Problem withs "complex loads" are only problems for poorly designed amplifiers, be they class AB or D or something else. There is no evidence that PA5 is one of them.
The FR curves pma showed aren't any surprise. Amir didn't measure the A07 FR at different loads, but he did for the A05. You can see the FR differences at the highs with different loads, and I expect the A07 to be similar. It is no big surprise that the FR of an A07 as measured with speakers is not the same as the Hypex UcD (assumed invariant to loads). When you look at the PA5 FR, there is no difference for 4 & 8 ohms until beyond 20 kHz. Again, not something to worry about.
I agree this shouldn't be a problem for pa5. It seems like pma tested 07 at 4 and 8 ohms
here:
Difference between 4ohm and 8ohm at 20khz is 0.8db for 07. pma managed to get FR to be off by 3.2db in his load. pa5's 4 and 8ohm curves in Amir's testing have difference of <.1db. 0.07db? This is like a x10 difference in FR difference at 20khz for 4 vs 8ohm. For pa5 to fail transparency its behavior would have to be way worse above 8ohms.
Now on to how much SPL we need. The good thing about digital is its max output level is fully deterministic. If you have a voltmeter, disconnect your speakers, set the volume knob at the highest you'll listen to, play a single frequency tone signal at a known level, say -10 dB full scale (you can use Audacity or some websites to supply the signal), and measure the output RMS voltage. (Use a 60 Hz signal if you are using a cheap multimeter.) You can easily convert that into power at whatever load resistance you are interested in. Scale that number to 0 dBFS and you've the max power your amp will put out. Apply whatever headroom you need for EQ, and you can easily judge whether you have enough and if your Metas can handle it. (I personally wouldn't worry too much with intersample-overs. But if you do, then multiply the number by a headroom factor of up to 2 to get your required maximum power.)
I don't own a voltmeter and I'm more worried I'm going to electrocute myself. xD
I *just* realized... Amir's graph is SPL unweighted. So while it's conceivable for 100hz or 120hz to be at 95db, there is no way thd at 1.6khz can be heard since playback levels will never, ever reach that high.
As frequency increases from 100/120hz so too does my ability to hear distortions, BUT playback level decreases. So, distortion does not look as bad as it might seem. 96db graph might be what bass thd looks like and 86db graph might be more what treble looks like.
...class-d amps are designed to take advantage that our audible range stops at about 20 kHz. They amplify audio by converting it into high frequency pulses (in the case of PA5, ~600 kHz), which the electronics can operate with little losses. The pulses are reconverted back to audio with low pass filtering. Therefore, they, by design, do not have bandwidth to >100 kHz.
It's rough, because clearly some technical members on ASR dislike most cheaper class D amps and other technical members on ASR think they're full of it. Beyond FR issues already addressed, I don't worry about pa5's performance.
---
So looking at a thd% graph of a speaker which struggles are loud bass, with a sub and crossover would the "real world" thd% actually be lower than what is measured since the speaker is no longer asked to produced 20hz-20khz? I guess that depends on how thd+n is tested.
For subwoofer crossover frequency I guess it's best to test it in the room myself and see if I can localize bass. Some people said room modes helped them localize bass but with EQ and right placement in music it's not localizable. 4th or higher order slope with dsp?
If my math checks out (please correct me if it doesn't!), then the PA5 is clipping at 18.26Vrms into 4Ω, which at 19.1dB gain equals 2.01Vrms input.
With 8Ω load clipping point is at 19.63Vrms ergo 2.16Vrms input.
So it seems like the PA5 was designed with 2.0Vrms DACs as the intended signal source. Any higher signal amplitude at the input will produce significantly more distortion at the output.
The input sensitivity of pa5 is 2.6v according to spec. I thought using the unbalanced adapter to get my o2/odac to work with pa5 would lead to half the voltage and thus less power to Metas, but I guess it worked out.
Voltage Sensitivity = [sqrt(83w*4ohm)]/9.02 = 2.02v
Max input drops at lower impedances... Meta hits 3.7ohm, so go for slightly under 2.02v ideally? Too close to be audible likely but still. 1.9v?
