This is my last "weekend" project. The idea came because I have recently built a new bigger UV exposure box for making big PCBs which employs four 15W F15T8/BL tubes. I was quite unhappy with the perfornamce, wheight and size of the conventional ballasts and starters that I tried at first, so I decided to get some IR2156 chips and try to drive the tubes in an optimum way.
The following pictures show what I've got after a couple of days of tweaking and thoughtful testing. There are also links to Eagle and EPS files of the ballast PCB (fortunately this is not one of my commercial projects so I can disclose everything).
Schematic:
PCB layout:
Some pictures showing the thing actually working (the ballast will be mounted inside the box with the timer):
The following picture shows the final version of the PCB of the own ballast after 2 minutes 30 seconds of UV exposure with the tubes being driven with a preliminary ballast version, and being developed for a few seconds in a two-week-old solution of plain NaOH and distilled water (from my dehumidifier) that had been already used several times!!. Note how the positive green layer has been completely dissolved, something that did never happen with my previous setup (2x11W with lousy capacitive ballast) and a new and clean NaOH solution, despite exposure times were in the range of 5 minutes!!
I can explain in detail how that kind of circuits work and how to wind the magnetics if someone is interested... I can also provide links to some PDFs worth reading.
Related files:
sch
brd
eps
The following pictures show what I've got after a couple of days of tweaking and thoughtful testing. There are also links to Eagle and EPS files of the ballast PCB (fortunately this is not one of my commercial projects so I can disclose everything).
Schematic:
An externally hosted image should be here but it was not working when we last tested it.
PCB layout:
An externally hosted image should be here but it was not working when we last tested it.
Some pictures showing the thing actually working (the ballast will be mounted inside the box with the timer):
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
The following picture shows the final version of the PCB of the own ballast after 2 minutes 30 seconds of UV exposure with the tubes being driven with a preliminary ballast version, and being developed for a few seconds in a two-week-old solution of plain NaOH and distilled water (from my dehumidifier) that had been already used several times!!. Note how the positive green layer has been completely dissolved, something that did never happen with my previous setup (2x11W with lousy capacitive ballast) and a new and clean NaOH solution, despite exposure times were in the range of 5 minutes!!
An externally hosted image should be here but it was not working when we last tested it.
I can explain in detail how that kind of circuits work and how to wind the magnetics if someone is interested... I can also provide links to some PDFs worth reading.
Related files:
sch
brd
eps
Jimbo:
My raw PCB material comes already coated with that green positive layer shown in one of the pictures, which is positive because it gets dissolved only where it has been exposed to UV. It's available single and double sided and I buy it from some Spanish company (http://www.metrotecnia.com/) but I'm sure that you are going to find better sources in USA. Check the image attached and the following papers:
http://www.cpes.vt.edu/_files/public/courses/modules/ballast.pdf (fluorescent basics)
http://www.irf.com/product-info/datasheets/data/ir2156.pdf (simple control IC)
http://www.irf.com/technical-info/refdesigns/cfl-2.pdf (example)
http://www.irf.com/product-info/datasheets/data/ir21592.pdf (dimming control IC)
http://www.irf.com/technical-info/appnotes/an-1021.pdf (dimming example)
http://www.irf.com/technical-info/refdesigns/cfl-3.pdf (exotic dimming mode)
http://www.irf.com/technical-info/appnotes/an-1013.pdf (series lamps)
http://www.irf.com/technical-info/appnotes/an-1039.pdf (parallel lamps)
http://www.irf.com/technical-info/appnotes/an-995a.pdf (funny passive PFC & stuff)
http://www.cel.com/pdf/appnotes/an1036.pdf (more funny stuff)
Choco:
The pins of the U15 are crossed as this is the easier arrangement for winding, see the picture. Polarity is unimportant. The purpose of that transformer is to force the same current to flow through all the filaments during preheating (and it proved to work very well). Initially I tried splitting in two halves the resonant capacitor and including the "center" filaments directly in series between those halves, but that didn't give satisfactory results because one lamp is always triggered before the other and that effectively bypasses one half of the resonant capacitor and detunes the system preventing the other lamp from being triggered most times.
The circuit with the component values shown is capable of keeping the lamps lighting down to 140V AC input, and is capable of starting with as little as 160V AC at 20ºC. It's adjusted to draw exactly 60 watts when peak rectified mains is 330 volts, which is the usual situation at my place. I also tried to keep light output as constant as possible against line fluctuations and I found that operating the filter at its resonant frequency (72Khz) provided the best results. At 140V AC the power drawn by the circuit drops to approx 40W and at 265V AC it rises to approx. 68W.
Also I investigated preheating a bit and adjusted it to cause the filaments to start barely glowing for just a fraction of a second before ignition with my typical mains input voltage. At higher voltages the filaments may suffer a bit as they are allowed to glow longer before ignition, and at lower line voltages they don't glow visibly.
My raw PCB material comes already coated with that green positive layer shown in one of the pictures, which is positive because it gets dissolved only where it has been exposed to UV. It's available single and double sided and I buy it from some Spanish company (http://www.metrotecnia.com/) but I'm sure that you are going to find better sources in USA. Check the image attached and the following papers:
http://www.cpes.vt.edu/_files/public/courses/modules/ballast.pdf (fluorescent basics)
http://www.irf.com/product-info/datasheets/data/ir2156.pdf (simple control IC)
http://www.irf.com/technical-info/refdesigns/cfl-2.pdf (example)
http://www.irf.com/product-info/datasheets/data/ir21592.pdf (dimming control IC)
http://www.irf.com/technical-info/appnotes/an-1021.pdf (dimming example)
http://www.irf.com/technical-info/refdesigns/cfl-3.pdf (exotic dimming mode)
http://www.irf.com/technical-info/appnotes/an-1013.pdf (series lamps)
http://www.irf.com/technical-info/appnotes/an-1039.pdf (parallel lamps)
http://www.irf.com/technical-info/appnotes/an-995a.pdf (funny passive PFC & stuff)
http://www.cel.com/pdf/appnotes/an1036.pdf (more funny stuff)
Choco:
The pins of the U15 are crossed as this is the easier arrangement for winding, see the picture. Polarity is unimportant. The purpose of that transformer is to force the same current to flow through all the filaments during preheating (and it proved to work very well). Initially I tried splitting in two halves the resonant capacitor and including the "center" filaments directly in series between those halves, but that didn't give satisfactory results because one lamp is always triggered before the other and that effectively bypasses one half of the resonant capacitor and detunes the system preventing the other lamp from being triggered most times.
The circuit with the component values shown is capable of keeping the lamps lighting down to 140V AC input, and is capable of starting with as little as 160V AC at 20ºC. It's adjusted to draw exactly 60 watts when peak rectified mains is 330 volts, which is the usual situation at my place. I also tried to keep light output as constant as possible against line fluctuations and I found that operating the filter at its resonant frequency (72Khz) provided the best results. At 140V AC the power drawn by the circuit drops to approx 40W and at 265V AC it rises to approx. 68W.
Also I investigated preheating a bit and adjusted it to cause the filaments to start barely glowing for just a fraction of a second before ignition with my typical mains input voltage. At higher voltages the filaments may suffer a bit as they are allowed to glow longer before ignition, and at lower line voltages they don't glow visibly.
An externally hosted image should be here but it was not working when we last tested it.
Are you using normal fluorescent tubes that have a phosphor coating on the inside? They look very white in your photo, but that may just be UV saturating the camera pixels.
If they are normal white tubes, you might want to see if you can get uncoated UV fluorescents. I have used both clear glass tubes and tubes with a blue-violet transparent filter coating as UV sources, and they are both much stronger than a white fluorescent tube.
If they are normal white tubes, you might want to see if you can get uncoated UV fluorescents. I have used both clear glass tubes and tubes with a blue-violet transparent filter coating as UV sources, and they are both much stronger than a white fluorescent tube.
Uncoated germicidal fluorescent tubes produce UV-C light and are not suitable for PCB making, they are only useful for EPROM erasing. My tubes are F15T8/BL which means that they have a coating that produces mostly UV-A light but also visible light. The colour as perceived by the eye is a mix of pale blue and white. As you have figured out, they completely saturate the webcam (and it's also quite uncomfortable to stare at them because the light intensity produced with the electronic ballast is quite high).
....yes, the connection of the U15 according your posting #4 is better than what I was seeing in your first posting, where windings seemed to be from 1==>2 and 3==>4....
The windings from 1==>3 and 2==>4 make much more sense! And yes, then polarity should be unimportant.
Have a look to your resonance circuit after the lamps are lighted up. Your adjustment to the resonance frequency of the filter is not anymore the resonance frequency after the lamps are ignited. As soon as the lamps ignite your load circuit is changing. The additional (mostly resistive) path through the lamp will bring down the dominant resonance frequency of the load circuit (L2, C14, lamp, filaments, T1, C12, C13).
At 72kHz this load circuit will show inductive behaviour to the half bridge, which is exactly what you need for nice softswitching of your snubbered (C9,C10) half bridge with some dead time.....
...should close my mouth now ...you probably know all this anyway...
The windings from 1==>3 and 2==>4 make much more sense! And yes, then polarity should be unimportant.
Have a look to your resonance circuit after the lamps are lighted up. Your adjustment to the resonance frequency of the filter is not anymore the resonance frequency after the lamps are ignited. As soon as the lamps ignite your load circuit is changing. The additional (mostly resistive) path through the lamp will bring down the dominant resonance frequency of the load circuit (L2, C14, lamp, filaments, T1, C12, C13).
At 72kHz this load circuit will show inductive behaviour to the half bridge, which is exactly what you need for nice softswitching of your snubbered (C9,C10) half bridge with some dead time.....
...should close my mouth now ...you probably know all this anyway...
Yes, the lamp adds variable damping to the LC tank resonator, but it's not resistive, it's negatively resistive That in turn helps a lot to keep somewhat constant lamp power against mains voltage fluctuations, because the Q of the LC tank increases for lower lamp currents (thus providing more voltage boost) and decreases for higher lamp currents (providing more attenuation).
Then, the trick to get the proper lamp power at the expected mains input voltage is to choose a suitable operating frequency for the inductance value employed, so that it does not provide too little or too much ballast effect.
I could have used a much lower frequency, like 35Khz, but I wanted to use some EF20 formers and core halves with 0.17mm gap that I had lying around, which provided quite limited energy storage. Then I estimated that I would need up to 12dB LC resonance gain to trigger the lamps and finally I found that the maximum inductance that the EF20 cores could provide while allowing for enough current to reach 12dB boost was 1mH (already considering the 80Khz range due to the low inductance available).
Anyway, the half bridge is always operating in resonant mode (achieved by choosing a dead time and a resonant capacitor value suited to drain current level at turn off), so the MOSFETs would dissipate the same power even if the operating frequency was 300Khz (but losses in the magnetics and the driver IC would increase quite dramatically). I learnt all that by experimentation
That in turn helps a lot to keep somewhat constant lamp power against mains voltage fluctuations, because the Q of the LC tank increases for lower lamp currents (thus providing more voltage boost) and decreases for higher lamp currents (providing more attenuation).Then, the trick to get the proper lamp power at the expected mains input voltage is to choose a suitable operating frequency for the inductance value employed, so that it does not provide too little or too much ballast effect.
I could have used a much lower frequency, like 35Khz, but I wanted to use some EF20 formers and core halves with 0.17mm gap that I had lying around, which provided quite limited energy storage. Then I estimated that I would need up to 12dB LC resonance gain to trigger the lamps and finally I found that the maximum inductance that the EF20 cores could provide while allowing for enough current to reach 12dB boost was 1mH (already considering the 80Khz range due to the low inductance available).
Anyway, the half bridge is always operating in resonant mode (achieved by choosing a dead time and a resonant capacitor value suited to drain current level at turn off), so the MOSFETs would dissipate the same power even if the operating frequency was 300Khz (but losses in the magnetics and the driver IC would increase quite dramatically). I learnt all that by experimentation
Maybe if I could work here at my home lab without moving, but I feel that it's not a valid option...
Anyway, since dimming conventional fluorescent lamps is so easy, something that I want to do one of those days is to experiment with somewhat harder stuff like metal halide.
I'm only happy when I work in different areas of electronics at the same time. For example, my next project is very likely to be an integrated switchmode PSU with a low voltage high current driver for diode-based medical lasers for a local company. It has to be smart enough to work with the mains available in any part of the world so I will use my heavy duty PFC and some of the stuff from my 0-15V 0-125A PSU that you may remember
The boss did not fully believe me when I told him that I can replace their really bulky 600 Joule capacitor bank and their water-cooled linear driver by a switchmode integrated driver having universal AC/DC mains input, capable of powering several types of laser diodes and almost the size of the commercial 48V 600W PSU that they are using now to power their arrangement. I want him to become puzzled with some nice happy-new-year present (a preliminary prototype).
Anyway, since dimming conventional fluorescent lamps is so easy, something that I want to do one of those days is to experiment with somewhat harder stuff like metal halide.
I'm only happy when I work in different areas of electronics at the same time. For example, my next project is very likely to be an integrated switchmode PSU with a low voltage high current driver for diode-based medical lasers for a local company. It has to be smart enough to work with the mains available in any part of the world so I will use my heavy duty PFC and some of the stuff from my 0-15V 0-125A PSU that you may remember
The boss did not fully believe me when I told him that I can replace their really bulky 600 Joule capacitor bank and their water-cooled linear driver by a switchmode integrated driver having universal AC/DC mains input, capable of powering several types of laser diodes and almost the size of the commercial 48V 600W PSU that they are using now to power their arrangement. I want him to become puzzled with some nice happy-new-year present (a preliminary prototype).
Eva said:
The basic idea is simple, but real life customer demands are sligthly higher.
Try to dim down to 1% of nominal lamp power without damaging the lamp... still showing stable light.... and ignite without visible flash ....all this in the temperature range of -25C...50C ambient +EMI+immunity+safety+handling all typical lamp failure modes+relamping feature+mass production suited design+reliability+costs.
The UC3871 is nice for that. I bought a few back when there was a Unitrode. Too bad they are still stupid expensive. I don't think that when it was designed they knew how severely the 2 transistor circuits would dominate the driving of these tubes. You could use any of the off-line ballast circuits you could find on the net if you were willing to stuff a step up converter in front of it. That would be a minimal design cost approach although not highest performance.
Thank's,
I need highest performance. I tested some industial balast which use only descrite: no RF noise (sinus output), maximum efficiency and brightness and maximum realable.
Regards
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.