First off, the size you have used for <assuming> each Motor will pretty much cover the voltage drop.
The 5% max quotation in the NEC is a FPN [Fine Print Note], which is more of a suggestion - not a mandatory "Design" limit [NEC is not a design manual... yadda yadda yadda].
Figuring the volts lossed would be done using the normal apparent items [such as distance, amperage and etc.], but could also be done using more elaborate methods - which include total Reactance of Conductors, Ambient Temperature adjustments and the like.
Which ever method you choose to do depends on how "Strict" you are and / or how "Strict" your Client is.
In the long run, what you measure at the Load's end is what matters.
Conflict... derating and applying Voltage Drop calcs together...
Solutions... Apply simple calcs, size one AWG size larger, complile advanced and complex calcs, "experiment"; all will be NEC compliant but will this satisfy the Client?
So what I am getting at (in a nutshell) is you have many choices on how to address this situation, but the simplest way to do everything is to figure a wire size that works with the
Derated Value, then check it to see what a simple voltage drop calculation turns up with that wire size. If that size "Borderlines" any suggested value (3% branch circuit, 5% from service to load per NEC FPN, or something a Client / Vendor suggests), then upsize until the numbers are better.
Using your numbers, here's what I see:
<OL TYPE=A>
[*]3 - 40 HP 3 phase Induction Motors,
[*]40 HP 3 phase Induction Motor @ 480 VAC = 52 Amps,
[*]3 - #6 THHN cu minimum per Motor,
[*]At least 15 volts lossed per Motor -52 amps, 350' run,
[*]9 current carrying conductors = 70% derating
[*]75 amps x 0.7 = 52.5 amps - derated value of #6 THHN cu
</OL>
Using the #3 cu conductors, I came up with the same voltage drop figure that Redsy did - using the "Simple Formula" (Resistance only), which is 7.8 volts lossed. This equals out to 1.625% of 480 VAC.
The "Best" approach here is to make sure the
Complete Voltage Drop is not excessive. This would be from point of service to the load, covering what the Actual Voltage is at the Motor[s] by measuring the voltage at the service and / or the MCC, applying voltage drop calcs to THAT voltage, then find out if it will "Work" with the nameplate voltage of the motor.
Wish I could give a clearer message and direct answers, but you might be able to use some of this for a good answer!
Scott s.e.t.