How derating, voltage drop and the inverter's MPPT window actually decide a cable schedule, why a string length that works in Chattogram can be marginal in Panchagarh, and what a combiner box has to get right.
Two quotations for the same 500 KWp rooftop can differ by a noticeable margin and use the same modules and the same inverter brand. The difference is usually buried in the cable schedule and the string layout, and one of those two plants will be quietly losing money every sunny day for the rest of its life.
Copper is a commodity with a visible price, so it is the easiest place to take cost out of a bid. Drop the DC cable one size and the AC cable one size and nothing visible changes at handover. The plant energises, the monitoring shows generation, everybody signs. The loss shows up as a number that is always slightly lower than it should be, forever, and nobody can point at it.
A cable is correctly sized only if it passes all three of these, and the governing one is often not the obvious one.
A bid built by picking a cable off an ampacity table and stopping has done one third of the work.
Published current ratings are quoted at a reference ambient, commonly 30 degrees C in air, and for a single circuit in a defined installation method. A cable tray on a factory roof in May, in the sun, with a dark tray under it, is nowhere near that reference condition, and the cable's own losses add on top.
Four factors compound, and each one is a multiplier below one:
Multiply three factors of roughly 0.8 together and the usable rating is about half the number in the table. That is the arithmetic that separates a cable schedule from a guess, and it is why the design deliverable you should be asking for is the derating calculation, not just the size.
DC string cable lives outdoors, in the sun, on a hot roof, for the life of the plant. The specification that matters is a purpose-made photovoltaic cable to EN 50618 or IEC 62930, the type usually designated H1Z2Z2-K: cross-linked insulation and sheath, halogen free, UV and ozone resistant, rated 1.5 kV DC, with a conductor temperature rating well above what building wire allows and a service life expectation stated at elevated operating temperature.
Ordinary PVC single core building cable used for the same job will do two things. The plasticiser migrates out under heat and UV and the insulation goes hard and cracks, generally within a few years and generally first at the point where it leaves the module frame and takes the most movement. And because it is single insulated, a chafe against a rail edge becomes an earth fault, and a second chafe on the other pole becomes a short circuit with a current source behind it that will not stop feeding the arc. Double insulation on PV cable is not a luxury, it is what allows positive and negative to be routed with confidence.
Tinned copper conductors are worth insisting on in a humid climate. So is a stated bend radius that the installers actually respect at the cable entry to a combiner or inverter, because a tight bend at a gland is where water eventually finds its way in.
Loss in a cable goes with the square of the current, which means the loss is at its worst at exactly the moment the plant is producing its most valuable energy. It is not an average condition, it is a noon condition, and it repeats every clear day for the life of the plant.
Sensible design targets, which you can write into a specification, are a low single figure percentage on the DC side from module to inverter, and a similarly tight figure on the AC side from inverter to the point of connection. The exact number is a commercial decision, and the way to make it properly is to compare the lifetime energy value of the additional loss against the incremental cost of the next cable size up. That comparison almost always favours the larger cable on a long run, and it is a calculation your EPC contractor should be able to show you rather than assert.
The point that gets missed is on the AC side. On a large shed the inverters sit near the array and the LT panel sits in a substation a long way off. That run is often where most of the drop is, and on a plant that exports, an oversized drop does more than waste energy. The inverter has to push its terminal voltage above the grid voltage to deliver current, and the further the grid voltage is from the inverter, the higher that terminal voltage has to go. Add a drop across an undersized AC cable to a distribution voltage that is already high at midday, and the inverter reaches its over-voltage limit, derates, and eventually trips. The owner sees a plant that cuts out at noon and blames the inverter. The fault is in the cable schedule.
The remedies, in the order they are usually worth considering: move the inverters closer to the point of connection, increase the cable size, or on a genuinely long run at a large site, step up to medium voltage and run at 11 kV. That last option adds a transformer and a switchgear panel, and on a big enough site it still wins.
Two calculations bound the number of modules you may put in a string, and both are done with the actual module datasheet and the actual inverter datasheet, not with a rule of thumb.
The cold case sets the maximum. Open circuit voltage rises as cell temperature falls, at the rate given by the module's temperature coefficient of Voc, which on modern n-type cells is a small negative percentage per kelvin. The string's Voc at the lowest cell temperature the array will ever see must stay below the inverter's maximum DC input voltage, and below the system voltage rating of every component in the DC circuit: modules, cable, connectors, fuses, isolators, SPDs. Exceeding it once, on a cold clear morning, can destroy an inverter input stage, and no warranty covers it because the log will show the voltage.
The design cold temperature is a regional decision in Bangladesh and it is more consequential than people assume. The coldest morning in Panchagarh or Dinajpur is a good deal colder than the coldest morning in Chattogram, and the relevant condition is the cell temperature at first light with the array at open circuit and irradiance just beginning, which is close to ambient. A string length that has a comfortable margin on a Chattogram rooftop can be marginal on a north Bengal site with the same equipment. Any designer working from a single national figure is designing one of your two plants wrong.
The hot case sets the minimum. Maximum power voltage falls as cell temperature rises, and if the string's Vmp drops below the inverter's minimum MPPT tracking voltage on a hot afternoon, the inverter can no longer hold the array at its maximum power point. You lose output at the peak of the day, which is exactly the energy you cannot afford to lose.
The mistake here is using NOCT as the hot-case cell temperature. NOCT is measured at 800 W/m2, 20 degrees C ambient, 1 m/s wind and an open rack with free air behind the modules. A close-mounted array on a corrugated shed with a small gap and no cross flow underneath runs considerably hotter than that. Use a realistic cell temperature for the actual mounting arrangement, and if the mounting is close-coupled to a metal roof, say so in the calculation and use a higher figure. Then check that the string still tracks.
A few related points that belong in the same calculation:
The cleanest DC combiner box is the one that is not there. Modern string inverters carry many independent MPPT inputs, and on a rooftop laid out sensibly you can often take every string directly to an inverter input and delete an entire class of enclosure, an entire class of fuse, and an entire class of fire risk. On a rooftop that is worth designing towards deliberately.
Where combiners are needed, this is what a proper one looks like:
The most common serious defect on a rooftop DC system is not the cable and not the fuse. It is a mated pair of connectors from two different manufacturers. They fit. They click. They look identical to anybody standing on a roof. The contact geometry and the spring material are not the same, the contact pressure is wrong, and the joint slowly heats until it degrades, arcs, and starts a fire that a DC array will happily sustain because it is a current source with no zero crossing.
The discipline is simple and it costs nothing: one connector brand across the whole plant, male and female from the same manufacturer, crimped with that manufacturer's matched die tool, torqued to the stated figure, and a sample pull tested at commissioning. Not pliers. Not a mixed box of whatever came with the modules and whatever the extension leads had on them. Ask your contractor which connector brand they are using and what tool the crimps are made with, and if the answer is vague, that is the answer.
Every one of these decisions is invisible on handover day and permanent afterwards. That is precisely why they are worth asking about before the contract is signed rather than after the first hot afternoon.