Hi, recenterly i recovered an 125w GE Kolorlux 125w from scrapeyard, but it was too late for its ballast, so i took 2xT8 ballast in parallel for drive this lamp.

Actually, current is 1.2A, but voltage is too hight: 130v (i have 230v main).
Because i don't want to overdrive this lamp, can i use an 36w ballast?

You can order a ballast for 125W mercury lamps, from your local wholesale or online. The mercury ballasts are still available.

Actually, current is 1.2A, but voltage is too hight: 130v (i have 230v main).
Because i don't want to overdrive this lamp, can i use an 36w ballast?

Don't worry about the voltage, it is dictated by the lamp and yours is apparently asking for a bit more. This voltage would not change on any other ballast...
The problem may be only, when the voltage become too low (below 70V at steady state, e.g. arctube leak,...), as such lamp may fry the ballast.
For the lamp is important the current you feed there, while the 1.2A is completely OK for the 125W MV.

So according your description, the setup behave absolutely correctly...

thanks medved, but this system drav 160W!
I know about ballast losses, but i'm bit scared about tube loading.

thanks medved, but this system drav 160W!
I know about ballast losses, but i'm bit scared about tube loading.

A questions:
- How did you calculate the power?
- What instrument did you used to measure the arc voltage? Was it the true-rms meter?

Meter problem:
Low cost meters measure the "rectified average" and then multiply the result by a correction factor so, the reading of the sine wave voltage match the rms value:
Reading = average(abd(Vin))*1.11
But with non-sinewave shape the reading of this meter is incorrect, with the rectangular shape of the discharge arc voltage it mean the "1.11" factor, so the real arc voltage would be 130/1.11=117V.

Power calculation:
When calculating power, the only correct way is multiply the current and voltage waveforms (multiply each time point separately, so you get the waveform of the power transfer) and average the result.
This method is rather impractical for everyday use with resistive circuits with only in phase sinewaves of the base frequency, so the voltage and current single number values were defined (= the rms values) to allow simple calculation of "P=Vrms*Arms".
But as this simplification is designed only for sinewave in phase, it give the correct result only with sinewaves in phase (well, the result remain correct with any shape, when the voltage and current waveform shape exactly match and are exactly in phase), so only for pure linear, noninercial resistive loads.
But the arc is not a linear load: Current fed by the ballast is nearly sinewave, but the voltage is nearly rectangular. So when you use the "P=Vrms*Irms" equation, the result would be higher by the factor of 1.11 (you may do the integration math; for exact sine and an exact rectangle it would be 2*sqrt(2)/Pi).
This shape mismatch correction factor (better say it's inverted value) is then called "power factor" and it is unity only when both shape and phase match (not the case with nonlinear component like discharge, not the case for inertial components like L or C).

So with the discharge fed by sinewave current the real power would be (I use your case):
Plamp = Irms * Vrms * 0.9 = 1.25 * 117 * 0.9 = 131W

And that is not as much off the rated power...

i used my multimeter in V and A settings.
After i used an VA meter.

But important is, what measurement method these instruments use.
Reular digital multimeter use a two way rectifier (an active one, so behaving like it would have ideal diodes), low pass filter and either modifier ADC reference or an extra 1.11 gain stage. So when used for the arc voltage measurement, it read incorrectly (by the factor 1.11, as with rectangular shape with 1:1 duty ratio the "rectified average" is exactly the same as the rms, while the meter multiply the value by 1.11)

Cheap home energy/power meters frequently use the "rectified average" (above) synchronous rectification for the voltage channel to measure the voltage and synchronous rectifier for the current channel (so they eliminate only the phase shift).
The current channel would measure correctly, as the current is sinewave. But the voltage reading would be wrong.

Better ones do expect non-sinusoidal waveform (the ones used for billing), but only in the current channel. Some use a low pass filter to keep only the 1'st harmonics of the current path signal (as with sinewave voltage only the 1'st harmonic of the current carry the real power) and then the synchronous rectification system (as above). So with the voltage being not sinusoidal they suffer from the same error as simple multimeters (maybe slightly reduced by the filter in the voltage path, but the voltage filter is not designed to reject higher harmonics as the signal path for the current).
 
Only some sample both voltage and current signals (and convert to digital data), on each time instant multiply the "volt" and "amps" and then average the result. And only these are capable to measure the power correctly.

If you have a two channel digital scope and an insulation transformer (Beware, scope is grounded, so the measured circuit should be isolated, otherwise you may damage the scope or probes by an excessive currents to the probe gwound connections) with current probe and with some math capability, you may program it there:
- Use one channel for voltage, second for current
- Set the timebase so, there are either only complete mains periods on the screen, or many of them (~20..30)
- Multiply both channels (by the "Math" function; it would be the shape of the instant power)
- Let it "Measure" the "average", you get the wanted real power.

thank for the answer! 

another question: does anyone know some safe frequency for hf ballast?
for an project, i'm thinkink to build an 400hz sine/square generator
with an proper inductor.

What do you mean by the "safe frequency"?

Aircraft "magnetic" ballasts are designed for 120V/400Hz and as any other precision magnetic, generally it need sinewave shape and steady frequency within 1..2% range.

HF ballasts usually use on board electronic inverter (hence the term "electronic" ballasts) to generate the required 10's kHz AC. The operation frequency in these ballasts is not fixed, but it adopt to the actual operating mode need and it should correspond to the frequencies the ballasting LC network is designed for, so there is nothing like one, common "safe" frequency.
This inverter is then supplied by a DC voltage (about 300..400V for higher power lamps) coming from the rectifier stage (either simple bridge for 230V, or doubler for 120V, or an active PFC stage for wide range and high PF ballasts). Here it depend on the rectifier structure:
- Bridge rectifier and doubler you may feed by the designed AC voltage with frequency up to ~500Hz.
- Active PFC you may feed by AC of maximum 120Hz
- Doubler could not be fed by the DC at all
Bridge could be fed by the DC equivalent to the rated mains peak voltage (so 320V for 230V ballasts). But you need separate DC rated fuse to the supply line, the internal one is not rated for DC circuits (so it may not disconnect should a ballast fault happen)
- Active PFC ones could be fed by DC equivalent to the rated AC rms voltage. Unless the ballast is explicitly rated for the DC, you need the separate DC rated fuse as well.

What do you mean by the "safe frequency"?

Aircraft "magnetic" ballasts are designed for 120V/400Hz and as any other precision magnetic, generally it need sinewave shape and steady frequency within 1..2% range.

sorry, safe frequency for hid lamp (so resonance don't occour).
i know that it depends to a lot of factor, but i read that under 500hz there aren't problems.

400Hz is too low to be practical for a HF style (resonant) electronic ballast - it would be complex, at the same time bulky and lossy.
HID ballasts are usually done as a constant power (so with hyperbolic VA curve; for fast changes they behave as a constant current) DC source using buck DCDC stage for mains  and flyback type DCDC when supplied from the car battery, followed by a comutator bridge, what swap the polarity to the lamp each few ms (so with 100..500Hz). As the current to the lamp is nearly rectangular, it have two onsequences: The arc have constant intensity over the whole periode, so no accoustic waves are exciteed at all. And the second, as the rectangular current have no spikes, (it have quite low crest factor), it does not stress the cathodes by the peaks, so they sputter more slowly during noemal opereation, so the slower arctube blackening.

Other option is to use high frequency (20..50kHz; so it is easy to implement with high efficiency and reasonable size), but with the frequency modulated to spread the spectrum, so avoid resonance problems in that way. But as the freequency modulation influence the actual arc power, the amplitude of the HF AC inverter output have to be modulated together with the frequency, so thee arc current does not change (otherwise the lamp qould flicker). And this is the trickiest part of the HID resonance free HF ballast design.