Review, Teardown: Philips LuxSpace G3 ANZ (DN498) LED Downlight

Following on from the same vein of LED retrofit solutions as my last posting, those with suspension ceilings or existing recessed lighting will also have a desire to retrofit older PAR globes, halogen downlights and CFL based “can” style downlights with LEDs to maintain a consistent appearance, and to reap the benefits of reduced energy consumption, heat and maintenance requirements. As expected, there are products to suit those requirements as well.

In this review, we look at the Philips LuxSpace G3, specifically an DN49823W4KDWHMW DN498 1xDLED-4000 PSD-E M-A WH CAU (12NC: 911401803199) which is a 23W wall-wash LED “compact power” recessed downlight with a 4000K neutral-white 2000 lumen output. This product is specially designed to be used where accenting a wall feature (e.g. along a hallway) is the primary aim, with an asymmetric beam.

This product is just one of the LuxSpace series of recessed downlight solutions which encompasses many options including 3000K or 4000K colour temperature, glass front accents and various trims, different optics, and fixed or DALI current drivers with a range of power levels.

This review was made possible by Voltimum and Philips Australia which sent a sample for review under the review challenge terms.



The unit comes in a large durable thick cardboard box, with the unit itself suspended in a cardboard cradle. Damage in transit is highly unlikely. The unit itself appears to have a production date of 21st March 2015, and is rated for 23W with a claimed 55 LEDs. It is part of their Compact Power series, developing 2000 lumens and is suitable for replacing 2 x 32W CFLi downlights.


The unit comes pre-assembled ready for installation. The main unit is the downlight itself with a hefty metal rear heatsink, and metallic front bezel which are not easily separable. The unit features plastic spring-loaded rat-trap style clips for easy installation, requiring a 200mm diameter hole.


A close look inside through the grilles allows us to see the LED engine itself, which has its rear fixed to the metal heatsink and a substantial height likely with internal optics.


The unit is a non-IC unit and requires clearance from insulation. The installation sheet provides further information, however, it seems to want about 50mm clearance from structural elements on the side of the downlight and the current driver, and 170mm height from the front to the ceiling. It also recommends 150mm between the downlight and the driver which is sensible to ensure best lifetime by reducing thermal build-up. Installers are also warned not to let the current driver rest on the downlight heatsink or to loop the mains cable over the current driver


On the other side of the unit, the product nameplate label provides further details, including a power factor of 0.95, and the Australian regulatory compliance mark. Other specifications from the installation sheet include a 4000K colour temperature, a UGR rating of 22, a weight of 1.86kg with a nominal current of 0.12A and an inrush current of 20A for 150us. It is double-insulated, and claims an IP20 rating for indoor use only at an ambient of 25 degrees C. Up to 28 units can be used on a single circuit with the Philips UID8520 DALI Dimmer as the recommended controller.

The cable connecting the downlight to the current driver is a pre-installed multi-conductor four core cable with special connectors and a plastic flexible conduit.


The unit can be disconnected from the current driver for replacement, but there is a warning that this must not be done while power is applied or else damage may result.



The current driver is in a beige plastic box, and is a Philips Xitanium 25W LED driver. It claims a case temperature up to 90 degrees C with ambient up to 65 degrees C. From its nameplate, it can take in 30W to drive 25W of LEDs resulting in an approximate efficiency of 83.3% which is good but not class-leading. This may be due to the overhead of the DALI capable circuitry which consumes additional power. It claims a power factor of 0.9 (less than the 0.95 claimed on the product). It is made in Poland.


2016021117182408The primary side comes pre-assembled with a 1.5m flexible cord from Qiaopu. This is a two-pin plug with no earth as the unit is double-insulated. The cable itself is printed with an SAA approval number of NSW18298.

No DALI cable is supplied or installed, and you as expected to provide an appropriate round-sheathed mains-rated cable of 0.5 to 1.5mm (or 0.2 – 0.5 mm^2 / 20-24 AWG according to the leaflet). Flat sheathed cable should not be used.


The front of the unit has a white matte aluminium finish, with a sandblasted elliptical asymmetric reflector. The light source is a slightly yellow (possibly remote-phosphor) disc with a matte translucent plastic front.

Digital Addressable Lighting Interface (DALI)

This particular current driver employed in this unit is denoted “PSD” indicating that it is DALI compatible. DALI is an IEC 62386 standardized protocol for lighting control, being purely digital and capable of individual and group luminaire commands. It succeeds the earlier 0-10V analog control. The protocol itself is “open” and employed by several manufacturers of lighting equipment allowing for interoperability of equipment.

The DALI bus itself is a 2-wire bus which can be connected either way to DALI devices (i.e. wire order is unimportant). It is designed to run in mains-rated insulated cable alongside mains with noise-resistant signalling voltages of 16 +/- 6.5V for high and 0 +/- 4.5V for low. The bus is current-limited to 250mA maximum, with each device allowed at most 2mA, and data signalling performed by pulling down the bus to signal a zero. The data is modulated as 1200bit/s Manchester encoding with two-bytes (16 bits) per forward message (with one start bit and two stop bits). Backward frames are one-byte long. Forward frames must be spaced at least 11 bit-times, with backward frames being spaced 3.5 to 11 bit-times after the end of a forward frame. Some commands require repetition within 100ms on the DALI bus to ensure they are carried out.

A big advantage of DALI is simplified cabling compared to earlier systems as it is possible to individually address luminaires with their single address (up to 64 on a single bus, multiple buses can be connected together by gateways). Further to this, luminaires contain configuration memory which allows for group-addresses and scenes to be programmed allowing for multiple luminares to be addressed by a group-address command and particular scenes to be recalled without command latency which can cause a noticeable delay in turn-on. Dimming functionality is also built-in to DALI devices.

On the downside, configuration of the units is required to take full advantage of their capabilities, and DALI controllers and power supplies can become rather expensive if only a few lamps need to be controlled. For a hobbyist like myself, who has no real need for DALI, purchasing a controller is cost-prohibitive. Seeing as DALI is one of the major features of this unit, I decided to hack together something for demonstration purposes.

My DANGEROUSLY Simple DALI Interface

!!! Unless you have a death-wish or you really know what you’re doing and you’ve taken steps to ensure your safety, do not (under any circumstances) try this at home! I will not be held responsible for any damages or injuries due to your ability or inability to use the following information or your attempts at emulating what I have done. You are advised not to use the following without first understanding all dangers and risks it poses to you and your property. !!!

While DALI is specified with ELV signals, the interface itself can cross mains cable and ballasts are only required to implement functional insulation to the DALI interface. In case of any faults, it is likely that the DALI interface could be at mains potential and must be treated as mains along with appropriate isolation.

Unfortunately, designing an isolated interface requires a few more components than I have on hand, and is something I don’t have much experience in. Another thing to watch out for is that many development/evaluation kits for DALI have some level of isolation but not enough to meet DALI standards. It also requires a current-limited supply of power on the bus, with fast current limiting a necessity for ensuring that the drivers on the ballasts are not destroyed in trying to communicate on the bus.

The first step in making the interface was to choose a microcontroller – what else but the Arduino? Thanks to the work of edinburghhacklab and Hugo at Tech Toy Hacks, code was available for basic DALI communication. The code from Hugo implements a DALI transmitter and receiver, however, it has some direct register access on the AVR which could not compile correctly under current versions of Arduino IDE. However, he does provide very important documentation on the DALI message format itself, namely that for a broadcast all direct lamp power command, the first byte is worked out to be 0xFE, with the second byte containing the power level (1-253, with 0 as off, and 254 as do-nothing).

As I did not intend to wrangle with the intricacies of programming a short address or scenes on the DALI controller, and I was to have just one device on the bus, I decided to modify edinburghhacklab’s code to make a sketch I called dali-simple. This basically takes a two-letter hex value light level from the serial interface at 115200bps, and puts out a DALI message on pin 8 corresponding to 0xFE followed by the light level. This is only a DALI transmitter and has no ability to receive the backward frames, and is hence not useful for query-based special commands.


The code was tested with FF as the value, and indeed, with an oscilloscope, the timing and Manchester encoding was verified. The output is inverted as it is to drive a driver circuit where high input voltage pulls down the bus, but the sketch can be modified to change this behaviour.


The next step was to create an interface using as few components as possible, preferably those I had on hand. The Arduino Uno was already conscripted, so the next step was to have a >9v DALI bus supply with appropriate current limiting. For this, I decided to use two 9v batteries in series, as I know they intrinsically have a limited current (in case I goof up). To further ensure current limiting, a resistor is placed in series with the supply – I chose a 3.3kohm resistor which is what I had on hand for a short circuit current limit of 5.45mA. As the maximum one-device consumption is 2mA, the bus should stay above 11V even with the maximum consumption occurring which is enough to signal an “idle” high. Theoretically, you need a resistor of about 72 ohms, but power dissipation will be your enemy (4.5W). For common 0.25w resistors, the smallest value is 1.2kohms for a current limit of 15mA.

To actually modulate the bus, a logic-level high-current MOSFET (IRLZ34) was used to short the bus out in time with the data, itself connected through a pair of 560 ohm resistors to limit back-flow of current if it failed and to bleed the gate capacitance when turning off. A lower resistance might have been better to ensure faster transition times, but it was enough for testing. Finally, a diode was placed to ensure that the MOSFET would not be destroyed if the batteries were connected backwards, and a capacitor was used to buffer the input voltage to ensure that the modulated data was clean.


The circuit was built as efficiently as possible on a scrap of veroboard and looked like this. I even soldered to the battery terminals since I had no battery snaps at the time. A terminal connector block was used to connect the DALI bus wires to the thinner header jumper wires. Of course, I didn’t just blindly employ this circuit …

Insulation resistance testing between all terminals was undertaken and proved to have more than adequate insulation to ensure safety. Despite this, extra precautions were taken to ensure safety. The controller itself was connected to a tablet running on batteries, and remotely controlled via Wi-Fi so I didn’t have to have contact with anything directly connected at any time. Further to this, the luminaire was run off an isolation transformer and placed on an insulated surface to minimise any other risks.

Initial testing proved that the circuit worked as intended and DALI control was indeed achieved. This allowed for more comprehensive testing to be undertaken than otherwise possible, without costing me an arm and a leg …

… still, please don’t try this yourself, because it could end up costing you a lot more.

Performance Test & Subjective Opinion

In this section, I will present the results of various tests undertaken on the luminaire along with some interpretation and subjective experiences. Please note that I am a hobbyist with a limited amount of equipment, and as such, some of these results are not comparable to those which are derived from a test laboratory. Please make a note of the methodology and interpretation when in doubt.

Insulation Resistance Test

For safety reasons, a thorough insulation resistance test at 500V was performed to ensure the separation of outputs and inputs from the mains. The full report is here, but the summarized results are as follows:

  • Active to Case – 18.53 gigaohms
  • Neutral to Case – 47.48 gigaohms
  • {Active, Neutral} to {DALI 1, DALI 2, Red, Black, Orange, Blue} – >260 gigaohms

The insulation was excellent and complete, which was mostly out of range of the meter but at least five orders of magnitude better than the 1 megaohm requirement. The measured results to case may have been due to capacitance charging or leakage through surface contact rather than actually insulation leakage, as the only paths all measured >260 gigaohms on their own.

Power Consumption vs Voltage, DALI Dimming, Warm-Up and Efficiency

Testing was performed with my Tektronix PA1000 Power Analyzer and a variac to see how well the device coped with line voltage variations.


It was found that when power is applied, the luminare operates at full brightness without any DALI command, making DALI units capable of operation in non-DALI enabled installations. However, there was a slight time delay from the application of power to when light was emitted, closer to one second. On the whole, well regulated performance was had for voltages of 166v and above, with a power consumption of about 22 to 23W. The power factor remained 0.95 and above throughout this whole region, which makes good news for industrial consumers where high power factors are a requirement. Between 230-240v, the power factor was 0.97-0.98. Below this, the current driver appears to go into a low-power-consumption standby down through to about 12v.


When dimming with DALI, it seemed that the minimum dimming level is between 0x50 and 0x40. The power level can be seen to follow an exponential curve to match the logarithmic characteristic of our eyes. The dimming functionality could possibly save energy and heating if used properly. I might have miskeyed the F0 step as FF which resulted in no change in brightness from E0. It was discovered that the dimming range is about 10:1, but the power factor also follows the dimming, with lower power factors when dimmed. This is generally not a significant issue, as the power consumption also falls as it is dimmed.


During warm-up, the power consumption falls slightly as expected with LED products. The 15-minute stabilized power is close to 22.3W.

As I do not have two PA1000 meters, I could not measure input and output power simultaneously. Instead, by performing sequential measurements, we determined the power in was about 22.385W and the power out was 19.226W representing a calculated approximate efficiency of 85.9% which is quite good and exceeds the nameplate value.

DALI Standby Consumption

As most DALI lights are used with the mains power continually supplied to the luminaire so it can respond to DALI bus commands from automation devices, the standby power consumption can be of importance.

By using the Tektronix PA1000 Power Analyzer in the IEC Standby Full Compliance mode using the PWRVIEW software, a full 15 minute test was performed using a stabilized inverter power source. The resulting consumption was 427.88mW (full report here).

As a result, over a whole year, you can expect each luminaire to consume 3.748kWh in standby energy. At residential power costs of about AU$0.25/kWh, this would be about 94c/year/luminaire.

In-rush Current

The specifications claimed 20A inrush current, however, this was not repeated with testing with the PA1000. By cycling the luminaire 10 times in a row, the in-rush current measured 36.02A, which is slightly higher. This is likely because the inrush current protective soft-start mechanisms may have been defeated by continual power cycling, however, due to the random nature of in-rush (i.e. it matters where in the mains cycle power is applied, and possibly the previous state of the device), a single trial is not sufficient. However, the first in-rush reading was 16A, which suggests that the one-off inrush current is likely to be close to the magnitude claimed.

Output Waveform and Flicker

Unfortunately, due to the EMI noise of all of my gear, measuring flicker using a red LED as a photodiode was not particularly fruitful. Instead, I had to resort to direct current measurement to determine the behaviour of the ballast, as it was rather complex. As it was determined that it is SELV, direct measurements could safely be made by splicing the wires and using a burden resistor (0.1 ohm equivalent non-inductive).


At full brightness, we can see that the current is not steady but continually varies in a way which is in sympathy with the mains. The current does not drop to zero, which means that flicker is unlikely to be perceived.


At full brightness, the variations cycled in the hundreds of kHz region, which would be imperceptible, with an average current recorded of about 618mA.


At a DALI dimming level of 0xD0, we find that the dimming is a rather complex function, resulting in a smaller amplitude ripple output.


As a result, this can be considered smoothed PWM output which is much less likely to irritate eyes and which is less likely to provoke safety issues from stroboscopic effects when dimmed. Flicker is a non-issue at this dimming level.


However, when dimmed completely to the minimum level, PWM does resurface, but at a rate of 84.4khz which is too high for humans to perceive. On the whole, the light output was relatively stable when dimmed, and the dimming range is about 10:1.

Current measurements were also made with the PA1000 for higher precision. It reported the following:

  • Vrms = 31.121V
  • Arms = 624.90mA
  • Power = 19.226W
  • Apk+ = 825.13mA
  • Apk- = 439.57mA
  • Vcf = 1.0137
  • Acf = 1.3204

This very much corroborates the current measurements achieved with the oscilloscope which show a current at full brightness that varies somewhat (Acf > 1) but does not reach zero (Apk-). The power delivered to the LEDs is 19.226W, which is less than the expected 23W.

Spectral Output

As with other reviews on this site, the spectral results were measured with an Ocean Optics USB2000 which is a little out of calibration and has blue response issues. A correction curve based on solar simulation data was developed, but is not perfect due to quantization noise and other factors, so the figures presented below are to be taken with a grain of salt and can only be compared with other reviews on this site.


On the whole, the output exhibits the same wide-band greens, oranges and reds as the other 4000k Philips SmartPanel 2.0, so probably has a mix of phosphors between the 2700k/6500k style LEDs. The calculated CCTs seemed to imply the unit was slightly cooler than 4000k, and the CRI figures seemed to imply a slightly lower CRI than the SmartPanel 2.0 (probably just barely meeting 80).

A second opinion of CCT was had with the Nikon D3300 taking a raw photo of the front of the light, which returned 3950K, which suggests the colour temperature is spot-on.


Other Observations

On the whole, I was quite impressed with the unit, and its inbuilt DALI dimming capabilities. The quality of the current driver construction, design and components were outstanding. The heatsinking on the LED engine seemed more than sufficient and the heatsink was only barely warm to the touch, as was the current driver. Only a very quiet buzzing was audible from the driver, only at close range (30cm).

The current driver also operates without the presence of a DALI bus, thus allowing for the use of DALI drivers in non-DALI installations. However, this is probably not a preferred route as such drivers are typically more expensive.

The mechanical construction of the device did leave a little to be desired. The front reflector optic does not match the LED engine surface perfectly resulting in a slight light leak behind the unit, which is probably of no great consequence.


However, the front reflector on the unit is slightly loose and can be rotated slightly. I’m not sure if this is a “feature” to allow slight alterations to beam alignment, but it could also cause noise during warm-up/cool down as parts expand and contract.

It was also found that by rotating the front, that the reflector can be trapped “below” flush, and other times, be bought back to flush.

2016021117262423 2016021117252422

Further to this, the white interior trim also exhibited some bowing at the ends on both sides with a significant gap, which might not be noticeable when installed onto a ceiling, but is noticeable at close range. This suggests the mechanical construction (Made in China) may be somewhat lacking in quality in comparison.


There are many variations of the LuxSpace G3, however, the core engine, heatsink and driver units are likely to be common, and these components were generally well designed. The claimed lifetime is still 50,000 hours to 70% (L70) which may be slightly conservative, as other products do occasionally claim more (e.g. L80/L85).

Once concern was with the LED engine which had an input voltage of about 31V. As each white LED has a voltage drop about 3v, it seems that this is a five parallel strings of 11 LEDs in series design. The high number of devices, and the high number of parallel devices especially, can increase the risk of current imbalance in one string accelerating the failure of the assembly as a whole. This design decision was probably necessary to ensure the output voltage remained below 48V, so it could be SELV rated.


When it comes to LED lighting products, it seems Philips Professional products show good quality design and good performance. In the case of the LuxSpace G3 reviewed here, the driver unit showed excellent design with quality components, and had good performance with DALI dimming in regards to flicker. It also exhibited good efficiency levels, excellent power factor and more than satisfactory levels of mains voltage variation tolerance. All parts were only barely warm to the touch under regular operation, with nothing considered “hot”. It also exhibited perfect insulation resistance readings which bodes well for safety.

The only let-down was a slight buzz from the driver at close range which is not audible under regular circumstances, and problems with the mechanical fit of the front reflector and side trim which had noticeable gaps and looseness, along with a small light leak.

The LuxSpace G3 is available in a wide range of power levels, optics and trim/cover glass options with 3000K and 4000K CCT and fixed/DALI capable driver options to suit.

Thanks again toΒ Voltimum and Philips Australia for providing this sample for review.


Teardown, PCB images removed as per Philips request.

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 Electronics, Lighting and tagged , , , , , . Bookmark the permalink.

6 Responses to Review, Teardown: Philips LuxSpace G3 ANZ (DN498) LED Downlight

  1. Andrei Rusan says:

    very interesting reviews on the LED light sources lately – thanks!
    I was wondering if you have access to any of the new IKEA led bulbs already, as I saw that you tested/torn down some Ikea CFLs earlier. I got one of these Ikea LED bulbs (the 1800 lumen version) and it seems to flicker weirdly when dimmed down (and sometimes even when not fully dimmed). The buzzing is also quite strong. I wonder if it’s the bulb or something else causing it, but I don’t have the necessary equipment to test. My dimmer should be a quite ok Schneider Electric one and worked well with GU10 lamps in the earlier light source. Unfortunately I am in Europe and sending one bulb over to you would not make sense, but I am really curious πŸ˜›

    • lui_gough says:

      This is a rather sophisticated question and really does bring in a few issues which I’ll try to explain without getting too technical or confusing.

      The first thing is that dimmers are not just dimmers. There are several sorts of dimmers, each optimized for a particular sort of load. There is the leading-edge type dimmer, which is commonly used with incandescent loads and is made from TRIACs/DIACs. These dimmers are generally very robust against short circuits (e.g. when a filament blows) but they aren’t optimized for capacitive loads such as that posed by most LED retrofit globes. The better-suited trailing-edge dimmer normally uses more sophisticated circuits and requires MOSFET switching, but is not robust against transient short circuits and are generally more expensive and are so less often fitted. These are generally more compatible with LED globes but there too are issues which can crop up.

      One thing to realize is that dimmers in general (except for rheostatic/variac style dimmers which are large, chunky and rarely used) work by basically “lopping off” part of the nice AC sine wave cycle. The difference between the type of dimmers is where this “cutting” of the sine wave occurs. However, whenever a nice smooth sine wave is “abruptly” changed, this imparts a lot of harmonics which in turn cause harmonic currents to flow in the LED globes. Due to magnetostriction, every time we get a sharp impulse where the wave is cut, if the current driver circuit isn’t made from high quality transformers/inductors that are wound tightly or mechanically damped, they tend to “jump” along with the impulse creating an annoying buzzing sound. The buzzing sound can also come about because of changes in the operational frequency of the driver circuits within the globe as the lighting level changes.

      More than this though, is that because of the load characteristics of LEDs, the dimmers themselves may not be able to switch on or off cleanly and instead “ring” or “misfire” causing flickering, uneven light levels and or excessive buzzing. This is quite a complicated part, so I’ll let LED Journal do the explaining:

      As a result, even though many dimmers may be rated for 400VA loads, this does not mean you can connect 400W of LEDs and expect it to run well. In fact, a lot of the time, even modern dimmers have issues running more than 25W of LED globes, and have a maximum limit of 80W total. Others require the use of “corrective devices” (which are likely to be harmonic filters) similar to these made by Schneider Electric/Clipsal ( (not an endorsement – I haven’t used nor tried them myself) which try to correct the load to match something more close to resistive so that the dimmer can operate more reliably.

      That being said, retrofit LED dimming is an inexact science and many people are in a similar situation to you, having to hunt around to find globes that work reliably and satisfactorily with a wide enough dimming range, stable and quiet operation.

      As I don’t travel to IKEA very often, and I don’t really need any more globes, I probably won’t have a chance to sample their latest product offerings. I’m not even sure if the Australian customers get the 1800lm version – last I checked, we didn’t. Products can indeed vary throughout the world, so I wouldn’t take it for granted that the products I can buy here would be the same internally to the products available overseas.

      – Gough

  2. matson says:

    Why should flat sheathed cable not be used? Or do you mean one (flat sheathed) cable should not be used for power and DALI?

    • lui_gough says:

      Good question. From what I can tell, the reason is that the cover and screw form a cable retention system that prevents the cable from being pulled out of the “push fit” terminals in case somebody trips on a cable in a roof or tugs on a cable during other installation. Because both sides of the cover have “jaws” of the same sort, and that round-profile cable is used for the mains, if you use flat sheathed cable for DALI, the cable retention force will either be uneven (i.e. insufficient for the thinner DALI cable) or may result in too much force being applied to the round profile mains cable potentially damaging and compromising its insulation over time (i.e. imbalance).

      Aside from that, it’s also possible that the choice is made just to ensure greater physical cable insulation as well, as the round profile is a little thicker and more easily identified as mains-type cable so someone (who might be half asleep) is less likely to mis-identify that line as an SELV output and potentially get themselves into trouble.

      – Gough

      • matson says:

        Good, eye-opening, answer! I am not smart enough to have thought of that.

        • lui_gough says:

          I don’t think I stressed it enough but there is an additional issue – if someone does tug either mains or DALI cable and it falls out of the terminals (and push fit terminals can let go if you pull hard enough), it will be “floating around” in that space and it’s quite likely that DALI and mains will come into contact. While technically DALI standard says that all bus participants should be isolated and should survive such an event, if installers are complacent and take DALI on face value (i.e. signals between about -4.5v and 22.5v at the most), then they will get a shock.

          You might be thinking that there are two sides of the ballast – but if you place the DALI terminals on the SELV side, and you mix it with other DALI devices which are designed such that DALI could (in a fault) be at mains potential, then you can compromise your SELV output as that can become mains if the wires fall out.

          – Gough

Error: Comment is Missing!