Home Photovoltaics Xantrex Inverters SMA Inverters After 7 Weeks After 18 Months

Xantrex Inverters



This page is obsolete!

My disappointment with the four Xantrex inverters led to replacing them with three SMA inverters. You'll find the SMA  Inverters page much more interesting.

It would be unfair to not mention that Xantrex subsequently redesigned the STXR2500 inverters which resulted in a much improved product. The new Xantrex models perform about as well as the Sunny Boys and offer some advantages with their support of low voltage PV strings. And Xantrex did right by their customers when they replaced all the old models in service with new ones.

May 17, 2002
I just received the first STXR2500 replacement and used it to replace enemic inverter #4. We'll soon see if performance improves on that PV cluster.
Outside the new XR looks identical to the older model. Inside it has a modular jack to which a remote display can be connected. It is supposed to have a redesigned combiner with input terminals split between right and left sides (instead of all left side). The new combiner supposedly provides direct connection terminals to the combined node to receive wiring from external combiners. But the unit I received still has the old style combiner.
The displayed data has been seriously reduced. The XR shows only the cumulative watt hours for the day. No more instantaneous power output or input/output voltages. This will make my performance evaluation more difficult.
Perhaps the presence of the power detail created too many complaints from disappointed users.

Efficiency Test

Test setup

Analog DC voltmeter is connected to a pair of the inverter's unused combiner input terminals

Digital DC clamp meter is around the loop of wire between the DC GFCI breaker and the DC main breaker

Output power is read from the inverter display.

The best way I found to capture and coordinate varying readings was to shoot a lot of digital photos of the setup. Looking at the four cropped close-up examples to the right, one might suspect loss of synchronization among these very dynamic readings due to differing convergence times among the instruments. Therefore dependable readings can only be obtained when the inverter has been at a stable point for a few seconds. Shot 1357 is the only sample that could be considered dependable by this criteria.

The four shots shown here were captured within a 1.5 minute period at about 2:45 PM on a cool May 16, 2002, and under a perfectly clear sky.  During that time the inverter made two significant power reductions and began to ramp up again. I was trying to catch the peaks and valleys as I watched the analog voltmeter, but missed them badly. Someday I'll try a video camera.

Inverter #2 seems to be my best. Most of the time it is well-behaved and holds a stable power point longer than the others. The best efficiency I've measured at stable points of about 1,600 watts is 90.0%.

Inverters 1, 3, and 4 are more erratic. They operate normally for varying periods of time and suddenly drop the load back, sometimes way back, and slowly ramp up to rediscover the maximum power point. Often they overshoot and drop way back again instead of finessing to the final point. I was occasionally able to derive efficiencies approaching 90% for inverters 1 and 3, but most often they were 87% to 89%. Inverter 4 showed lower efficiencies ranging from 76% at mid morning to 86% at solar noon.



Shot 1347

Shot 1347 close-up
88.4% efficiency
Shot 1351 close-up
98.7% efficiency ??
Shot 1352 close-up
87.5% efficiency
Shot 1357 close-up
89.9% efficiency
Analog meters are helpful because they allow a human observer to follow the constant fluctuations that occur as the inverter probes along the VI curve for the maximum power point.
If this topic becomes more important, I might install a 50 mV shunt and combine the voltage and current readings into a four-quadrant analog multiplier. The multiplier output would indicate power directly, and if plotted would reveal the proximity to the maximum power point.
Or, if the topic becomes really exciting, I might search for a ready-made instrument to do the job.
One opportunity for measurement error might be a small AC current that was detectable in the DC circuit. I've almost convinced myself that the clamp meter integrated this into the DC reading. If this is not the case the measured efficiencies would be slightly higher.
How does MPPT really work?
According to Xantrex literature, the inverter makes an "efficient sweep" every minute to find the MPP, after which it rides with that load until the next minute.
My observations are that it constantly probes along the VI curve, gradually increasing and decreasing load in small increments with a dwell time of about 1/2 second. But often it seems to increase too far and then makes a sudden curtailment, sometimes nearly back to zero.
I'm puzzled by this behavior. Xantrex tech support explains that if the MPP is exceeded on the heavy side, a rapid withdrawal must be made to avoid "collapsing the array".

Output Power Test

With the power of only five PV strings to draw from, the maximum output power is limited to about 1,850 watts. I have not yet seen any of the inverters reach this level. To push this limit I externally combined three strings, borrowed from inverter #4, and added them to inverter #3. Eight strings at clear mid day should have sourced enough power to easily reach the 2,500 watt maximum, but the inverter never reached 2,200 watts. In fact it moved about 7 volts higher on the VI curve, away from the maximum power point, as though trying to comply with a programmed power limit.


Let's look at a VI curve. Here's one:

Click to enlarge and look at the green curve. Given that MPP is in the vicinity of the knee corner, it's easy to see why loading beyond that becomes fruitless. Much beyond the curve goes constant-current and small current increases cause big voltage decreases. Could this be "collapsing the array"?
If so, couldn't the MPPT algorithm simply retract the step that got it into trouble?  Why must it drop back so drastically?
If the VI curve had a negative resistance characteristic, like the red curve in the picture, one small step over the line  would drop voltage to zero. Now that's collapsing, and the only recovery would be to release the load and start over.

Daily Energy Delivered

One gratifying observation was that the delivered watt-hour readings shown on the inverter displays, when totaled for a day, track closely with that registered by the private kWh meter. The inverter sums are always above the private meter reading within 0.5% to 1.5%. This implies that the individual readings are quite accurate and constitute a sensitive performance indicator.

From my observations, inverter #2 is the best performer. It starts early after sunrise and quits late. Inverter #1 follows closely, #3 is somewhat disappointing and #4 is a real letdown.


Inverter #4 was replaced on May 17. The new model STXR2500 appears to reduce some of the malperformance issues mentioned above.

A question comes to mind after seeing the chart below. Click to enlarge and look at inverter 2 on days 12 and 18.

For every day of the period except 12 and 18, inverter 2 leads the pack in production. Why would it fall below its peers on two days?
Similarly, why would inverter 1 sag so obviously on days 13 and 14?  Or poor old lagging inverter 4 on day 8?  I would expect a lot more consistency within the group.
Note:  I experimented for a short time on day 27 with the entire array feeding inverter #4, thus creating the abnormal excursion seen in the chart.


Its pretty clear to me that the most serious deficiency with the Xantrex inverters is their abysmal MPPT algorithm.


At least one of the ST2500 inverters that I own cannot meet the manufacturer's specifications for maximum output power, and it would appear that none are likely to reach the specified 94% conversion efficiency. Fortunately my system design doesn't expect them to operate above 1900 watts output.

Final conclusions are reserved until all of the original ST2500 inverters are replaced with the new model STXR2500.