I have noted that I have the ability to blindly distinguish between an american and an european HPS lamps, simply by looking at their arctubes. It seems to me that the arctubes of american HPS lamps, have different external design than the ones of the european HPS lamps.
For example: In this picture , I could almost automatically recognise the lower lamp as american, without even know that it is a GE lamp, and totally ignoring the longer arctube length compared to the upper lamp.
Is this true?

"Coincidence" due to 2 other factors :

Electrical specs - USA lamps may be different from European lamps of the same wattage, what more, for some lamps more than 2 specs may exist for the same wattage rating. the arc tube will be of different size

Technology - Different manufacturers use different designs and machines, and designs may be significantly different between old lamps (with external amalgam reservoirs) and newer lamps of all manufacturers. It may e so you recognise as "American" some designs which indeed are used by American manufacturers but that's probably cause it "happened to be" this way (possibly with cloning and reverse engineering being part of the process) but nothing beyond

At dor123 the reason the arc tube is differences because of the American voltage on lower wattage HPS. As we use a choke on those lower wattage lamps as those lamps of 35 to 150 watts have about 50 to
60 arc voltage. Then s56 150 watt HPS is the eu spec lamps.  We have to use auto transformers on 250 watts and up

If you try and use a stepdown transformer in reveres to light a 4w T5 meaning eg. 24v ac to 230v ac no ballast you'll know Y US lamps are different. You get power loss and lots of flicker at the electrodes so the arctube has to be different to account for the losses and arc stability.

For lamps of same voltage, there doesn't need to be a difference between USA an EU HPS lamps, but in practice there often is for cost reasons.  This is because many American HPS lamps are used on constant wattage control gear, and with such arrangements the arc tube voltage is less important.  Within a fairly wide range the control gear will compensate for changes in arc voltage and drive the lamp at same power.

In EU this is not possible because we use almost exclusively simple choke ballasts to run HPS lamps.  To ensure that the lamp runs up to the correct wattage, it is essential to make it very precisely with the smallest possible variation in arc voltage from lamp-to-lamp.  The result is that if you take randomly a dozen HPS lamps made for EU market and measure their voltage, they will probably all be within about 5-10V of each other.  If you do the same with lamps made for USA market, the voltages will often vary by as much as +/- 30V either side of the nominal value.  That is why you’ll almost never come across the arc voltage of an HPS lamp in an American catalogue, it’s simply not important in their applications.

Since the EU lamps have had to be optimized very precisely, the manufacturers have learned over time that when targetting one specific arc voltage, there is also an optimal arc tube geometry which will give the highest efficacy and best lifetime.  Over time most of the European manufacturers products have therefore evolved to have broadly the same dimensions as each other.   In America this has not happened to the same extent and you still see quite significant differences from one factory to another.

It’s important to be aware of this difference in voltage control between lamps intended for different markets when you run American lamps on choke ballasts in EU.  Although the nominal values may be the same, because of the great spread from lamp-to-lamp it’s quite likely that the lamp will either be significant over or under-run on EU circuits.  In the other direction it’s no problem to run EU lamps on American circuits.

@James: I guess you were misled by the "constant wattage" term with US ballasts, the ballast behavior are rather opposite:
The European style series choke quite significantly lower the current when the arc voltage increase, so prevent the real power from changing as much.

The CWA have the power nearly proportional with the arc voltage, these ballasts are not constant power, but constant current sources. The term came in the era of mercury lamps, when the arc voltage was pretty independent on operating conditions (unsaturated vapor concept, so always all the fill evaporate), so the only variable was the mains voltage. With that lamp, once the ballast keep the current constant over mainly mains voltage fluctuations (as this was the only operational variable), the power become constant as well.

But with constant current, the real power delivered to the lamp become directly proportional to the arc voltage, so when the arc voltage rise, the power rise too. Now if you take the saturation vapor style lamp, the amount of evaporated fill depend on the arctube temperature (more precisely it's cold spot). But this temperature rise together with the driving power. Now as the gas density dictate the actual arc voltage, we got a kind of feedback in the system: The real power influence the temperature (positive coefficient), what influence the fill density (positive change as well), that influence the arc voltage (again positive change) and this, via the ballast characteristic, influence back the power delivered to the lamp. And with CWA, this is quite strongly positive as well...
Now what we got is a system with a positive feedback, creating a tendency to thermal instability. As no one want an unstable setup, somewhere have to be placed a "brake".
As the CWA for US lamps mean the power rise quite a lot with arc voltage, the "brake" is then within the lamp design: The arc voltage dependency on the real power have to be somehow suppressed. And there are at least two ways I'm aware of:
Either keep the temperature less dependent on the power. As the coldest spot is the only thing what matter, the solution was moving it away from the arc - viola, we have an "amalgam reservoir" style lamps.
Another approach I know is the departure from the saturated vapor concept, so the use of unsaturated vapor. Then the fill pressure, so arc voltage would become rather independent on the temperature. Viola, the feedback is broken. However this mean the sodium loss can not be covered from the reservoir, so this require other measures to keep the lamp lasting long enough.
And I guess there are way more methods in how to make the HPS stable on a constant current CWA ballast. What is clear, each method will require different arctube shape. And as there are variety of such methods in use, the arc tubes differ.
But the constant current nature of the ballast allow the lamp design to alter the real transferred power to match the light output, so to compensate for different efficacy - allow to hide a less efficient lamp design by using higher power. So another possible cause for different arctube proportions (mainly with cheepeese lamps)

The EU style series choke lower the current when the voltage increase, keeping the real power pretty constant.
That mean the lamp stay stable, even when the arc voltage is strongly dependent on the temperature, so there is no need to solve the thermal stability within the lamp. That mean the advantage of the saturated vapor concept could be utilized to full extend, more-less without any restriction (the fill material losses are covered by the rather large reservoir).

But the series choke mean at least two drawbacks:
There is no trick to conceal the worse efficacy by increasing the real power, so worse efficacy design would mean only less light.
And second drawback is in the mains voltage being the only thing available to restrike the arc after current zero cross. And that dictate the arc voltage to be less than half of the mains (the series capacitor on the CWA mean the restrike voltage is nearly twice the ballast OCV, due to resonance effect). So the lamp can not operate, when it's arc voltage rise above ~120V.

Now as the tube blacken over the lifetime, it absorb more light, so the temperature rise, what mean the arc voltage rise as well. So the highest arc voltage is just before the EOL (in fact the life is given by the arc voltage becoming too high for the mains voltage).
So when the lamp shall last long enough, it should have sufficient room for the arc voltage to rise, so the voltage on a new lamp shall be low enough. But as too low voltage mean low efficacy, there become an optimum trade-off between lifetime and efficacy. And this settled on about 70V on a new lamp.

While the EU design does not need any extra special features, the design remain the simplest possible. As these fundamentals are common for all makers, the resulting design (tube shape and dimensions,...) become virtually the same for all of them as well.

Then why in the USA a CWA ballast is used for HPS ? Wont an autotransformer that imitates "plain step up voltage transformer + plain choke" be better ?

Ash: American HPS lamps, aren't used on CWA ballasts, but on HX ballasts. These are Eltam that calls their autoregulator ballasts for american HPS lamps "CWA", and I don't know about their operation reliability.

US penning start and internal starter HPS retrofits for MV lamps, usually retrofit CWA mercury ballast (Since most MV lamps operates on CWA and not HX ballasts). To allow a reliable operation on a CWA ballast, these lamps have to contains an unsaturated sodium vapour, meaning that their arc voltage don't rise during life, which may allow CWA ballasts to drive the lamps at dangerously high power loadings. This adds their benefit that they aren't cycling at EOL.

@dor:
  HPS are used on both CWA, as well as HX ballasts. The CWA is a more expensive option, but allow "dirtier" supply.

Then why in the USA a CWA ballast is used for HPS ? Wont an autotransformer that imitates "plain step up voltage transformer + plain choke" be better ?

The CWA still does it's original function: Compensate out the fluctuations of the input voltage.
In US many installations the input voltage tend to vary quite a lot, the HX autotransformer would yield too much power variations there.
The EU style ballasts are designed for +/-5% mains tolerance, what already correspond to about +/-10% lamp power variation, what is quite on the edge.

There are ballasts able to compensate out the mains fluctuations and at the same time allow to keep the real lamp power about constant over varying arc voltage - three coil "mag-reg" ballasts, but their losses and weight are too high to be practical for most installations.

Moreover the CWA's series capacitor block any DC current, so prevent the lamp from flashing widely even when it starts to rectify (during warmup, response to some mains disturbance,...). Similar events on the European series choke frequently lead to lamp arc extinction, so the CWA allow way more severe mains disturbances to occur, without causing the lamps to extinguish.

In the Europe, the lower losses of the simple series choke were so strong argument, it mainly economically "allowed" things like separate, thick (for the power transferred) wiring only for streetlights, with no sources of disturbance as a measure to go around their sensitivity, the savings on the electricity consumption were simply high enough to cover for these extra installation costs. Moreover the distances in the Europe are way shorter than use to be in the US installations, what keep the extra costs from being too high.

Thanks for the detailed explanation Medved.
Just one point then that I don't fully understand, is why the American lamps are not manufactured with the same tight control on arc tube voltage as the EU versions?  From your explanation about the different circuits, I would have thought it even more critical to keep the arc voltage within tight limits on CWA gear.

I just wanna say CWA ballasts are acshely constant current devices not constant wattich if that was the case wen u get a 400W SON that the arc V is 75V then the ballast if it was CWA wood try and output more amps (to keep the W the same) and wich wood overload the electrodes if they were the same but on a Constent current ballast the ballast wood keep the A the same wich is watt a US CWA ballast dose (the same lamp on a EU choke wood mean more V drop across the choke and means more current flowing thro the lamp and the reverse if a lamp has a hier arc v wich meens less v acrose the choke and less current thats how i can run a F50T8 wich runs a 165V .38A on a 35/55W SOX choke wich is normaly ment for a 35\55W SOX wich is 70\100V .6A just fine). from my own tests with a CWA ballast i can conferm that a CWA ballast is realy a constent current ballast so CCA lol (even wen i short the ballast out the A stays fairly constent no more then 100Ma more)  olso i have a US 400W GE SON for the EU market from 1981 and apart from the cap looks no difrent to one for the US market

hope this all makes sence and helps out hear 

Thanks for the detailed explanation Medved.
Just one point then that I don't fully understand, is why the American lamps are not manufactured with the same tight control on arc tube voltage as the EU versions?  From your explanation about the different circuits, I would have thought it even more critical to keep the arc voltage within tight limits on CWA gear.

There is a difference between arc voltage tolerance and it's dependence on the input power variation.
The previous described, why the US lamps have to maintain their arc voltage ideally independent on the actual power to prevent thermal instabilities, but it does not mean the arc voltages could not vary due to other reasons.

To maintain the arc in the AC circuit, you need two conditions to be met:
At first the OCV measured as first harmonic have to be higher than the arc voltage first harmonic, so the current would flow and deliver power into the arc.
Second because the current is AC, it have to change polarity at some time. That mean the current first decay to zero and then it could be reestablished in the other direction. But with mains frequency there is a problem: The arc essentially disappear when the current drop to zero, so after the polarity changes, you have to essentially reignite the arc again, so the current could flow again.
For the reignition you need hot electrodes (quite easy, they are already warmed up and they have the inertia) and some extra voltage to "repopulate" the arc with ions.
Now this voltage have to be above the arc voltage. The arc voltage is in fact an equilibrium point, when the new ions are generated at the same rate as they decay. So having the voltage equal to that mean the ions won't populate. So you need there something extra. And that mean higher the arc voltage, you need higher voltage for the reignition as well, the "extra" margin for the ion breeding should be always there. So higher the arc voltage, higher voltage should be available in the ballast for the reignition.

Now with the series choke, the only voltage you have is the mains voltage. Given the phase angle conditions, this lead to the generic limitation of the maximum usable arc voltage to be half of the mains rms (higher power arcs are more stable, so the arc voltage could be a bit higher, but it is still not that far). For European mains it mean ~110V arc voltage at all times over the lamp life. As the HPS voltage rise over it's life, the initial arc voltage should be lower. Higher initial voltage would mean the lamp reach sooner the 110V EOL limit, so it's life would be shorter.
Now creating a lamp with too low arc voltage (to boost it's life) is not a winner too: It mean ,lower efficacy, higher current, so higher ballast load, so when the voltage would be below what the ballasts are designed for, the ballast would overheat. That mean lower ballast efficiency as well. At the same time lower arc voltage mean (practically on all wattages, maybe except the lowest ones) lower efficacy, so the system efficacy would be severely compromised. So the standard set a minimum value (as a trade off balance), what all ballast have to be designed for and what no lamp should cross. Now as makers want maximum life from their products, they are pusheing the voltage tolerance range as close as possible towards the minimum limit.


The CWA can tolerate way higher arc voltages without the danger of extinguishing it, because the series resonance effect (it is in fact a series LC operated below the resonance) keep the capacitor charged about at the peak voltage during the current zero cross. So for the lamp reignition, there is not only the (transformed) mains voltage, but it's sum with the capacitor voltage, so aboput double voltage compare to the series choke, what is available to the lamp to reignite the arc.
Practically this mean the arc reignition is not anymore the limit for the arc voltage. So the arc voltage could really approach the ballast OCV, so it extinguish only when the ballast is not able to feed any power.
The other limit is not as hard as well: As the ballast keep the current constant over wide range of conditions, the lower arc voltage does not present such risk for ballast overheating, so lamp designs may go safely as low as their designers wish (there is the efficacy and ballast efficiency penalty with doing that; this is, how the "energy saver" MH's reach the lower power input).
As the limits are by far not that strict, the makers are not bound to so tight voltage tolerance as they are for the European market.

Of course, to keep some margin for the ballast to compensate the main variation, the target arc voltage is about the half of the OCV, but it is by far not that critical as with the series choke only.

So if I could sum up:
As you can not get all sweets, with lamp design you have to make compromises. And one of the copmpromise is the achievable lamp-to-lamp tolerance range, vs thermal stability.
For EU market you need to tighten the lamp-to-lamp tolerances, but you don't care as much for the thermal stability, you may use the saturated vapor concept to it's largest extend (pressure is a function of temperature and nothing else, so the effect of dimension and dose variations are suppressed), the series choke keep it stable anyway.
For US market you need lamps helping to thermally stabilize the stuff, but you don't care as much for the lamp-to-lamp tolerance, so e.g. an unsaturated vapor concept works well, even when it is very sensitive to dosing and aging, the CWA cover that spread.

Thanks, superb explanation :-)  It is a pity that the USV concept is so expensive and technically very difficult to produce in reality, so that it was never really widely adopted in either the EU or US markets.

I guess the reason, why the unsaturated vapor is so expensive and so was not widely adopted are mainly the manufacturing difficulties, the sodium have simply too small and reactive atoms, so making anything sodium-gas tight is nearly not possible...

I saw few examples, but only for the US market.
I guess, the tight tolerances required for the EU market are even not achievable with the unsaturated concept, so I even doubt such lamp was ever marketed.

USV HPS lamps are only marketed for EU in the form of mercury retrofit lamps, such as Sylvania SPX EcoArc and Iwasaki Sunlux Ultra Ace.  To make the PCA sodium tight is quite feasible, but the problem lies in chemical reactions with the other arc tube materials.  For a start since sodium reacts with ordinary emitter coatings, it is necessary to use quite exotic materials for the electrodes.  Iwasaki uses sintered cermet electrodes and Sylvania lamps have an yttrium based emitter on a tungsten coil.  Then there are reactions with the frit seal which consume sodium over time, and need further optimising.  Also the dosing and composition of amalgam has to be rather more critical than in standard HPS.  All these challenges mean that even though every major manufacturer has made USV lamps at some point, only two really mastered the techniques and could control the processes well enough to be able to market a product - but even then, only for the niche sector of retrofits for mercury lamps that have higher performance than normal HPS retrofits.