800mA respectively.
- L70 of 50,000 hours with Tj <=135 degrees C.
- Absolute Maximum Rating of 1200mA for DC forward current, with Peak Pulsed at 1375mA for series connected packages.
- Ripple with frequency >= 100Hz and amplitude <=1375mA acceptable as long as average current does not exceed 1200mA.
This brings us all the way down to the metal core printed circuit board itself (MCPCB). From the printings, we have determined that this unit is produced by Opulent, a solid state lighting professional with many years of MCPCB and thermal management solutions experience.
It seems that the unit itself can be “field serviced”, as after pulling on the board, it can be detached from the four pin connector with ease. Only two of the four pins are used, however. The rear of the board is covered with a black “rubbery” thermal pad, which helps conduct heat from the MCPCB to the heatsink. It’s probably not the most thermally conductive way (thermal paste comes to mind), but it eases the assembly process and adds a further layer of electrical insulation.
It can be seen that the screws perform double duty – it holds down the whole sandwich of the optic against the LED but it also holds the whole MCPCB assembly against the heatsink itself.
The rear aluminium has been machined to a smooth surface to improve heat transfer, although the machining pattern is still visible. The power cord from the rear power supply unit can be seen to come in through the cable grommet, and is of a two conductor cable only. Interesting is the use of mains-coloured cable colour coding, rather than the more familiar red/black of DC systems. An unused cable entry point, below, is covered by a grey coloured cap.
One major question is how the LEDs are connected on the MCPCB? This can have a direct impact on how the unit performs when faced with a failure and can contribute to failures to some degree. By shining a light at an angle, it was possible to outline each trace with a permanent marker. Forgive my crappy line-drawing skills …
That looks a bit messy, so I decided to give it some colour coding to ease interpretation.
The unit itself is arranged as two parallel sets of nine LEDs with one LED position shorted out by a resistor. A close examination of the PCB shows that another two positions can be shorted out with a resistor by design.
The positive input is routed around the edge of the board to the top right and bottom right LEDs which feeds the strings until the end at the central right LEDs.
Each LED has an associated copper area which it uses as a thermal pad, to help move the heat from the rear of the LED into the MCPCB’s core, which then moves it to the outer heatsink. These copper areas are shaped around the six holes to mount the optics (to avoid shorts) as well as four holes which are “keys” to ensure the optic sheets are placed correctly. As a result, there is a suboptimal design choice which may result in some LEDs (especially the edge ones) running hotter, which would reduce their lifetimes and efficiencies, as well as potentially changing their chromacity/colour temperature slightly.
This arrangement is somewhat common, but suboptimal, as it places two sets of LEDs in parallel without any current mirror arrangement to ensure the current flow in both sets of LEDs are evenly balanced. In cases of mismatch, which is inevitable due to component differences, one string will have a lower voltage drop than the other, and will take more current. This will cause that string to heat up more, which further reduces their voltage drop which further increases the current. This is somewhat mitigated to a limited extent by being mounted on the same MCPCB, and possibly by binning, but over time this may be the cause of failure of one of the strings.
Once one LED of either string fails, catastrophic things might happen. If one LED fails short circuit, then that will force that string containing the failed diode to take almost all the current (as current is exponentially related to voltage drop) which may cause them to change colour (become more blue) and then fail soon (due to potential overheating and overcurrent). This is somewhat related to the “cheap christmas lights” problem.
If one LED fails open circuit, that will cause the string’s current consumption to drop to zero immediately, which will force all the current through the other string, suddenly doubling its current. This is likely to be unhealthy and result in the same outcome.
The reason for this arrangement is likely to avoid excessively high voltages on the output for safety reasons. Each LED unit has a voltage drop of about 11.2v, and with nine in series, that brings the array voltage to 100.8v. If all the LEDs were in series, you would need to run 201.6v, which would not be suitable for a buck (voltage reducing) converter circuitry running in 120v countries, and would be a fairly high voltage DC.
That being said, with all units in one series string, a short-circuit failure would result in the loss of one LED, whereas an open circuit failure would result in all LEDs being disconnected.
Regardless, it’s a common design choice, and it isn’t particularly worse than anything else on the market. Either way, you won’t win, so probably the take home message is not to tempt probability. Less emitters, but more powerful emitters, are probably the safer choice regardless of connection. As soon as one LED begins to fail, the whole array will likely be at risk.
Lets turn it around and take a look at the power supply unit.
Opening up the rear we see the current driver unit screwed into the rear cover. This isn’t exactly the best way to ensure heat gets dissipated, as it’s in its own enclosure, in an enclosure!
The unit’s enclosure is reminiscient of 12V DC power inverters, in an aluminium “trench” with end plates. As this unit is waterproof, it has rubbery silicone gaskets at both ends, and rubberized cables coming in and out. The full specifications are given on the top, as well as the colour coding for the wires. It claims a case temperature of 85 degrees C, which is much more healthy. The input current is 1.5A maximum, and the power input is 120W maximum. The power output is 105W maximum, with a 1A current and 110V maximum voltage.
A basic calculation of efficiency would give this unit 105/120 = 87.5% efficiency which is a good result. This means that the unit itself would need to dissipate a maximum of 15W, which is manageable.
The current output of 1A means that each LED string is actually being operated conservatively at about 500mA despite being rated for 700mA. If one string fails, the remaining string sees 1000mA which is still below the 1200mA absolute maximum rating, so there could be a chance that the unit can operate with one string failed, although it depends on how well the heat can be transferred from those diodes.
The longevity of the current driver unit really depends on the quality of the design and components used inside. Lets take a peek …
… wait a minute! This thing is potted! Filled with silicone and epoxy! Damn. I’ve been thwarted!
Specifications and Subjective Experience
Having never dealt with something of this size, once I plugged it in and turned it on, I was instantly dazzled by the brightness of the light. There should be a big “do not stare directly into beam” label, because it’s that bright. I couldn’t even get a picture of it because it was so blinding to stand in front of. It claims to put out 11,000 lumens, and I can definitely believe that. The neutral white is also pleasant – it’s not yellow like warm white, or slightly blue like cool white. It’s more like a champagne colour.
Its 110W rating, with 100 lumens per watt represents a very high benchmark of LED performance for a complete luminaire solution. Its resulting light output is likely to be slightly more than a 500W halogen lamp. However, a quick look at the luminous efficacy table seems to suggest that 100 lumens per watt is at most twice where the better flourescents, metal halide and high-pressure sodium lights get. However, due to the optical properties of LEDs, there is not as much light lost in reflectors, so there is still some gain to be had there. There isn’t much of an on-off switching penalty or need to warm up, and there are no restrike delays in case of power loss, which are all benefits.
The unit itself powered on in under half a second with no flickering to report. Depending on when it is energized, sometimes it comes on immediately, sometimes it’s got a very short delay before it comes on. There seems to be no visually detectable flicker at all, which can be especially important when around rotating machinery due to the stroboscopic effect.
Running the unit for two hours, the heatsink itself felt only mildly warm, and was easy to keep the hand on. This suggests that either the thermal conductivity is limited, or (more likely) the heatsink is more than adequate and high temperatures should not be a problem for the unit.
Running it from normal mains, there was no audible noise from the unit and its power supply, which is definitely something which cannot be said about some of the lights it replaces.
Unfortunately, I don’t have the necessary equipment to perform photometric or radiometric measurements to confirm the performance of the luminaire in that regard.
The quality of the insulation is an important parameter in determining electrical safety, and so, the unit was tested with the Tenma Insulation Resistance Tester. Resistance tests were made on the primary side to ground, as well as on the secondary side to ground, and to the primary. Testing involved cutting the wire to the LED panel to patch in the test wire, violating the IP68 rating, so that the resistance with the LED panel involved is also measured.
The test result was a resistance of greater than 2 Gigaohms for all combinations (i.e. A to E, N to E, + to E, + to A, + to N, – to E, – to A and – to N). This means that the insulation is safe and effective.
This also means the unit is an isolated transformer, meaning that it would be safe to perform measurements on the secondary, although paying careful attention to the high DC voltage.
Input Power Parameters
Testing of the input power parameters was undertaken with a Tektronix PA1000 Power Analyzer, and using a 1A Multicomp Variac module. Mains power, as well as power from a Pure Sine Wave inverter was used for these experiments.
With mains AC power, through a variac, the LED luminaire performed excellently. While the luminaire was rated for 220-240V AC, the power input was swept from 100V to 270V and the luminaire operated correctly for the most part. Below 175V, the ballast began to make some audible noise (likely overstressed), and the power regulation suffered slightly.
The power regulation as a function of input voltage is exemplary, and the light output remained at about a constant 108W throughout the sweep, with the input current increasing to compensate for reducing voltage. This is likely to mean that the lights do not dim during brownout conditions and are likely to make such conditions worse by drawing more current as the voltage is falling …
Power factor was excellent, remaining above 0.96 at all tested points.
The current crest factor value was 1.5, which implies that the load “follows” the sine wave somewhat, but has some distorting elements. A check of the waveform confirms this, with the current consumption dropping to zero around the zero crossings (green waveform). The mains here has a nicely flattened top due to the large number of switching supplies in use (red waveform).
Interestingly, powering it off the inverter seemed to produce a poor result just around the 220-240v range, possibly due to resonance in the system (as inverters aren’t great at sourcing reactive power). It implies there’s been some sort of “tuning”, but this finding is likely to be of limited significance.
The waveform with the inverter seems very similar. The oscillations in the current waveform may be implying the switching frequency of the secondary output.
Warm-Up Power Consumption
As the LEDs warm up, their forward voltage drops fall, which reduces the power consumption. This can be seen by watching the unit’s power consumption as it warms up. The final consumption stabilizes near 108W after 90 minutes and 80 minutes of warming up under mains and synthetic mains sources respectively.
Ballast Output Parameters
The ballast output parameters were tested by using the PA1000 to measure the current and voltage. The results are as follows:
- Vrms 100.69V
- Arms 1.00000A
- Watts 100.69W
The calculated efficiency (roughly, as input and output were not simultaneously measured) is 100.69/108 = 93% which is an excellent result, above the nameplate implied 87.5%.
A more important result has to do with the output waveform of the ballast. This will determine if there is any flicker component and the frequency. This was determined by using the Picoscope 2205A USB Oscilloscope with a 0.1 ohm 5W wire-wound resistor as shunt resistor. Math channels were used to display the correct scale and value.
The oscilloscope plot shows the current waveform has a decent amount of ripple to it. The DC average current was measured at 1.013A (5% resistor tolerance, so probably trust the previous result more) with a ripple of about 650mA instantaneous. The current swings between about 722.9mA and 1372mA. This in itself is not harmful to LEDs, and as it doesn’t fall to zero, is unlikely to result in stroboscopic effects.
The perception of flicker is extremely unlikely as the oscillation rate is around 81.59khz, significantly above perception thresholds. There wasn’t any significant 50Hz residual on the output either.
The DC voltage measured by the Picoscope was 99.94V, which is close to the value above – keep in mind that this device only has 8-bits resolution, so I’d be inclined to trust the PA1000 values more.
Switching the unit into frequency mode, we can see that there are harmonics of 81.59khz showing up until about 2.5Mhz. When tested with a radio for radio frequency interference, it seems a signal was receivable from the device at a distance of 1 meter up to about 1Mhz. This may affect low-frequency radio enthusiasts if they’ve got many of these running nearby.
Aside from these low-frequency contributions, no significant harmonic energy exists above about 2.5Mhz.
Measurement of maximum in-rush current was performed with the PA1000 in the Inrush current mode, using the 20A shunt with 100A range. Testing for inrush current is difficult, as the inrush varies depending on when you power on the device in regards to the AC waveform.
I decided to power the light on and off quickly, 25 times in succession, and measure the peak inrush current. Doing this produces a lot of stress on any soft-starter components (e.g. thermistors) and provokes the highest inrush current possible. It can also cause failures (many manufacturers advise against closely spaced on-off switchings for this reason). I’m glad to report the unit survived … with a higher than expected inrush current value.
A positive peak of 73.74A was recorded, and a negative peak of -76.15A was recorded. The actual inrush current is likely to peak at about 80A in the worst case.
Thinking that this sounds relatively unreasonable for a 1.5A rated device, I decided to give it some time to cool down (30 minutes) and do single-power-on tests. After three cycles, I got a fairly similar result …
The peak was negative in direction, but over 70A. Yikes.
This value is mainly of interest as it can cause degradation of switch contacts and nuisance fuse blows or circuit breaker trips. Such a high inrush current makes it correspond closely to a capacitive load’s inrush current for switch derating, resulting in a derating factor of about 4. As a result, a 10A switch can really only handle one of these units optimally, and a 15A switch, just two units!
A check of the thermal circuit breaker trip diagram for single cycle trip seems to show that for a half-cycle at 50Hz, breakers can take about 2000 times their rated current for inrush without tripping, meaning breaker trip is unlikely due to inrush as it would exceed its sustained current rating before inrush would become a problem.
The same cannot be said of fuses, depending on the type of fuse. A 16A slow-blow small fuse can handle about 1000A for half-a-cycle, so it could handle the inrush for 12 units, but only the sustained current for 10 units, so again proving inrush is no problem. But a 16A quick-blow fuse can only handle about 250A for half-a-cycle, so it can only handle the inrush of 3 units in the worst case, despite 16A being sufficient for the sustained current of 10 units.
When the time-current characteristics are taken into account, it seems that it’s not really much of an issue, provided you don’t put too many on a single switch.
Through an anonymous donation, I have had a weekend’s worth of work … oh, I mean fun! The Philips luminaire seems to be well designed, with some common compromises which aren’t specific to just this one manufacturer. Quality components, from the LED to the MCPCB are used, and the performance of the current driver is quite exemplary. The thermal design is also excellent, with a non-dust collecting large heatsink which is durable and does not heat up too much. The construction quality and documentation could have been improved somewhat, but overall, are still a good product and package given the intended audience (commercial retrofit installers).