Well there you go. I did surprise me, and this now does NOT surprise me. Hypex does their own measurements and shows them on their website. When you look at them you'll think they're simulated. They are not!The Hypex has 7 dB higher max output (440 W vs 86 W). A usable 7 dB vs a 7 dB that doesn't matter.
OK! Usable 3.7 dB vs useless 7 dB then.
Not necessarily. You can have tons of feedback and loop gain and still have significant output impedance by design for a variety of reasons.
Negative output impedance is only one crude way to make more use of the back-EMF, the microphonic voltage. There are more elaborate schemes that can handle, for example, the varying DC resistance of the VC with temperature. Also provided that the reactive part of the impedance is stable one can well compensate for that as well.
There is zero damage on the frequency response. You simply EQ it to the desired target and obviously, same frequency response as seen by the driver terminals gives exact same SPL frequency response -- we did not change the driver, after all.
Actually, for the transducer designers who are "source impedance aware" it turns into beta = (BL)²/(Rout+Re). Until we finally have superconductors at room temperature the beta factors B² and L²/Re (the latter being proportional to the amount of copper in the gap, in the end) is brickwalling beta. We can only "outsource" some part of this parasitic resistance into a negative resistance and try take care of the side-effects.
Well, since normally both BL and Re are given there is no need to redundantly state beta.
OK, in that case let's view my comments as striving for being that friend of mine, who always inspires new idea's in my head on the subject under research, the fact he actually doesn't understand what is really going on notwithstanding.
So in the end, what does the load-invariant frequency response of Putzeys power amps accomplish? That does not flatten out a speaker that would normally have very erratic FR w/a non-load invariant amp (which would make sense to my limited intellect)?
I wasn't aware that you already have an application in mind, and you were referring to your design instead of for general use. In your case, the extra power of the NCx500 is wasted on the HF driver and the B200 should be a fine match for the LF driver too. However, I am also quite sure that the NCx500 will work just as well in your application, perhaps not as cost effective and/or convenient to source in Thailand (even though the NCx500 are probably made next door in Malaysia).
In the class-D regime, load-invariant frequency response at high frequencies is one of the most important design goals simpler designs with weak or absent post-filter feedback cannot achieve. By this class-D finally can compete with linear A/B.
Shorting means in form of copper rings and caps or even full sleeves are know for decades and are very effective at mid and high frequencies but at LF they do almost nothing, notably for flux modulation. Either you apply some balancing counter flux with a static helper coil or make sure the magnetic circuit is saturated so it cannot be modulated any more (the latter is what Purifi does: https://purifi-audio.com/blog/tech-notes-1/some-speaker-problems-that-needed-solving-1)
Definitely. Modern driver designs like from Purifi even use varying winding pitch on the VC to get a flatter BL(x) over the widest possible range (https://purifi-audio.com/ushindi)
Tx KSTR, learning all the time.
Not sure about Hypex in particular, but white papers published by for-profit businesses aren't generally peer reviewed. In any case, the claims by the AI about the benefit of the feedback design and resulting ultra-low output impedance on diaphragm motion are largely incorrect. This is evident from the basic physics: once the output impedance is a small fraction of the voice coil resistance, pushing it ever closer to zero provides no meaningful benefit in terms of motional control.
Wow, great article quote!
Ahem, the microphonic voltage of a VC is always proportional to velocity at all frequencies. The non-resistive, rather (semi-)inductive V/I-transfer impedance is the problem here, make the resulting force lag behind.
Thanks for your clear explanation. But also please note we were already at the point the DF itself was not a factor of increased control, but an indication of it, because it rises as a side effect of effective negative feedback and loop gain application, which, as you indicate, DOES improve load-invariant response and thereby pistonic driver behavior. Am i getting it right now?
