You'll learn more from that video with the volume turned down because what he does makes more sense than what he says. BBI mentions all types of problems from uneven trace lengths between transistors and transformers, to Rob using the wrong type of 50 ohm coax on the output transformer. He then shows us "the correct 12.5 ohm white coax”. If you look at the end result, most of what he says on audio, does not match what he did on video.
None of the trace lengths have been made symmetrical. The 12.5 ohm coax was never used and just replaced with the same turns ratio in heavy black Teflon wire. The biggest change here was adding some capacitive padding across the primary and secondary of the output transformer. What that did was match the transistors output impedance to deliver more PEP power into the 50 ohm load at this frequency and power level.
Make no mistake, this design is still broadband push pull even though he says otherwise. Having learned about spectrum analyzers, the biggest problem I have with this video is what looks like the intentional overdrive of the fundamental frequency done on the first test, to make the results appear horrible. BBI runs the stock test with the fundamental frequency well above the correct attenuation point, where it is now driven right off the top of the screen!
Using too little attenuation so that the fundamental frequency is driven off the screen is the fastest way to make harmonics appear much worse than they are, to the untrained eye. Luke is careful not to make the same mistake when he makes his second improved test. Then, you will see the fundamental has been attenuated down to about 10db below full scale. This is absolutely critical when determining harmonic content since you must be able to compare the size of the harmonics, against the full size of the fundamental frequency. You can't even start this process when part of that signal is hidden off the top of the screen.
The last issue is, you see how hot that output transformer was running at 1,400 watts? Improving the match made the RF power through that transformer double, at about the same DC current. The thermal gun never gets pointed at that output transformer after the power increases to 2,800 watts for a reason. 25 years ago "Dave" designed that style and size core as a heavy duty replacement for a pair of 2SC2879 transistors having a total dissipation of 500 watts.
The idea that this transformer core is suitable to handle four sections of transistors with 3000 watts of total dissipation, is insane. I would avoid any amplifier that tries to stuff four 1,500 watt transistors through one push pull output transformer. It's hard enough to design a transformer that can efficiently match two of these transistors without core saturation. Doubling that RF while still passing twice the DC current through it, guarantees core over heating.
Things like running the amplifier output right up against the power supplies current limit in order to provide amplifier protection, has other problems. If you get this at a correct point where it can protect the amplifier, the power supply can easily cause all of your positive peaks to flat top without you ever noticing it on the watt meter because the voltage must start to drop at or around the level of these word peaks. You've given yourself no headroom in the power supply because it's now being run at 100%, with no more current available. That increases the likelihood of power supply failure too.
The video also reveals some lack of technical understanding with things like "Flyback resistors" that are found in older CRT based televisions and confusing them with "Negative feedback resistors" used in amplifiers. Then, a "Crowbar" over voltage protection circuit is mistaken as an over current protection circuit. How big of a deal are these mistakes? Well, that depends on some other things like results. While I'm quick to point out mistakes, the same can be said for pointing out the constructive things.
Luke has indeed addressed a well known problem with this particular pallet that has driven the people that purchased it, absolutely nuts because nothing they can do, gets it to make more than half of rated power. The fact that Luke was able to identify which section of the amplifier was causing this problem and one way to fix most of it (other than transformer heat), says more about his technical abilities, than his words in the video. His clean construction skills have to be in the list of top builders in this field too. I'd just like to see competition, not cause any temptation to "fudge" test results on the SA.