Coupling capacitor sizes and headphones

[LongRunner]

As I mentioned in some other threads back in April, I have an awesome set of headphones. Being nominal 32Ω impedance, if coupling capacitors are needed (as with every common PC audio codec), they should be fairly large — I would recommend 470µF — to minimise loss of bass. At least, that's the idea.

But the Gigabyte H87M-D3H (among other boards) has only 100µF capacitors there. For feeding an amplifier input or 120+Ω headphones this is enough, but with low-impedance headphones there's a significant roll-off of the lowest octave. Yes, this can be compensated by using the equaliser (for the combination of 32Ω with 100µF, raising the 31.25Hz, 62.5Hz and 125Hz sliders by +5dB, +2dB and +1dB respectively will do the trick), but you do need to be aware of it (knowing both the capacitor value and the impedance of the headphones) to get the best response. As it happens, allowing for the somewhat accentuated bass of the headphones themselves, I didn't actually need to raise any of the sliders above ±0dB with the equalisation I use now:

31.25Hz: ‒1dB
62.5Hz: ‒3dB
125Hz: ‒2dB
250Hz: ‒2dB
500Hz to 16kHz: No changes (as the review explains, the high-frequency roll-off is intentional; I did attempt to compensate for it earlier on, which of course worsened the sound experience)

If the coupling capacitors weren't undersized, the desirable settings for the HP50 would be:

31.25Hz: ‒6dB
62.5Hz: ‒5dB
125Hz: ‒3dB
250Hz: ‒2dB (again)
500Hz to 16kHz: No changes (again)

Anyway, I got it to sound great, and that's what matters in the end. Getting the sound right is totally worth the effort; the thing about good sound is that it should be comfortable to listen to, in contrast to poor sound which results in fatigue.

(On an unrelated topic, why does Waterfox get so slow!? It's frust r at i n g . . . Especially given that SRWare Iron ran just fine on my old system. )

[c_hegge]

I'd try it, but I don't have an equaliser in the sound driver on my main PC. My GA-Z77M-D3H-MVP has a VIA audio codec. The stock windows 7 driver has basically no adjustments, and the VIA driver stops the rear audio jacks from working. VLC's equaliser doesn't go any lower than 60Hz, which is too high for your suggested adjustments on my 24Ω Sennheiser HD201. I did try lowering the 60, 170 and 310Hz sliders by a dew dB, but it makes absolutely no noticeable difference. I don't listen to enough music to make it worthwhile investing more than 30 seconds in tweaking it.

[LongRunner]

Well, that's a huge shame. Every half-decent PC audio manager that I've seen has the equaliser. IMHO, we would be much better off if we rid VIA from the industry (preferably with the same gusto with which Y1 capacitors kill signal circuitry — not even a BC546 can survive that without added circuit resistance). VIA

[c_hegge]

Yah. On a freebie motherboard, though, I won't complain. I do have a Creative X-Fi PCI-E card, but I have no spare slots on my board as my Graphics card takes up three and the last one is used by a TV Tuner. Also, creative are hardly better for drivers than VIA. The driver for the X-Fi crashes FSX all the time.

[LongRunner]

Well, it's known among real audiophiles that Creative lost the plot when they went from ISA to PCI cards. Of course, as no mainstream board has had ISA slots since about 2002, there is no worthwhile Creative card that can be used in a modern PC.

[LongRunner]

Well, guess what: I missed something. Having since obtained the PDF for the ALC892 (the audio codec used on the H87M-D3H), it shows 75Ω series resistors for each analog channel. These result in attenuation with headphones connected, but provide a more even power response into different headphone impedances and - more to the point - reduce the high-pass corner frequency for a given coupling capacitance.

100μF is still marginal, though, giving a −3dB frequency just under 15Hz. Actually, even the 10.6Hz −3dB frequency given by my earlier-suggested 470μF into 32Ω (based on no series resistance) can be said to be a bit on the high side. If we take 5Hz as the reference −3dB frequency, 330μF would fit well with the 75Ω resistors shown in the application circuit. Then again, lower-value resistors might be used in the interest of higher volume…

Anyway, I've further tweaked my equaliser settings in the meantime, ultimately settling on what gave the most uniform perceived volume at the 31.25Hz, 62.5Hz, 125Hz, 250Hz, and 500Hz points. This does have much more sub-bass than my previous set (which attempted to approximate objectively flat response), although the higher bass range is as before (unlike the "booming" characteristic of cheap speaker systems aimed at the general public). I also added a slight treble boost, giving:

31.25Hz +11dB (yes, you read that right)
62.5Hz +3dB
125Hz −2dB
250Hz −3dB
500Hz to 4kHz flat
8kHz +1dB
16kHz +4dB

The extra decibel at 8kHz is probably barely audible and the somewhat larger boost at 16kHz is partly based on a suggestion in the InnerFidelity review that "maybe 3dB more energy" in the highest octave would be an improvement (on which I stacked a further dB).

And as it apparently wasn't totally clear last time - this set is only valid for the NAD VISO HP50, and will not be accurate for other headphones (unless they happen to have an extremely similar frequency response). Besides, even with the HP50, it won't be to everyone's taste - although I do think I've nailed it for myself.