It’s been a while since I had to make a trip to my local IKEA (Rhodes, Sydney, Australia), and as usual, I always like to stop by the lighting department and see what’s new. Last time, they were just introducing their LED retrofit globes as the beginning of the end of them selling CFL globes with their mercury disposal and recycling issues. This time, there was not a single CFL left on the shelves – everything had changed to LEDs.
Their original LED series, the LEDARE, was still being sold, while a newer lower-cost product line named the RYET was being promoted. This post will look at the RYET 5W 400lm and 7.5W 600lm E27 globes as they were the cheapest globes available, and I wouldn’t feel so bad about destroying them in the name of science. Also of note is that a 13W 1000lm version was also available (although it seems it is being replaced by an 11.5W version), and a few other decorative globes in other bases.
The RYET E27 globes are sold inside a folded cardboard package that has a cut-out used as an internal separator and as a viewing window to see the product inside. This design, while somewhat attractive, does expose one globe to possible transit damage – in one case, my globe had some scratches on the plastic which were purely cosmetic, but otherwise not desirable. The globes are sold in packages of two, no doubt, as a cost-saving measure compared to the individual plastic-bubbles that protect the LEDARE globes.
This particular globe is their 5W, 400lm version which was on special for AU$5.49 per two, making it effectively AU$2.745 per globe. That’s a pretty low price, so I really wonder what the quality would be like inside. The product code is 903.057.42 and has a batch of Week 1 of 2016.
As with all of the LED products on sale, they all reach the EU energy efficiency level of A+. This particular globe claims a 15,000 hour lifetime, with 15,000 power cycles, a CRI > 80 and 80 lumens per watt. This isn’t as high of a lifetime as some of the LEDARE or competing globes, but isn’t too bad either. The luminous efficacy, however, is quite good considering smaller globes tend to have lower figures due to power overheads. The packaging shows that the lamps are not dimmable, and comply with Australian requirements, bearing the regulatory compliance mark.
Compared with other globes I’ve previously handled, this one is relatively tiny, featuring a small neck and light weight.
It claims to consume 48mA and 5W, so therefore, its power factor is likely to be about 0.453 which makes it a low power factor device. Maybe this is one reason why it costs less. As usual, the globe is Made in China.
On my scales, it only weighed in at 33.68 grams.
The other product is the RYET 7.5W 600lm version, which is physically slightly larger but otherwise comes in an identical type of packaging. This particular product has a code is 703.216.58. The batch is Week 12 of 2016. The specifications are otherwise identical to the 5W version, although the pack is priced at AU$9.99 which makes it AU$4.995 each. Still, a very acceptable price for an LED globe.
This globe follows a more classic GLS “A” shape.
It claims to consume 65mA, implying its power factor is 0.5. Another implied low-power factor. The globe is also Made in China. It has a slightly more substantial weight …
… but not by much, weighing in at 57.04 grams. Low weight is generally undesirable, as it implies less metal is used inside, and thus less heatsinking is supplied. This would mean that the LEDs would run warmer, and their lifetimes could be compromised. This may be the main reason behind the more “conservative” 15,000 hour lifetime rating.
In the past, owing to the high cost of LED globes, and my desire to actually use them in the house (after all, I bought them with my own money), I did not really go too far with my teardowns. This time, it’s different.
As with the other A-shaped LED globes, the front plastic cover can be pried off to reveal the internals. Operating in such a manner is not recommended, as the voltages inside can be dangerous, and often, the outputs are not isolated, as the cover is presumed to provide insulation.
The RYET 5W lamp can be seen to have an internal model code of KT-A400-MJT3030-05 and KT-A55N-S0002-V00. The design is dated to 29th September 2015, and the production date of the MCPCB is Week 43 of 2015. The board is made of an aluminium substrate for better heat dissipation, and has three 3030 type LEDs mounted. As 3030-types are usually used for approximately 1W per unit applications, this unit seems to be driving them a little on the high side. The MCPCB is secured to the shell with two tapping screws.
Interestingly, in this design, it seems they have learnt from the previous messes that have been seen in soldering wires directly to MCPCBs mounted to heatsinks – getting the right amount of heat and a good reliable joint is hard. Instead, they seem to have designed a two-sided PCB shim with through-hole vias that is “pre-soldered” to the MCPCB with the LEDs, and then, a solder “bridge” is used to connect the driver to the board. As the fibreglass of the shim insulates the pad and the vias can only sink a limited amount of heat, this makes soldering the connection a lot easier – and avoids the cost of a proper connector. Smart thinking.
The soldering of the LEDs is acceptable, with the pads slightly oversized compared to the LEDs and what appears to be proper filling of solder onto the thermal connection in the rear. However, it seems the LED may have sustained some mechanical damage somehow.
Desoldering was necessary to remove the MCPCB from the rest of the housing.
Only a very minimal contact area is made between the MCPCB and the small can heatsink, namely at the edges. This is not the ideal situation, although a small amount of thermal paste seems to have been used. Some of it may have also mixed with the silicone used at the edges to hold the diffusing dome in place.
In this image, we can see the tip and can connections to the crimped E27 base. The wire is sandwiched between the plastic body and shell, and as it is stranded, it should make good contact. The tip lead is also wedged into place, and is one leg of a fusible resistor, which provides protection in case of lamp failure.
The driver PCB is also similarly named, and seems to be a design specific for this lamp. Another cost saving feature can be seen – namely the use of a single sided PCB, with very few components. There is no discrete surge protection or transient filtering capacitors, although in many space-constrained applications, it can be considered normal to omit. At least an RF-suppressing inductor is provided.
The capacitors are both Aishi capacitors from the CD11GHS (6000h) and CD11GD (6000h) series, both long-life 105 degrees C capacitors designed for use in lighting drivers. These should not pose any major issues, as my experience has been rather positive with their capacitors which are widely adopted by other brands as well.
Most of the components are mounted on the underside, including a bridge rectifier, several resistors, a diode, a capacitor and the controller IC (of which only 7 pins are used). The controller itself could not be identified, although the markings seem to show 7YL5K53K being marked on the chip.
Seeing such a fully-fledged integrated driver inside a low-cost LED lamp is somewhat surprising, as some other “Chinese” efforts on eBay use very crude capacitive dropper arrangements for simplicity and low cost.
The 7.5W version took a lot more plastic-cracking to get open, which also resulted in me gouging the MCPCB slightly. The unit also comes from the same manufacturer (it seems) with a code of KT-SS3030-34 and a date of 19th November 2015. The PCB manufacture date is Week 1, 2016. The board uses a similar two-screw mounting arrangement, and soldered shim to MCPCB contact, although the soldering is slightly worse in this model. It’s good to note that both MCPCB have wide copper areas for better heat transfer, so it seems that the manufacturers are doing something right. This one claims 7.5W and uses six 3030-type LEDs, so is not pushing them as hard as the 5W one seems to be.
Unlike the 5W version, the footprints on this MCPCB are a little on the small side, with the solder being completely hidden underneath the chip save for a small bubble at one side. The silkscreening seems to be slightly misaligned and larger than the actual chip, which gives it an odd appearance. At least the chip seems to have survived well, and although it’s a 3030 chip, it’s clear the LEDs are different in the two products (more on this later).
The same connection to the heatsink and to the end-cap is made, thus the comments made before apply equally to this lamp.
Unlike the other lamp, this board has a completely different design using a beefier inductor, a few polyester (non-safety rated) capacitors for transient suppression and filtering and a double-sided PCB although not strictly necessary. It seems slightly surprising that the polyester capacitors seem to have a 450V DC rating, which may mean they won’t handle mains transients so well as compared to safety-rated AC polyester capacitors.
The output filtering capacitor is also an Aishi capacitor from their RF line, rated for 105 degrees C lifetime of 3,000-6,000 hours.
The rest of the components are mounted underneath, and it seems to be based on an AP8022 switchmode power supply chip. There are a few SMD joints with solder blotches and evidence of touch-up and over-use of flux, so construction quality is slightly lacking. That being said, it seems that this driver belongs to an older product – KT-400LM-56 dated 25th October 2014. It seems that it may have been adapted from a former 400lm globe.
A quick check of isolation showed that both globes did not have isolated outputs, so should not be touched if opened. Both of them had 0.5-0.6v (one diode drop) to either live or neutral in testing, thus touching the pins when running may put you at risk of electric shock. Of course, it is designed to be used “as is”, not opened up like I do in my teardowns.
Ultimately, as I didn’t want to waste them, I decided to cannibalize two of the adapter sockets and solder in the wires directly and tape them up so they can still be used.
Electrical Performance and Testing
Driver Output, LED output
Out of curiosity, I decided to measure the driver output voltage. To my amazement, the 5W globe had a driver voltage of 141V, whereas the 7.5W globe had a 37.58V output. I double-checked and reconfirmed that this was not a mistake – the output of the 5W globe could hurt if touched!
The reasoning behind this is due to the LEDs themselves. Despite both units using 3030-style LEDs, their internal configuration is vastly different.
5w globe’s 3030 LEDs 7.5w globe’s 3030 LEDs
In the case of the 5W globe, we can see each 3030 LED package has two wire bonds – one for the positive coming into the LED chip, and one for the negative going out of the LED chip. However, the LED chip has been made in such a way that it seems to resemble an array of 15-series-connected LEDs in a 5-across by 3-down arrangement. As a result, each package has about 15 *3 = 45V voltage drop, and three of them in series gives 135V – close to the 141V measured when cold. This arrangement is quite interesting as it means that the current is kept low, which could improve the cost of the components needed to make the ballast, but at a consequence of making the output voltages not safe for accidental contact and increasing secondary insulation requirements. At least, the 15 internal junctions seem to be on the same chip, thus no additional wire bonding is necessary.
In the case of the 7.5W globe, each 3030 chip actually contains two separate LED chips, each in a long rectangular shape, connected together by an internal wire bond link. As a result, it has an extra two wire bond connections that are made, and each chip hence has a two-LED voltage drop of about 6V. For the 6 chips that are used, the approximate voltage drop is calculated to be 36V, very close to the measured value. However, this arrangement is not a preferred solution, for the same reason that some “massive array” 100 LED floodlight chips fail – wire bonds are subject to expansion and contraction stress due to heating and cooling, and these bonds are often the cause of failure resulting in LEDs that flicker until they get somewhat warm and fail altogether open circuit. The only way to know how reliable these are is to run them and see.
Based on very casual usage, it seems the light output is fairly close to the promised levels, and the light quality is acceptable for home usage. It’s not particularly special, but it’s not particularly poor either.
Warm-Up Power Consumption Trend
Both globes were left for 20 minutes to run-up from cold, with power measured by a Tektronix PA1000 Power Analyzer.
It seems that both globes were below the label rating for power consumption. The 7.5W globe settled out at just below 6.5W, whereas the 5W globe settled out at about 4.7W. This in itself is not a big issue, as it just means the power consumed is likely to be less than expected based on nameplate rating, although I cannot tell if this is at the expense of light output.
The voltage provided to the globe was swept with a variac to see how it behaves under a range of line conditions.
The 5W globe seems to function correctly about 150V. This means that this globe is pretty much a 220-240V globe which cannot function in 120V countries such as the USA. The globe has a well regulated output which tails upwards slightly with increasing voltage. The power factor was about 0.55, which is considered low. This is not optimal for “helping” the power distribution network, but is common amongst households where reactive power is not something they are charged for. The converter is hence a buck converter – it reduces the input voltage, and thus cannot operate at an input voltage below that required by the LED array (approximately 135V, plus internal converter losses).
The 7.5W globe, however, showed a much wider operational voltage range, operating correctly even down to about 62V. This globe could well run in 120V countries without modification, and at a fairly close-to-unity power factor, it seems. At 230V, the power factor is 0.8, which is not as bad as expected based on the label, and is much better than 0.5.
The RYET series tries to offer an even lower cost entry point to LED retrofit globes. The packaging is spartan, and the globes are sold in two, but the price is quite attractive and the internal circuitry was of a higher standard than I would have initially expected based on the price tag alone. As usual, compromises had to be made in achieving the lower cost, with light weight, limited heatsink contact, minor construction quality issues and potentially questionable LED chips in use. The power factor was somewhat poor for the 5W globe, but better for the 7.5W globe.
However, keeping in mind the price – the advertised lifetime is less than some of the more expensive competitors, and the components used are of mostly sufficient quality, thus it seems it is a good balance for especially residential users where lights may not be used for extended periods (e.g. 24/7). The light output is of an adequate quantity and quality, based on the label expectations.
Their ultimate reliability, however, cannot be gauged on the design alone. The only way to know is to try it and see.