What’s inside the Cree T8 LED fluorescent-replacement lamp

After evaluating the performance of the new Cree T8 LED fluorescent replacement lamp it was time to take one apart to see how they are designed. A quick recap: The lamp uses 21W to produce 2100 lm of 90CRI light.

Cree T8 linear LED heat sink and lock annot_400

As you may remember from the article, the lamp’s cross-section is oval rather than round.  Cree gave it rotatable end-cap plugs so that you can orient the lamp if it needs to mount at an angle to fit into the fixture. (I should have tried this on the first fixture I bought, but I thought they were there to lock the lamp in place.)

T8 LED lamp heat sink and pcb

I assumed that the LEDs and their pc board would mount directly onto the long aluminum heat sink. Wrong: The heat sink has a channel on each side that supports the pc board. The pc board just slides along the channels. Other than the lip of the channels, it has no heat conduction path to the heat sink. This surprised me, but, on the other hand, this is a huge heat sink area compared to the densely-packed LEDs on an incandescent replacement bulb. These LEDs run cooler and don’t require the close contact to their heat sink.

Cree T8 LED lamp inside

Hand-soldering and reliability

One of the clever assembly approaches used in the Cree A19 60W-replacement bulb was the use of clips to snap together the bulb’s electronic components. These clips eliminated the need for hand-soldering the connector wires to the pc board, and hand soldering is always a quality and reliability leak for high-volume electronics manufacturing. So I was disappointed to see that this lamp reverts to a hand-soldered wire attachment method. As you can see in the photo above, the solder joints, which are surface mount, not through-hole, are more like blobs than joints. These T8 lamps are also assembled in China; the A19 bulbs are assembled in the US, so this may be the reason for the hand-soldering.

The power control components are located at the ends of the lamp. The black rectangles may be machine vision alignment marks.

The power control components are located at the ends of the lamp. The black rectangles may be machine vision alignment marks.

Cree T8 components with inductor

The LEDs used in the lamp look like Cree XLamp XH-B ceramic package LEDs. There are 120 of them, each putting out 17.5 lm. The large square component is the inductor, the board’s sole magnetic component.

Leveraging advances in LEDs

The LEDs you see in the photos above look to be Cree XLamp XH-B LEDs, which are mid-power LEDs. Because they use a ceramic rather than a plastic package, they can handle heat better, and this probably contributes to the simple heat sink design. Having to use thermal grease to attach the pc board to the heat sink would add to the cost and complexity.

I previously speculated that the lamp used a mix of LEDs to achieve its high CRI, but the specs for the XH-B show that it’s the LED itself that delivers the 90 CRI – a much simpler approach, and simple = lower-cost. To save you jumping back to see the gorgeous spectral density for this lamp, I’ll repeat it:

Cree T8 linear LED lamps spectral power density with R9 values

The Cree T8 LED linear lamp has the highest color quality of any T8 replacement lamp, with a CRI of over 90.

In addition, the pc board is a single-layer board with no metal core, a significant cost savings. Most pc boards for LED lamps use metal core boards to get the heat away from the LEDs as quickly as possible, since heat is the major culprit in lowering lifetime for LEDs. Using these LEDs , which are well-spaced out (compared to an A19 replacement bulb) means that the LEDs are less fussy about heat sinking.

To show the impact that improvements in LED components have had on LED lamp design over the past four years, take a look at an LED T8 lamp from 2010, which was one of the first available. It consumed 18W to produce 820 lm. It used 288 plastic packaged LEDs – yikes. Not only does the higher number of LEDs add to cost and complexity, but more significantly it increases the number of components and solder joints, which escalate the failure rate.

2010 LED T8 lamp

I was surprised (I keep saying that) to find no power management IC in the lamp — the few components that your see here (and they are all on the one side) are passives. And, no electrolytic capacitors. How does Cree do this?

Fluorescent ballast compatibility

The real heavy-lifting in T8 fluorescent lights is done by the external ballast. For my tests I used a Philips Advance Optanium:

Philips Advance Optanium ballast info

Fluorescent ballasts operate at around 50k-60kHz, and this one lists its open circuit voltage (OCV) as 600Vac, the voltage applied between the lamp electrodes to generate the arc necessary to start fluorescent lamps. (Cree says that it expects these lamps to be compatible with 90% of the existing ballasts, including dimming ballasts.) So this LED lamp is going to have to accept 600Vac initially.

The first part of the circuit that I traced out was the LED matrix:

Cree T8 LED matrix

The LEDs are arranged in six strings of 20 LEDs each: 3 parallel strings, and then three parallel strings in opposite polarity. My thinking currently is that the circuit uses the LEDs as rectifiers with half on during the positive portion of the 50KHz cycle, and the other half on during the negative. Since the frequency is so high there is no perceivable flicker for the human eye.

Huh. Sounds good, but does the surrounding circuitry bear that out? (This is not a rhetorical question: I don’t know, but I think it does. I’m hoping for some suggestions below in the comments, gentle readers.)

Below are the circuits for the components on each end of the lamp. TP20 and TP21 are test points on the pc board. Here is the circuit that connects through R25:

Cree T8 LED J1_J3

…And here is the circuit for the L1/R26 section:

Cree T8 LED J4_J5

And here is a close-up of the ballast wiring:

Philips T8 ballast wiring

Dimming fluorescent ballasts keep the voltage across the lamps the same,but lowers the current, which works out well for LED lights since LEDs are current-controlled components.

Why the 5-year warranty?

I asked Jeff Hungarter, product portfolio manager at Cree, why these lamps have a 5-year warranty when most other Cree products have a 10 year. He explained it was because of the wide range of ballasts the lamps will have to work with. Because Cree has no control over the quality of the ballasts, it is being conservative about lamp lifetime. In the luminaire where Cree controls the whole power path, from wall plug to LED component, they offer the higher warranty.

He also explained how to compare fluorescent  vs LED lamp lifetimes. Although fluorescent lamps state lifetimes of 30,000 hours and longer — apparently making them a close contender to the Cree’s 50,000-hr lifetime — the failure modes are completely different and affect replacement estimates.

Fluorescent lamps fail catastrophically: One day they’re on, the next day they’re flickering a bit and then, poof, they’re completely dark. That 30,000-hour lifetime on a fluorescent lamp means that after 30,000 hours you can expect that half will have failed, and half will still be illuminated so, after 30,000 hours be prepared to have replaced half of your fluorescent lamps. LEDs, for the most part, don’t fail catastrophically — they gradually lose their light output, until after 50,000 hrs they are putting out only 70% of their initial output. This number is known as their “L70” life, and the light loss is barely discernable to human vision. After 50,000 hours your LED lamp will still be on, but it will only be putting out 70% of its original light. So yes, the greater lifetime of LED lamps can make them worth the additional cost in large commercial installations.

So that’s the gist of what I saw inside the Cree T8 LED linear lamp. With the exception of the hand-soldered wires, the manufacturing points to a high-reliability lamp, and we already know from the previous review that the color quality is superb. This lamp is aimed at the commercial market, especially installations where, as the July 2014 ban on T12 fluorescent lamps and low-efficacy T8 lamps goes into effect and some buildings will be replacing and re-ballasting fixture by fixture. (Refer again to the Cree T8 review article for more details on theT12/T8 legislation.)

Comments

  1. marc riel says:

    I appreciate the hard work!

  2. Excellent information and work, Margery! Thanks for your research and observations.

  3. coal_burner says:

    awesome teardown!
    since the Cree lamp doesn’t have it’s own current regulator, it seems its brightness will be determined by the ballast factor.
    i couldn’t quite make out from the pic if the ballast factor was printed on the ballast’ label.
    if so, what was it?
    could you test this lamp with a few other ballasts to see what the ballast factor does to the brightness & if there is any color shift?
    Thanks again for your cool reviews.

  4. Coal_burner, I agree I need to test this with other ballasts. However, I don’t expect to see a color shift since the performance of the Cree LEDs is stable over their operating range. The specs for the Philips ballast are here: http://www.usa.lighting.philips.com/pwc_li/us_en/connect/advance/assets/1-01_to_1-69_Atlas2012.pdf and the ballast factor ranges, based on the load from, about .9 to 1.05. It’s a pretty high-quality ballast.

  5. Fantastic writeup and product tear-down.

  6. Bruce Aleksander says:

    Wonderful work on the tear-down. I just wanted to add a note regarding the way fluorescent tubes fail. Although not noticeable to the average consumer, there is a gradual reduction in lumen output with fluorescent tubes that is similar to the loss we see in LEDs. By the time they fail completely, most fluorescent tubes are delivering much less than their initial light. We should note that these tubes still use the same power, dramatically affecting your lumen-per-watt calculations.

    There’s a decent paper on lumen maintenance in LEDs from DOE here: http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=8&cad=rja&uact=8&ved=0CE4QFjAH&url=http%3A%2F%2Fwww.pnnl.gov%2Fmain%2Fpublications%2Fexternal%2Ftechnical_reports%2FPNNL-22727.pdf&ei=Mvm6U7nADpGJogTu34LoAg&usg=AFQjCNHm4VSEq1WuKEV9wBrqnktYQXy8Yw&bvm=bv.70138588,d.cGU
    You can find a relevant graph on page 6 of this PDF file.

    While I’m not an expert on phosphors, I suspect that the phosphors simply fatigue and break down in much the same way, whether they’re LED or Fluorescent.

    Thanks again!

Speak Your Mind

*