Some Chinese designed LEDs are now using resistive series ballast

[Electric-Light]

One Energy Star certified, as well as non-certified integral ballast LED lamps were evaluated for their construction. They keep finding ways to carve cost, but not always without affecting important performance. These two were using a series linear current regulator.

Bridge diode x 1
160v 10.6uF x 1
3 chip LED elements x 16 connected in series to form 144v, 48 LED chip string.
China made transistorized linear amplifier regulator x 1.
Chip resistors to make 8.55 ohm external current setting



More thorough documentation on cost cutting, practical performance cutting linear LED ballast is explained in an ON SEMI datasheet.
http://www.onsemi.com/pub/Collateral/AND9041-D.PDF

Currently, Energy Star specification says
https://www.energystar.gov/products/spec/lamps_specification_version_2_0_pd

"All tests shall be conducted with the lamp connected to a supply circuit of rated frequency. For lamps with multiple operating voltages, the lamp shall be operated at 120 volts throughout testing. If the lamp is not rated for 120 volts, it shall be operated at the highest rated voltage. For dimmable or multi-power lamps, measurements shall be taken at the highest wattage setting listed for the model, unless otherwise specified."

LED Vf shrinks as it warms up, input current remains the same. The amount of energy flared off in transistor is increased.
As a way of satisfying power factor, the LED industry's answer was to reduce or eliminate DC bus capacitor which lowers light quality parameter by increasing flicker which is not adequately restricted in Energy Star criteria and allows LED products with inferior light quality to slip into the market.

This documentation is for a series linear drop out LED ballast chip found in an Energy Star certified Greenlite lamp. It has a much sparse implication on performance when deviated from ideal conditions.

The lamp, Greenlite 9W/LED/OMNI/D (UL 3UT1 E344320. LED board 10th week of 2017. Lamp case code F11 ) is rated for 800 initial lumen and they forecast the time needed for LED elements to deteriorate to 560 lumens (70%) is 15,000 hours. The lamp is listed for use in enclosed fixture. Test in a pancake fixture resulted in an evolution of objectionable chemical odor and the lamp surface reached 230*F.

According to the Chinese datasheet for linear drop out LED ballast, it is suggested for use with LED street lights as well as those controversial direct wire TLED. Vf change of LED as a function of temperature is very predictable and forward voltage setting can be fine tuned in 3v, 6v, or 9v increments for specific application which leaves manufacturing variations as the only variability. Laboratory testing for Energy Star certification as well as LM-79 testing are conducted on a regulated precision laboratory power supply set to 120.00v and lamp test environment is conducted in precisely temperature controlled space.

The very nature of linear regulator makes it become less efficient as the voltage difference between LED Vf and line voltage increases and loss of light output as the difference gets too small. Unfortunately, this allows LED lamps to be engineered around lab benchmark conditions and do well while becoming dimmer or inefficient in real life conditions.

I think every related lighting product qualifications and energy code such as Energy Star should expand to include +5%, -10% line voltage influence on consumption, case temperature and lumen output and flicker and dimming performance to avoid the temptation for tuning LED products for lab conditions. The practice of tuning products for a very narrow test conditions that is not relevant outside of a testing lab seems to be a common Chinese engineering practice to reduce cost by reducing performance. It is very common for many utility power service to have voltage at socket near +5%. This will increase dissipation and lower efficiency as well as raise temperature. Sickening burnt rubber or flux chemical odors I noticed on multiple samples of these Greenlite lamps, general increase in use of organic materials in LED lamps not used in conventional materials and toxic chemical concerns that surfaced on imported goods in the past, chemical analysis as well as health impact study of LED lamp outgas on sensitive domesticated animals such as canaries and humans maybe beneficial.

 

[mbrooke]

Honestly, thats incorrect lol. I've been using Greenlite bulbs and the light drops by 1/3 at the 1000 hour mark. :happyno:


But anyways- I don't think those chips are resistive ballasts. They act more like dimmers rapidly opening and closing, correct? While not perfect, they are better than capacitive droppers.


Let me read through the docs you posted.

 

[Electric-Light]

Honestly, thats incorrect lol. I've been using Greenlite bulbs and the light drops by 1/3 at the 1000 hour mark. :happyno:


But anyways- I don't think those chips are resistive ballasts. They act more like dimmers rapidly opening and closing, correct? While not perfect, they are better than capacitive droppers.


Let me read through the docs you posted.

I meant to say that the China chip for the Greenlite lamp has much sparse description of what happens as you deviate from ideal. The ON SEMI datasheet does a much better job of explaining. These most definitely are linear dropout current regulators. It most definitely is a linear regulator, which means resistive ballast. This is why change in voltage has so much effect on this design.

 

[mbrooke]

I meant to say that the China chip for the Greenlite lamp has much sparse description of what happens as you deviate from ideal. The ON SEMI datasheet does a much better job of explaining. These most definitely are linear dropout current regulators. It most definitely is a linear regulator, which means resistive ballast. This is why change in voltage has so much effect on this design.

If its a resistive ballast, then where is the heat sink or the heat sink requirements? A 9 watt LED would mean that chip gives off about 4 to 5 watts of heat. That amount of heat would burn such a small circuit board to a crisp. As is, why not simply use a plain resistor instead of a expensive chip?

 

[Electric-Light]

If its a resistive ballast, then where is the heat sink or the heat sink requirements? A 9 watt LED would mean that chip gives off about 4 to 5 watts of heat. That amount of heat would burn such a small circuit board to a crisp. As is, why not simply use a plain resistor instead of a expensive chip?

It's soldered on the metal core circuit board with the LED elements. It goes into a lamp base which heats up to 230 deg F. A plain resistor doesn't change in value. This thing adjusts resistance to hold current constant (as long as the voltage supplied is high enough to allow LEDs to conduct) through varying the resistance. Chips like this are cheap, especially when its entirely made in China.

This kind of inferior crude ballast is cheap but it is full of compromises. The power supply topology common for fluorescent lamps are vastly superior in almost every way and people should be aware that if they're sold ballast bypass retrofit, the light quality drops.

A fluorescent electronic ballast using boost active PFC achieves 0.99 PF, <10% THD, fantastic power regulation (constant input power, constant lamp power from 108 to 305v) while delivering steady light with a flicker index of 0.01 or so which is less than incandescent, all while maintaining >90% efficiency.

 

[mbrooke]

It's soldered on the metal core circuit board with the LED elements. It goes into a lamp base which heats up to 230 deg F. A plain resistor doesn't change in value. This thing adjusts resistance to hold current constant (as long as the voltage supplied is high enough to allow LEDs to conduct) through varying the resistance. Chips like this are cheap, especially when its entirely made in China.

This kind of inferior crude ballast is cheap but it is full of compromises. The power supply topology common for fluorescent lamps are vastly superior in almost every way and people should be aware that if they're sold ballast bypass retrofit, the light quality drops.

A fluorescent electronic ballast using boost active PFC achieves 0.99 PF, <10% THD, fantastic power regulation (constant input power, constant lamp power from 108 to 305v) while delivering steady light with a flicker index of 0.01 or so which is less than incandescent, all while maintaining >90% efficiency.

Do you have any info on how these chips works? I always assumed that they simply switched on and off very rapidly to maintain a constant current rather than mimicking a variable resistor.

 

[winnie]

The simplest approach is a combination of linear regulator and a resistor. The full output current passes through the resistor, and you regulate the value of voltage across the resistor. Constant V across R means constant I.

The 'linear regulator' could even be a simple transistor.
https://en.wikipedia.org/wiki/Current_source#Simple_transistor_current_sources

Linear current sources make good sense in small battery powered devices where the parasitic loads of switching regulators eat up any efficiency improvements.

I agree with Electric Light here: the testing standards should include normal and expected supply voltage variation.

-Jon

 

[gadfly56]

...I think every related lighting product qualifications and energy code such as Energy Star should expand to include +5%, -10% line voltage influence on consumption, case temperature and lumen output and flicker and dimming performance to avoid the temptation for tuning LED products for lab conditions. ...

Is there a way to provide public input to the standards? You've clearly given this more than a little thought and seem to have the technical chops to argue your case. You'd be doing something about the weather instead of just complaining about it. :happyyes:

 

[steve66]

Some Chinese designed LEDs are now using resistive series ballast

Sorry, but I think your initial premise is flawed. This doesn't look at all like a simple series resistor (or even a simple linear regulator) to me.

http://www.onsemi.com/pub/Collateral/NCL30051-D.PDF

I believe the power that would be lost by a linear regulator would make it impossible to use a linear regulator (or series resistor) for anything over a few watts without including a heavy, bulky transformer.

It looks to me like they have gone to extreme lengths to squeeze every drop of efficiency out of the total chip/driver/LED package.

 

[Electric-Light]

Hey Roger,
I want to clarify with you that what I posted is not anything that is privileged. It is publicly available on the web following a part number search and the part number was obtained from physical inspection of LED lamp available through the market.

 

[Electric-Light]

Sorry, but I think your initial premise is flawed. This doesn't look at all like a simple series resistor (or even a simple linear regulator) to me.

http://www.onsemi.com/pub/Collateral/NCL30051-D.PDF

I believe the power that would be lost by a linear regulator would make it impossible to use a linear regulator (or series resistor) for anything over a few watts without including a heavy, bulky transformer.

It looks to me like they have gone to extreme lengths to squeeze every drop of efficiency out of the total chip/driver/LED package.

Well the China datasheet is stamped "CONFIDENTIAL" for some reason so I had to remake the post without using screen shots or direct link to it.

Nah, that's not the one I linked. This is the one I referenced.
http://www.onsemi.com/pub/Collateral/AND9041-D.PDF

I have not come across LED lamps using linear dropper ballast until this year.

The no name LED lamp had an internal construction like this, which was determined wholly through product tear down.

The variable resistor dynamically changes to adjust current while flaring off energy in the package.
This is a more generic equivalent circuit. That resistor sets the current.


The Greenlite 9W/LED/OMNI/D* uses some China chips SM2082D and SM2082ED, although they work on the same principle.

The increase in line voltage causes increased voltage drop at the linear dropper LED ballast which increases thermal dissipation at the lamp which lowers LED element efficacy.

 

[mbrooke]

@Roger:

I'm with Electric-Light on this one. A lot of times when you Google the part numbers of the chips on LED fixtures you get plenty of declassified PDFs that say "confidential" An example, click on the first few PDFs:


https://www.google.com/search?sourc...sy-ab..0.5.803...0j0i131k1j0i10k1.7UdoHY9Ngj8

There are also a number of videos on You Tube where people will dismantle LED lamps, read the number off the chip and then download similar documents right off line in regards to explaining how the chip works.

See point 3:11 showing the exact same doc:

https://www.youtube.com/watch?v=KKd2L9Exw0M

 

[mbrooke]

Sorry, but I think your initial premise is flawed. This doesn't look at all like a simple series resistor (or even a simple linear regulator) to me.

http://www.onsemi.com/pub/Collateral/NCL30051-D.PDF

I believe the power that would be lost by a linear regulator would make it impossible to use a linear regulator (or series resistor) for anything over a few watts without including a heavy, bulky transformer.

It looks to me like they have gone to extreme lengths to squeeze every drop of efficiency out of the total chip/driver/LED package.

I am probably wrong, but I see it the same way. I don't think thats a variable resistor but rather a very fast switching variable "triac" It opens the gate for a drop in voltage and closes it when voltage spikes in an effort to modulate a near constant current. That voltage threshold determined by the resistor tacked across one leg of the chip.

 

[winnie]

I agree with Electric-Light's analysis of the lamp.

120V gets rectified to give about 170V DC.

The LEDs are arranged to have a voltage drop of about 144V.

A _linear regulator_ is arranged to drop the difference between the rectified mains voltage and the LED forward voltage.

This is _not_ a PWM style switching regulator because the LEDs are non-linear and if you rapidly turned them on and off the peak current would be much higher than desired. This is not like PWM dimming of a resistive load like an incandescent lamp.

A switching regulator for LEDs require the use of energy storage such as inductors and a much more complicated circuit.

The greater the LED voltage as a fraction of the rectified mains voltage, the greater the efficiency of the driver circuit.

At the nominal 120V supply voltage, the driver circuit is about 86% efficient, which is probably as good as you will get with a switching regulator at this power level. You have 24V * load current dissipated as heat by the regulator and (144V * load current - light output) dissipated as head by the LEDs. Put the LEDs and the regulator on the same heat sink, and it just runs a bit warmer.

With this circuit, the power consumption is essentially linear with supply voltage, with constant power dissipated in the LEDs string and the difference dissipated as heat by the _linear_ regulator circuit. If the supply voltage is 10% high, then the heat dissipated by the regulator circuit almost doubles. Still less than what the LEDs are dissipating, but a very significant change, and one that needs to be understood by proper testing.

With this circuit, if the supply voltage is 10% low, the system probably stays in regulation, but you might see an increase in flicker, again something that needs to be understood with proper testing.

Sorry for making this so long...I don't have time to edit it shorter.

-Jon

 

[winnie]

The variable resistor dynamically changes to adjust current while flaring off energy in the package.

One small error in your schematic: I expect the full lamp current (60mA) is passing through the sense resistor, with the ballast chip measuring (and regulating) the voltage across the resistor.

-Jon

 

[gar]

170909-2043 EDT

From post 11 there is a part number BP5131H in the circuit diagram. I can only find via Google one Chinese datasheet that is somewhat readable in English. This part number is probably a linear current regulator.

What is a linear current regulator. Fundamentally it is an adjustable resistor that is continuously adjusted by control circuitry that senses the voltage drop across a current sensing resistor that is in series with the current thru the adjustable resistor.

In the shown circuit this current sensor is 8.55 ohms. The stated controlled output current is 0.06 A. At 0.06 A the voltage drop is 0.51 V. In the datasheet there appears to be an 0.6 V value given for this control point. These seem to correlate.

The datasheet specifies 1.25 W maximum dissipation, and a maximum junction temperature of 155 C. Maximum current is listed at 80 mA (0.08 A).

In a series pass regulator, whether voltage or current, or in between, the power dissipated in the controlling element (the adjustable resistor) is the voltage drop across the regulator times the current thru the regulator.

First question is can I physically build the circuit (mechanical mounting) to dissipate 1.25 W at a realistic ambient temperature and stay below a 155 C junction temperature? It is a very small chip and I doubt it for what are probable chip case temperatures. I would never design for a 155 C junction temperature if I want long term reliability.

The circuit is moderately below the the maximum current rating and thus this is not a major concern.

The next problem is the needed power dissipation. Once the LEDs stabilize, then at constant current their power dissipation is constant. This power dissipation we are not concerned with except as it increases the ambient temperature of the regulator chip.

If we assume that worst case load on the regulator is 144 V at 0.06 A (meaning minimum voltage and maximum current), and maximum voltage input to the regulator is 120/0.707 = 170, then the maximum power dissipation in the regulator is (170-144)*0.06 = 1.56 W. To start with this is above the maximum rating, and for realistic heat sinking conditions and maximum junction temperature you would not wnat to even operate at a 1.25 W level.

It gets even worse when you make a realistic assumption that a bulb might work at 125 to 130 V. At 130 V the series pass dissipation is ( (130/0.707) - 144 ) * 0.06 = 2.4 W.

.

 

[Electric-Light]

:ashamed1:

steve66's comments based on what was provided is correct. I pasted the wrong link.

This is the linear ballast type LED driver I was talking about the whole time.
On Semiconductor Design Note – 05079/D

Something you will see in one of the tables is that their example circuit exhibits 9% efficacy loss when you go from the lower end of bandwidth to upper end of ANSI tolerance. 114 vs 126v. The increase in power is increase in ballast power dissipation which increases temperature. This means users closer to substation would chronically experience higher power consumption from these lamps as well as accelerated wear.

Someone evaluated the same model Greenlite integral-ballast LED lamp in this video.
https://youtu.be/zJWssuCGkC8?t=360s

the unit he evaluated has a switch mode converter ballast that drives two parallel strings of 5 three chip LED elements in series.

My sample drives 16 three-chip LED elements in series through a linear drop out ballast. They both have the exact same model number and there is no way to tell them apart. I don't believe the criteria for Energy Star considered the potential of cost engineers coming back around to chip away realistic durability and performance from the product design and leaving behind the absolute minimum to meet the required by specifications. When you go to Energy Star website, there are multiple Greenlite 9W 800 lm LEDs but do not allow positive identification of these samples. Two different date codes of these lamps were energized simultaneously and they have a dramatically different dimming curve which means they dim totally differently alongside on the same dimmer because they have different integral ballasts.

I think it's a concern that a product that has gone through a cost reduction engineering change so drastic that the there's nothing in common between the two products and they're still allowed to bear the Energy Star logo under the same model number. The version he reviewed used a 130C capacitor, the version I evaluated used a 105C capacitor. Presumably both of these products meet the specs under lab conditions but they provide different realistic performance despite the same model number. You can not tell the two apart without breaking them open. What prompted an extensive evaluation was the fact the several samples of the new version gave off burning plastic smell and case temperature of 230*F.

ANSI allows +/- 5% from nominal and I think its reasonable to expect the same performance within the parameter and currently, Energy Star standard allows any amount of flicker only for LED lamps as long as it is happening at 120Hz or above and there are too many LED products out there that take this literally and omit smoothing capacitor entirely.

With linear ballast, when line voltage goes up:
LED current is held constant. (constant ballast factor)
Power consumption and dissipation goes up (Lowering BEF, increasing Tc temp)
Indirectly, the lamp temperature goes up which lowers Vf (further reduction in efficacy, as well as life)

Several issues here:

Energy Star does not limit flicker in a meaningful way. It mandates HF operation for CFL but holds LEDs to the lesser standard of direct AC line flicker which gives nearly the same flicker as a 60Hz driven red neon lamp.

Products are allowed go under a total gut change and bear Energy Star logo and recycle model # without letting consumers know of difference.

No limit on performance or durability deviation when operated above or below nominal voltage but within the ANSI tolerance.

 

[mbrooke]

170909-2043 EDT

From post 11 there is a part number BP5131H in the circuit diagram. I can only find via Google one Chinese datasheet that is somewhat readable in English. This part number is probably a linear current regulator.

What is a linear current regulator. Fundamentally it is an adjustable resistor that is continuously adjusted by control circuitry that senses the voltage drop across a current sensing resistor that is in series with the current thru the adjustable resistor.

In the shown circuit this current sensor is 8.55 ohms. The stated controlled output current is 0.06 A. At 0.06 A the voltage drop is 0.51 V. In the datasheet there appears to be an 0.6 V value given for this control point. These seem to correlate.

The datasheet specifies 1.25 W maximum dissipation, and a maximum junction temperature of 155 C. Maximum current is listed at 80 mA (0.08 A).

In a series pass regulator, whether voltage or current, or in between, the power dissipated in the controlling element (the adjustable resistor) is the voltage drop across the regulator times the current thru the regulator.

First question is can I physically build the circuit (mechanical mounting) to dissipate 1.25 W at a realistic ambient temperature and stay below a 155 C junction temperature? It is a very small chip and I doubt it for what are probable chip case temperatures. I would never design for a 155 C junction temperature if I want long term reliability.

The circuit is moderately below the the maximum current rating and thus this is not a major concern.

The next problem is the needed power dissipation. Once the LEDs stabilize, then at constant current their power dissipation is constant. This power dissipation we are not concerned with except as it increases the ambient temperature of the regulator chip.

If we assume that worst case load on the regulator is 144 V at 0.06 A (meaning minimum voltage and maximum current), and maximum voltage input to the regulator is 120/0.707 = 170, then the maximum power dissipation in the regulator is (170-144)*0.06 = 1.56 W. To start with this is above the maximum rating, and for realistic heat sinking conditions and maximum junction temperature you would not wnat to even operate at a 1.25 W level.

It gets even worse when you make a realistic assumption that a bulb might work at 125 to 130 V. At 130 V the series pass dissipation is ( (130/0.707) - 144 ) * 0.06 = 2.4 W.

.

Ok- I guess I was mistaken then the whole time. Then again I've never really researched these chips in depth outside of translating Chinese which works out poorly in general let alone STEM concepts. Resistor it is instead of dimmer.

But- I am not saying any of you guys are wrong- but wouldn't a 7 watt poses a chip that gives off 3.5 watts of heat? I can't picture such a tiny chip dissipating that much heat into the cap of the lamp, that is unless its not dissipating that level?

 

[gar]

170912-2449 EDT

Electric-Light:

It looks like you have provided some misleading information.

In post 11 you showed a circuit that could imply a true linear regulator, but the capacitor implies an RC time constant of possibly 2.5 milliseconds.

You also provided a reference to a discussion on a regulator that was possibly not the BP5131H of post 11.

Then in post 17 you reference an ON datasheet that does not map to the device in your circuit of post 11. The ON chip and circuit appear to be a form of pulse width modulation, far lower power dissipation than a linear circuit.

In your post 17 photograph it appears that there is a larger power device, possibly an FET, and not just a single chip regulator.

Do some circuit tracing and provide an accurate circuit diagram. Also use a scope to analyze circuit operation.

.

 

[Electric-Light]

170912-2449 EDT

Electric-Light:

It looks like you have provided some misleading information.

In post 11 you showed a circuit that could imply a true linear regulator, but the capacitor implies an RC time constant of possibly 2.5 milliseconds.

You also provided a reference to a discussion on a regulator that was possibly not the BP5131H of post 11.

Then in post 17 you reference an ON datasheet that does not map to the device in your circuit of post 11. The ON chip and circuit appear to be a form of pulse width modulation, far lower power dissipation than a linear circuit.

In your post 17 photograph it appears that there is a larger power device, possibly an FET, and not just a single chip regulator.

Do some circuit tracing and provide an accurate circuit diagram. Also use a scope to analyze circuit operation.

.

The schematic with BP5131H is some no name lamp, which uses a 10.6uF parallel capacitor.

I provided the ON-SEMI page just for information. It does a much better job of explaining things than the sheets available for the Chinese chips relevant to the two products. The demo circuits they used for the graphs choose not to use a smoothing capacitor at all and voltage drops all the way to zero each half cycle. Remember that the load is a long string of LEDs. The PWM like effect you're seeing is from LEDs themselves commutating. The flat top is created by the linear regulator adjusting to flare off the top part of each half cycle. And it drops again when each half cycle goes below the on-state for LED stack.

The Greenlite9W/LED/OMNI/D* pictured uses
SM2082ED (IC looking thing)
SM2082D (3 terminal device)
This lamp used RTV and adhesive very liberally and it didn't yield easily and got quite banged up in the process.

The 8 pin (really 6 pin + 2 unused pins) appears to have two devices in one one package and I believe the purpose is help with diimmer compatibility to keep conducting even when LE Diodes have not reached Vf. I couldn't find the English version so I have no idea what they're saying, but the Chinese version shows formulas that prove it's a linear dropper. The 2082D is a a single channel linear driver.

Anyways for the purpose of relevance to this forum, the linear drop out topology accommodates line voltage regulation by flaring off excess voltage as a ballast loss. It's common for utility to aim for the top of the allowance so they make sure end the furthest point still gets 114v incoming.

The behavior in response to sag or swell depends on where the balance point is tuned for. The unique thing is that raising the line voltage does not increase the current. It just raises the ballast's share of voltage compared to LEDs, so the ballast factor stays the same while BEF goes down.