Output at Irradiance is often the most confusing step of the loss table. Here we will explain what drives that calculation, and how to interpret the results.
What does Output at Irradiance even mean? This step of the loss table accounts for the low-light performance of the module. Specifically, a module's performance doesn't change linearly with sunlight; the production generally drops at lower sunlight levels. For example, say you have a module with an STC rating of 200 watts (i.e. the power at 1,000 W/m^2 and 25°C is 200 watts). When that module is under exactly 50% of sunlight (i.e. 500 W/m^2), the power will not be 50% of the rated power - instead of 100W (50% of 200W), it might produce 98W. This is because one of the basic properties of semiconductors is that their low-light performance is not as good as their full-sun performance.
It helps to look at a simple example with a 305-watt crystalline module:
Note that while this module is technically a 305-watt module, the actual performance at STC is 306.2 watts. And furthermore, at lower sunlight levels, the actual production is lower than the pure linear prediction:
For example, at 40% sunlight, the module actually produces just 116W, versus the 'predicted' power of 122W (305 x 40% = 122W). And since the modules often spend much of their time at lower irradiance levels, these overall differences can easily add up to annual 'losses' of between 1-5%.
In fact, if we take a random sampling of 20 modules, we can see that many of them have losses in the 3-5% range, others have losses in the 0-2% range, and there is one outlier with irradiance 'gains' of over 1%:
How can this be zero (or even positive)? This becomes more complicated when you add in one more potential wrinkle: PAN file creators (i.e. module manufacturers) sometimes push the numbers up.
Consider a simple example: say you had a modules that are fundamentally a 300-watt module, with 3% irradiance loss in low-light conditions. But then on the datasheet, you decide to drop the rating of the module by 5%, so you sell it as a 285-watt module. The underlying physical properties haven't changed - so at STC it still behaves like a 300-watt module, and the Output at Irradiance will actually be +2%! With the new "ratings" change of the module, from 300 to 285, the total energy production is still the same. The way this will show up on the loss table is that the "Nameplate" step will be 5% lower, while the "Output at Irradiance" will be 5% higher (from -3% before to +2% now).
In practice, manufacturers might do this because they want to get credit for the positive-tolerance binning of their modules - in other words, if you purchase a 310-watt module with +0 / +5W binning, then the actual performance of that shipment is arguably 312.5W. If the PAN file physics is adjusted to account for that, then the "underlying" properties of the module might be 1-2W greater than nameplate - and so the Output at Irradiance will improve.
So technically, it isn't really a "loss", right? That is correct - its arguable that we don't need to show both the "Nameplate" step of the calculation (where we multiply the sunlight x module namplate power) and then the separate "Irradiance" adjustment. In practice, this is done to give some transparency around the PAN file behavior - and also, this is what PVsyst does, so this is what financiers have become baselined to expect to see.
So how should I interpret this number? Ultimately, there isn't a clear definition of what is correct or not. Fundamentally, this step combines two factors: the low-light performance of the module (a physics phenomenon) and the way the PAN file was developed (a business decision). So if the Output at Irradiance losses seem too good to be true, the best course of action is to check with the manufacturer.