Just been pondering on the flourescent dimming idea. I have made a dimmable flourescent fitting some time back using simple components and of course heating the electrodes. But what I don't get is that commercial dimmable ballasts cannot dim to 0%. The system I made can dim the tube up/down to 0% although it's not fully lighted but it still emits light so Y it cannot be done in comercial ballasts?

They may not want the user to see visual artifacts from dimmed fluorescents : Glowing cathodes, uneven light level in multiple lamp setup in the low dimmer settings, and so on. The user may not like that and generate bad name for their system

They may want to avoid extra ballast complexity related, esp when speaking of electronic ballasts (EWOL protection must work on all dimmer levels but how to make it so it won't false trip etc) - Its not really too hard but they may just want to avoid the development if it won't see much market....

They may not want to kill market share for other lamps like incandescent/halogen and LED which they promote using dimability as hteir advantage over fluorescent

Just been pondering on the flourescent dimming idea. I have made a dimmable flourescent fitting some time back using simple components and of course heating the electrodes. But what I don't get is that commercial dimmable ballasts cannot dim to 0%. The system I made can dim the tube up/down to 0% although it's not fully lighted but it still emits light so Y it cannot be done in comercial ballasts?

I don't know, what concept was used on your ballast, but there are few limitations for mass-production:
- It should remain simple, so suffice with frequency controlled inverter (maybe in a closed loop, but the steering element is still only the frequency). So independent inverters for filaments and arcare not acceptable - just too complex. The main reason is to keep cost and failure modes under control.
Maybe your ballast fulfilled this, you haven't specified it's concept.

- The arc should be stable at all power levels and over the complete lamp life, so no flickering at all. Here all the frequency control based systems tend to become unstable. It may be somehow treated by the regulation loop dynamics, but usually not below few percent.

- It should be robust against compopnent tolerances and their combinations, all that without any adjustment element. This is usual problem with many home-made designs: They are somehow tuned to work on the prototype, but when such design would be produced in quantity, quite significant percentage of the units would not perform as expected. So the design need quite significant margins. Well, with the previous it mean when one piece does work down to fraction of a percent, such extreme dimming should not be allowed, because few percents of the mass produced units won't be stable with that setting due to unfortunate component tolerance arrangement.
Here belong even lamp-to-lamp variations, across all makers.

- When igniting the lamp on a frequency sweep resonant start system, it will always create a flash with the low dimming setting (when the arc ignite, the frequency correspond to way higher power than desired). Lower dim setting, more disturbing is the ignition flash. So when the ballast is not allowed to be dimmed so low, the flash does not appear so bad.

- Lowering the light output, the power income stay about constant (about 20% of the full power), what make the thing not efficient, so it is expected, it won't be used in deeply dimmed state for too long, what make the deep dimming capability as not so necessary feature.

- And as Ash wrote, unevenly lit tube would be frequently considered as not acceptable.

It's a simple 2 wire out 12v dc ballast but produces what appears to be ac. The variable power supply kit serves as the dimmer module supplying the ballast from 0v to 11.5v dc. There are striations when dimmed below 10% but it's still stable and extinguishes as far as the arc can sustain but can relight when ramped up again from 0% to full tube lighted. The level of dimming is consistent from brand to brand but has some flashing sometimes at lower levels with 1/3 tube lighted and goes away after while. I made the circuit because I needed a dimmable flourescent that could go below 1% but non available. So far tube has been used for months with no sign of wear the other Osram one had some faint banding but that was due to the continuous heating being too hot.

So you have two inverters, one running all the time at full power supplying the filaments (or that could be only a simple transformer frm the mains, it is of no difference) and second, regulated, supplying the arc?
Then this belong into the "too complex" category. Such complexity would bring too high costs, while the demand for such low level dimming is nearly nonexistent.
Normally such systems are done using two light sources, one high power one and second low power arranged so, the beam is identical, where the controller is programmed so, it transition from both to the low power only. So with standard minimum 5% ballasts it is capable to reach system dimming down to 0.25% with the same cost (I have seen a system with a row of LED's behind a fluorescent tube, where below ~5..10% the fluorescent shut down completely and onlt the LED's light, using the fluorescent as their diffuser, so the light still appeared to be emitted from the tube)

Pretty common in large auditoriums and similar halls is to switch OFF lamps from the installation at lower setting, so the system retain it's efficacy. So again no need for such low level capable ballast...

Interesting about the LED lights behind the tube.

This gives me an idea on a dimming circuit which in theory should go down to 0% if the ballast provides sufficient current to the electrodes. This would probably work best with T12s. I'll think about it then post some more details about my idea ..


BG

If you may tolerate some flicker, dimming down to ~0.5% is not as complex. On such levels the lamp does not operate on the desired level, but alter between completely OFF and about 5..10% brightness, the duty ratio then steer the average brightness. This of course cause a flicker. And the problem is, the flicker frequency tend to go down at lower setting (as the minimum "ON" time is rather limited), so may become visible.
Lower currents without flicker you may achieve only by DC supply, but with that you will have troubles at higher power levels.

Given the fact the input power does not decrease below ~5% setting, using a light blocking shield would yield the same efficacy, but way more stable light...

It is better to use LED dimmer bulbs. The Fluorescent lamps flickers when they are dimmed.

@CrestwoodOhio: With LED's, you can not go below the ~10..20% without flicker either. The thing is, you may drive the new LED's even by 0.5% of their rated current and they will work well. But "dimmable" does not mean it can do just the low setting range, but the full power as well. And it has to be able to to dim to the rated minimum even after majority of it's life it was driven at full power. And there is the problem: LED's tend to develop quite high leakage current as one form of aging degradation. This leakage takes some current away without producing any light. And as the degradation is never uniform after all chips in the string, some chips suffer from higher leakage and some from lower or even none at all. So when you drive such chain by a current comparable to that leakage, some LED's will light wnd some not at all, creating quite uneven brightness. Not speaking about the low efficacy: Instead of producting light, the energy just creates heat due to that leakage.
So to overcome that, the current has to be always significantly larger than the leakage. And there we end up at about 10% of full power. So when you want to go lower, you have to use PWM and viola, you have technically a flicker...

And there is another reason a "dimmable" retrofit lamp can not operate below ~10..20%: The problem is not as much in the lamp itself, but with the dimmer. The standard phase-cut triac dimmer can not operate reliably below ~5..10% of the power, at the same time need certain minimum loading of about a watt in the lowest setting to operate with the necessary RF filters.
For the incandescents that was not any problem, as that mean ~1% of light output or even less and even the 40W one is way above that limit.
Fluorescents need about ~5% of the rated power for filament heating. But as the fluorescent ballast need RF filtering on the input, the minimum power for the dimmer operation is higher than with pure resistors. Together with the generally lower fluorescent full power rating it means the fluorescent ballast can not go below ~30% of the rated power input, what mean minimum light output of about 10..20% on a phase cut dimmer.
As the LED's do not have the power need for low level operation and their efficacy and filtering requirement is about the same as with fluorescents, the minimum light level for them is about 20..30% of the full power, unless they incorporate an artificial load. But that load would mean the LED will light at 0.1% of the rating, but still consume about 20% of the rated power.

Incandescent and halogen dimmers, operates by Pulse Width Modulation. Thats why fluorescent and LED lamps, flickering at a lower setting.

Incandescent and halogen dimmers, operates by Pulse Width Modulation. Thats why fluorescent and LED lamps, flickering at a lower setting.

PWM are called rectangular pulses with varying duty, but the mains dimmers just cut out part of the sinewave. This is then called "phase cut" (it could be both leading and trailing edge, depend which part is missing at reduced setting; the leading edge are then the most common)

With CFL that is not the reason. The ballast front stage makes the constant, filtered DC voltage of anything that come to the input, so it can virtually run at any power level of that, assume the front stage can handle it.
The ballast then measure the cut out angle and from that derive the information for the ballast dimming setting. So the ballast then power the lamp at a power level corresponding to the phase angle. With decently made lamps this dependence is artificially shaped so, the relative light output for any setting matches the relative output of an incandescent lamp powered by the same dimmer setting. The reason is, the dimmers have their button angle vs phase angle curve shaped so, it gives smooth and steady dimming perception with the assumed incandescent lamp.
So as the ballast functionality is independent on the dimmer setting, so it is it's eventual flicker: There is either none (higher power levels,...) or some, on varying frequency (when the discharge becomes unstable at low setting). And if that frequency fall below 100Hz, the flicker become visible.

With the LED's it could be the same as with the CFL's (when you really want to match the dimming control curve), but as the LED dimming is usually implemented by a PWM, it will flicker. But most likely at about 150..400Hz, which is the most common frequency range for explicit PWM LED dimming.
The other group means very simple ballasts with no low frequency filtering. That mean the LED light when it get's power and does not light, it doesn't. So then the LED light really follow the dimmer output voltage, so flickers at 100 or 120Hz. But that is not the worst thing. Worse is, the LED power follow the input power, but that mean the dimming is way shallower than with incandescents, so seemingly with small range (the dimmer can not go below ~10..20% of the full power, so that becomes the minimum dimming setting)
But still even the 100/120Hz flicker is normally nor perceived as a real flicker. The problems are elsewhere:

There is yet another problem with the CFL's and LED's: They form a way lower load with way more pronounced filter components, what makes the dimmers itself to operate erratically at lower setting. And it is this erratic behavior, what makes the lamps to visibly flicker, as the frequency components fall even way below 50Hz (it's a form of a subharmonic oscillation). To cure that, usually is sufficient to connect real incandescent lamp of the same rated power as the total power rating of all the CFL/LED's on that dimmer (so one 40W incandescent mixed with 4x 9W LED/CFL's). This usually make the dimmer to operate correctly and the flicker usually disappear.