Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Fluorescent Ballast IC problem

Not open for further replies.


Jun 22, 2008
Reaction score
Trophy points
Activity points

I am confused about the IR2520D offline fluorescent ballast control IC.

IR2520D Datasheet

I am confused about the way in which this IC is said to change its switching frequency in order to bring about Zero-Voltage-Switching (ZVS) of the two Mosfets.

-I am referring ONLY to cases when the lamp is lit.

On page 8 (bottom right) of the datasheet, it is explained that if the ballast operates just below the resonance frequency, then NON-Zero-Voltage-Switching occurs.

The IR2520D is said to be able to detect this non ZVS condition , and will increment its switching frequency upwards until it gets into ZVS operation.

However, -when the lamp is lit, the above statement is completely false…

Changing the frequency upwards will NOT bring about ZVS.

When one of the FETs switches off, the half-bridge node voltage slews to the opposite rail in about 300ns.
When the half-bridge node voltage has reached the opposite rail, it will then turn on the intrinsic diode in the fet at that rail.
This intrinsic diode then comes ON and “shorts” out the FET ….
-then, when the FET does eventually get switched ON, (after the dead-time)
it obviously switches ON at virtually zero voltage, and ZVS has occurred.

ZVS is only brought about by making sure that the dead-time between the FETs is long enough for the half-bridge node voltage to slew to the opposite rail, and thus turn ON the intrinsic diode of the opposite FET.

When a fet switches off, the half-bridge node voltage MUST slew to the opposite rail and turn on the diode there, ..-because the ballast inductor’s current MUST keep flowing…….as you will know.

So why does the IR2520D datasheet claim to adaptively move toward ZVS operation by incrementing its switching frequency.?

Incrementing the switching frequency will NOT bring about ZVS in a lit lamp.

Here is an LTSpice equivalent circuit of a fluorescent ballast as on page one of the IR2520D datasheet

LTSpice fluorescent ballast equivalent circuit.
(C2 changes switching frequency)
(Unfortunately max dead-time available in LT3721 is 300ns)

Your Download-Link:fluorescent ballast.txt

(need to convert from txt to .asc for LTSpice)

So why does the datasheet claim that the IR2520D varies its frequency to bring about ZVS?

Varying the frequency would NOT “bring about” ZVS.

The statement at the bottom of page 6 of the IR2520D datasheet is thus completely false………….

“If non-ZVS is detected then V decreases and frequency increases to maintain ZVS”

The claim of “Adaptive zero-voltage switching (ZVS)” at the top of the datasheet is also completely false.

All the Application Notes regarding this IC also have similar incorrect statements.

I am actually therefore wondering -what is the real “modus operandi” of the IR2520D ?

When the lamp is properly lit, what actually does determine the switching frequency?

The datsheet gives this explanation about the ZVS criterion tested by the VCO circuit:
During run mode, if the voltage at the VS pin has not slewed entirely to COM during the dead-time such that there is voltage between the drain and source of the external lowside half-bridge MOSFET when LO turns-on, then the system is operating too close to, or, on the capacitive side of, resonance.
In my opinion, the explanation is correct.

On the "capacitive side", the current phase is behind voltage. So, when the switch opens, a load current still exist, causing the output voltage to slew to the opposite limit. When the switching frequency decreases, this doesn't happen any more.


here is a better version of the above ballast simulation in LTSpice.

this one is in the nicer .asc form

**broken link removed**


Added after 15 minutes:



Your quotation, i believe, only applies to when the lamp has failed or failed to strike.
-not when the lamp is lit

if you run the simulation, you would see that the half-bridge node has slewed to the opposite rail well before the minimum dead time offered by the IR2520D.

in short, when the lamp is lit, there should be no problem with getting ZVS, and no need to change frequency to get it.

so i can't understand why the datasheet purports to do this.?

Added after 50 minutes:

QUOTE On the "capacitive side", the current phase is behind voltage. So, when the switch opens, a load current still exist, causing the output voltage to slew to the opposite limit. When the switching frequency decreases, this doesn't happen any more. UNQUOTE

-as you would be able to see by playing with the simulation, in a "lit" fluorescent lamp, switch current is always pretty well at its maximum when either FET turns OFF.

-No matter what frequency you are at. (within reason)

when the lamp is lit...
whether you are above or below the resonant frequency, or on it, it makes no difference, you will always get ZVS with the IR2520D's minimum dead-time of 1us.

so i don't understand why the datasheet of IR2520D and the many corresponding Application Notes all puport to do some kind of adaptive operation to get ZVS when the lamp is lit.

-changing the frequency will not significantly effect the half-bridge node's voltage slew rate when the lamp is lit.

and i can't understand why the datasheet infers that it does.

changing the frequency of a lit lamp ballast to get ZVS is a complete misnomer. (its nonsense)

-but the datasheet has it that way.

Your quotation, i believe, only applies to when the lamp has failed or failed to strike. -not when the lamp is lit
It exactly applies for the latter case, thus it's called run mode in contrast to e.g. fault mode. I think the datasheet is very clear in this regard. I also added my own explanation, why it can work this way.

I realized, that the low Q LC circuit dimensioning in your simulation setup actually can't resonate and thus not achieve ZVS at any frequency. The load current is almost inductive and switching occurs at maximum current.

Unfortunately, I can't determine if IRF's assumptions of fluorescent lamp operation or your dimensioning are unrealistic. But if you are able to achieve a resonant operation slightly above the resonance frequency, the variable frequency ZVS control concept would work.

Thankyou FvM, i appreciate your reading of the datasheet, and your comments.

My above simulation was poor as the dead time of that chip is limited at 300ns.

I have done three more much, much clearer ballast simulations - each with 1us dead time switching at 38KHz (below resonance) switching at 50KHz (at resonance) switching at 62KHz (above resonace)

they are all resonant at 50KHz.

though it must be said, when the lamp is lit, as in these simulations ,the resonant behaviour is all but damped out.

you can see that incrementing the frequency does not bring about ZVS.

-so the datasheet must surely be wrong.

ZVS is all down to the slew rate of the half-bridge node.

this slew rate depends on...
A. The peak inductor current (i)
B. The charge pump capacitor
c. The fets' DS capacitance.

...from the above, you can see that the slew rate will be quick...since

dv/dt = i/c

where i is peak inductor current and c is the capacitances as explained above.

here are the simulations on LTSpice which you can simply run and watch if you wish to.

38KHz ballast:

50KHz ballast:

62KHz ballast:

..if you run the above, you will see how incrementing the switching frequency can not bring about ZVS, and the datasheet must surely be in error?

so the datasheet must surely be wrong
It's not particularly a datasheet problem. The datasheet doesn't specify CRES and LRES values. To achieve resonant operation with sufficient Q, the LC impedances must be considerably lower than in your circuit. I don't know the dimensioning of existing ballast circuits.

I can assure most definetely, that ZVS does occur at frequencies above and below "resonance" when the lamp is lit.

the above LTSpice (free download) simulations at 38KHz, 50KHz and 62KHz show this.

I appreciate that the terminology used in the datasheet, and by myself, is misleading.....

.....when the lamp is lit the "resonant" capacitor (across the lamp) is not actually resonant at all. -it is only resonant during "pre-heat" and "ignition-ramp" phases.

FvM you are absolutely right in saying its more or less inductive load current when the lamp is on

The capacitor across the lamp is not actually needed when the lamp is lit and would "ideally" be switched out.

here is some component values of a 20W ballast

-schematic on page 2 and component values on page 16. switching frequency is 4E9/82K Hz (from Page 9)

here is another CFL ballast

**broken link removed**

-schematic page 3, component values page 4

I guess, you noticed, that the Fairchildsemi applications gives a similar explanation of ZVS operation as IRF.
As a difference, they control the switch dead time and keep a fixed frequency in running mode. The lamp impedance
is said to be about 500 ohm, see figure 21, which gives a somewhat higher Q than your simulation circuit.

Thankyou FvM, -i see you noticed how the FAN7710 works.

I started using the IR2520D , but when i realised that it appears to vary its "running" (lamp lit) switching frequency for bizarre reasons (as i explained above), i ditched the IR2520D and and am now using the FAN7710.

I am using a PL-L 18W tube, which the ballast design software on website gives as 228R when lit.

I don't know if any reader had chance to run the triplet of simulations of my post dated 08 Oct.

If you did, you would see that the IR2520D datasheet's claim to be able to give ZVS for a lit lamp by varying the frequency, is nonsense.

The fact is you get ZVS anyway, when the lamp is lit, as long as your dead time is long enough.

I am still wondering what is the reason for the IR2520D to vary its frequency when the lamp is lit and functioning properly.

I already reported, that you don't get ZVS operation with the present resonant circuit dimensioning (LRES, CRES, 228R load) beacuse of the low Q, I don't need more simulations to see this.
In my opinion, your complains are justified, because the operation mode claimed in the IRF data sheet apparently can't be achieved with their reference design.

You can achieve ZVS with a higher Q circuit, but I don't know, if it also results in a meaningful ballast design.

My doubts are however already starting with the actual lamp characteristic. I know, that a 50 Hz operated CFL has a constant voltage gas discharge characteristic. I don't fully understand, why a high frequency lamp apparently has a almost resistive characteristic, may be it's a matter of limited ion mobility.

Thankyou for looking in to this.

I can assure that ZVS always occurs when the lamp is lit.

I will get the waveform close-ups of the half-bridge node voltage along with the DS voltage of the FETs and post them here.

-It is clear that ZVS occurs.

(and as stated above -there is no need to change the switching frequency to bring about ZVS, becauese ZVS occurs anyway)


Here are "lit" fluorescent lamp waveforms in a circuit which is resonant at 50KHz (please remember "resonance" is only relevant at ignition stage)

The waveforms are for 38KHz, 50KHz and 62KHz,

they show mosfet gate-source waveform with drain-source waveform....

In each case, the order is
1. bottom fet coming on
2. bottom fet going off
3. top fet coming on
4. top fet going off

You will see ZVS in every case.

-showing that (contrary to IR2520D datasheet) changing the frequency does not "bring about" ZVS -because you get ZVS anyway (as long as your dead time is long enough)













here are the simulation files for the above in LTSpice.





  • fluorescent
    6.3 KB · Views: 60
Last edited by a moderator:

Optimizing the switching operation involves both edges. ZVS is only relevant for switch-on. The other aspect is to switch-off at low current. If I understand right, your waveforms show, that the load current is inductive in all three cases. But you get a different amount of switch-off losses depending on the load current during switch-off.

The purpose off frequency variation is to stay at the inductive side of resonance but reduced switch-off current. A certain circuit Q is required however to make the frequency variation effective.

thankyuo for looking in further,

though i must assure that this circuit, above, always switches off at its peak current....that is its modus operandi.......we cannot make this better by changing the frequency.

it switches off at peak current, but the rise of voltage across the switch is delayed by the DS capacitance and the charge pump capacitor -so the switch off loss is not at all bad.

if you open up a CFL, you will see tiny TO-92 transistors, evern in 240Vrms 15 Watt CFL bulbs.

the reason they can get away with this is the pretty well zero switching losses.

you can also notice how these CFL's have no controller in controller to change the its not needed.....yuo get pretty well lossless switching at whatever frequency when the lamp is lit.

thats why the IR2520D datasheet is wrong when it says it varies the frequency of the lit lamp to get ZVS..........this is a get ZVS anyway....

FvM you are correct when you speak of it as inductive when the lamp is lit...this is right....this is how its supposed to be,........the datasheet is wrong.

Not open for further replies.

Part and Inventory Search

Welcome to