Maybe somebody can shed some light on this (no pun intended), Ive seen the 360w metal halide lamps in the past that supposedly replace the standard 400w lamps in base-up fixtures such as high bays and offer a savings of 40-50w per fixture. Well today I stumbled upon some new 320w lamps at the commercial lighting store that are also made for 400w fixtures and allow even more savings than its 360w counterpart. How is this possible??? I dont even understand how the 360w lamp works and the new 320w one is a good 80w less than what the ballast was designed for. Both these "energy saving" lamps have the same ANSI ballast designation as a 400w lamp too. Doesnt a Constant wattage autotransformer deliver the same wattage regardless of which lamp is used??? If I run a 400w lamp on a 400w ballast and the kill-a-watt reads 450w, then I screw in the 320w lamp will the kill-a-watt read 370w?? doesnt make any sense. Also these 360w and 320w lamps look identical to a 400w lamp, arc tube and all. The only difference is that they're only rated for base-up installations. Still makes no sense. I ran a 250w Halide lamp on a 400w ballast and the kill-a-watt read 450w. This must work somehow or nobody would be buying them year after year. Anybody have any idea?

The "constant wattage" in the CWA name is a bit misleading name, in reality it is the lamp current, what these ballasts keep constant over both mains and lamp variations.
The name was coined, when all the MV's had exactly the same arc voltage (so there were no differences), but the mains fluctuation were a problem. The lamps were expected to be already of constant arc voltage: On MV's the arc voltage depend only on the arctube geometry and it's fill doses, so does not change with the actual current, so power delivered to the lamp.
So when the ballast was able to keep the current constant over the wide mains range (what was the main problem of the long road stretches - as the voltage dropped over the long wiring), it kept the lamp input power constant over the mains variation as well, so the marketing gave it the name "Constant wattage", even when in reality "constant arc current" would be technically more accurate.
 In another words the technically correct name for the "400W CWA" would be something like "3.5A CCA". But for ordinary lineman and/or lighting designer that won't be practical, as they usually don't care about what exactly happen between the lamp and ballast, but they need top match correctly the ballast with the lamp and know, how much load the thing would mean for the mains and how much light the thing would give, so the rated wattage become more used (well, except for the North America). In the US it is the code, what describes the ballast characteristic (and so the match with lamps), but it does not tell, what would be the real transferred power. The rated wattage is then only to give an estimate about the power category in numbers, but it does not mean at all the real delivered power.

But as the ballasts keep the current constant, lowering the arc voltage mean you lower the power delivered to the lamp. And this is exactly, how these "energy savers" do reduce the power input.

And as the ballast designation specify the ballast characteristics, even the 320W lamp should bear the "400W" ballast code, when it is designed to run on "400W" ballast, regardless of what real power would be during operation.

If the point of CWA is fo load stretches with vdop, why they did not use simpler HX with few taps on the input instead ? (at asy 5v distance frm each other) - the farther from the supply point, you use lower voltage tap

I would put the question the other way around: Why not to use the CWA, when with CWA you don't have to care about the exact voltage at given point, the size and cost is about the same, the losses are about the same as with the HX (because you need the transformer functionality anyway, because of the rather low mains),...

The point of the CWA is it's ability to keep the lamp current, so power relatively constant over wider mains variation.
The variations could be of whatever origin, including the varying load on long cables and/or unknown exact voltage at the given point.

The multitap HX would work, but only when the correct tap would be selected.
At first this would be VERY prone to wiring errors (each lamp wired in different way, together with the mess of 20 voltage taps would be a nightmare)
Second any load change (e.g. adding new or removing a part of the illuminated stretch) would require to rewire the existing fixtures - a lot of work and again "warm welcome" to many errors.

I still dont understand how this will work. For example a 250w lamp has a lower arc current than a "320w or 360w" rated lamp but a 400w ballast still drives the lamp at 400+ watts and the lamp is clearly overdriven, when a wattage reading is taken it still reads 450-460w. I cant imagine the contents of the "energy saving" lamp arc tubes are any different than the typical 400w lamp (sodium scandium etc..). I do know this though, I have a CWA ballast for a 175w lamp and I purchased a 150w "energy saving" lamp, there is a slight difference of 3w less with the 150w lamp (which is probably just due to differences in chemistry as both are made by different manufacturers). How would this be any different with 400w systems? The only way ive even known to cut energy costs with metal halide is to replace BOTH lamp and ballast with one of a lower wattage.

For example a 250w lamp has a lower arc current than a "320w or 360w" rated lamp

I think I see the basic of the notunderstanding:
You are used to regular mains circuit, where the voltage is given by the energy source (so the mains, battery,...) and current by the load (so e.g. an incandescent lamp,...). So if the energy source is given (e.g. 120V socket), the voltage is there pretty independent on what the load does. So if you connect there 120V/60W lamp, it would draw 0.5A and the transfered power would be 60W. If you put there 240V/60W lamp, the current would be something above 0.125A and real power something above 15W (even when the lamp is still rated 60W), as the lamp is not matched to the energy source.
Well, with the discharge lamp vs their ballast interface it is basically "the upside down": The lamp, so the load is, what dictate the voltage and n ot the ballast. The ballast is then, what dictate the current. 
So you can not say "The 250V lamp has 2.15A arc current and is rated for 125V arc voltage", but you have to say "The 250W MV is rated for 2.15A and has 125V arc voltage"
The current is, what the ballast feed it, so if you connect the 250W lamp to the 400W ballast feed 3.5A, the arc current would be 3.5A (dictated by the 400W ballast) and the lamp would have still 125V arc voltage (as MV's keep the arc voltage over wide range of current), so it would yield the same real power (~400W) as with the 400W lamp (because all the MV's and probe MH's happen to have very similar arc voltage of about 100W is designed to have 110V, 1kV to have about 130V across the arc). The delivered power by given ballast would be always the same with lamps having the same arc voltage, regardless what is their rating. So if the 3.5A is above the lamp's rated current, the lamp is overdriven (e.g. the 250W 2.25A lamp on a 3.5A "400W" ballast), if lower, the lamp is underdriven (1kW lamp rated for ~8A run on 3.5A "400W" ballast)

In other words the 3.5A "400W" ballast feed the 400W only into lamp having 125V arc voltage, but into whatever lamp having about the same 125V arc voltage.

When you use a lamp, which have lower arc voltage (let say 100V) and run it on the same 3.5A ballast, the current stay the same (as it is dictated by the ballast, not the lamp), the delivered power would be V*A*PF=100*3.5*0.9=320W (the lamp power factor is below 1, as the voltage and current have each different shape - the voltage is about rectangular, the current is more like sinewave, so PF=0.9 or from the CWA a triangle, so PF=0.87=~0.9 as well)

If I say it in different words: As on 120V mains would work any incandescent rated for 120V regardless of the rated power, on a 3.5A ballast would run any discharge lamp rated for 3.5A, regardless of the rated power. Don't look as much on the ballast "power", as that figure only mean for what lamp the ballast was originally designed for, but does not say anything about the real power delivered to the lamp.

With CWA and MV/probeMH the thing is simple, as the lamps have constant voltage (based on how they are designed) and the CWA feed constant current (based on their design).
With HPS the thing is way more complex, as the HPS arc voltage is somehow dependent on the arc current the lamp get from the ballast and in order to make the system thermally stable, the ballasts are designed to have their output current somehow dependent on the voltage the lamp have.

So in order to get the real lamp volts and ballast currents, you have to draw both dependencies into one VA graph and the resulting lamp voltage and ballast current (so the power) would be given by the point, where the two lines cross.

That's a very informative post, thanks. I knew something about how discharge lamps work but this makes good sense to me

I wonder how these work on simple choke ballasts (220-240V)? I would think there'd be more ballast losses and also higher arc current, leading to the lamp being overdriven and ballast running hotter?


BG

There are two aspects:
1. Phase shifts and summing the voltages: The series choke is an inductor, so the current lag behind the voltage across it. The lamp have the voltage in phase with the current. As both are in series connected, the current is the same in both the amplitude, as well as phase in the whole loop. And as the second Kirchhoff law dictate, the sum of both voltages should give the mains voltage. But it is not the straight sum of amplitudes. We are working with AC, where the voltage changes all the time. The second Kirchoff law dictate the relationships between instant voltages in any moment, but as there is a phase shift between the lamp and coil voltages, once the lamp voltage is around the center, the voltage across the coil become zero and vice versa. So when evaluating what the rms values do, after some painful math and approximating the lamp voltage by a sinusoidal shape, you end up with rather simple Phytagoras equation: Vmains^2=Vlamp^2+Vballast^2.
Now as the ballast is in fact an inductor, so it's current is proportional to the voltage across it (neglect saturation effects for this moment), the ratio being the ballast impedance. And that we assume constant for now.
So when we use it for the 230V mains and our lamps, we get:
Vmains^2=Vlamp^2+Vballast^2
As we want to know, what the ballast current would respond to the different lamp, we could rearrange it:
Vballast = (sqrt(Vmains^2-Vlamp^2)
So assume the 130V full power lamp give the nominal arc current, we get the ballast voltage corresponding to the "100% rating":
Vballastfull=sqrt(230^2-130^2)=190V
And this yield the rated 3.25A
And now we replace the lamp by 100V energy saver:
Vballastsaver=sqrt(230^2-100^2)=207V
So the ballast voltage increased from 190 to 207V, the same ratio would apply for the circuit current, what lead to increase in the lamp current by 9%. But as the lamp voltage is lower, the final power would be (ratio counting) Psaver=400W*Vsaver*Vballastsaver/(Vfull*Vballastfull)=335W
So the overdrive would be only a little...

Other problem with US MH's is the 230V is too low as OCV for their startup, so they need an aid in the form of an LV ignitor. But if the full wattage work, the energy saver would work even better, as it have shorter discharge path, so generally should suffice with lower voltages.