Hardware Insights Forum

This is so you aren't tempted to buy an (expensive) expansion sound card over an incredibly simple to solve, if not especially obvious, design flaw in many (if not most) PCs — which can be sidestepped at no cost, or repaired for real with a few small parts and some modification of the connector board.

The problem referred in the thread title is caused by either an unshielded cable being used from the audio header to the front headphone (hint, hint…) and microphone ports; what I'm guessing is an earth loop (though created within the PC, through its own case and by the current draw of its own components); or, indeed, the combination of both. The effect of those is that, while the rear sound ports on my D945GTP work more-or-less perfectly, the design flaws in the case wiring have taken the front ports aalllllllllll the way down to being completely unusable. I do hope to repair it eventually, but it will be a while before I have the means.

To sidestep the issue
Simply unplug the front audio cable, and if it's a very old board with the AC'97 type of codec, jumper pins 5/6 and 9/10 of the header (to pass the signal through to line-out). Sure, you won't get to use the front ports this way, but that's no different from what you would have got with the add-in card.

To repair the problem itself
Pulling this off will require you to be able to both solder and desolder double-sided PCBs, and probably also to have an engraver to modify the PCB itself — but if you succeed, it will definitely be worth it. First of all, if the front audio cable that originally came with the case was unshielded, you will need (if you use the microphone jack, at least, but it doesn't hurt to use the shielded cable as a matter of principle) to either assemble a shielded replacement from the parts, or salvage a shielded cable from another case that can be adapted to match the wiring of the subject (I imagine most of us will take the latter, easier way out). Over on the front connector board, you will probably have to desolder the ports themselves, remove them, and then reinstall them after the mod (and make sure not to mix up the headphone and microphone ports if they are indeed colour-coded — I'd suggest writing "H" and "M" on the bare PCB to mark their respective positions). You will need to take the front connector PCB and somehow split the earth plane of the headphone and microphone ports — and only them — from that of the I/O (USB, IEEE-1394, etc.) ports (which is, by extension, connected to the case itself). Pay special attention to the vias joining the top side and underside of the board (I'm presuming it's a 2-layer type), as if you forget about even one that does earthing, it will defeat your attempt at breaking the loop. After breaking the loop, a parallel combination of (A) a (low-power) resistor approximately 10Ω in value (any available type will be good enough) and (B) a ceramic capacitor (preferably, but not necessarily, C0G) of approximately 0.1µF (exact values are unimportant) is suggested "across" the break, to assist RF suppression, but this part may be considered optional. If you salvaged a cable with separate shield and return wires connected to separate pins at the case end, but the case you're fixing doesn't provide for that, you will need to figure out how to combine them yourself; without the return wire, the circuit won't be complete, and an unearthed "shield" won't. You will have to work out the rest for yourself.

So now you know why the expansion sound cards want nothing to do with the front audio ports. It would really be a trivial expense (i.e. the cost of shielded cable, a small resistor, and a ceramic capacitor) on the case manufacturers' part to make all the difference between unusable and perfect front audio, but in this sorry age of penny-pinching, some of them just don't seem to care.

For what it's worth, I have yet to notice the functional limitations of the sound chip on even this 8+-year-old board — most recordings have higher background noise anyway (yep, there's a statement swap, with the new statement being unequivocally true) — so the only reason I could justify an expansion sound card for my own use is if I run out of channels on the integrated version (which I don't see happening any time soon, as I'm currently using headphones for my main listening expectations). (Another statement swap here, as headphones are supposed to have HF roll-off and the treble boosting I set on the EQ for some months was really just giving me headaches.) The days when discrete sound cards were a necessity outside of professional audio and hardcore gaming are long gone indeed.

(Never mind the statement that was originally here; the transition from terrible monitor-integrated speakers just threw off my judgement. Another issue did crop up in the interim, though: The line-in on the Intel D945GTP picks up noise apparently from some other part of the board unless short-circuited. The Gigabyte H87M-D3H doesn't have that problem — not audibly, anyway; any frequency peaks can only be detected with a frequency analyser, and even then are all less than 15dB above the median — so I can only conclude that Intel's mainboard designers just suck. At least, that's one thing I do agree with Topcat on. Anyway, I understand now that 32Ω (let alone the output impedance of the headphone driver) is too low to really be audibly affected by (normal levels of) EMI, and only suffers from the earth loop part. EMI would still attack the front microphone input if the cable is unshielded, of course.

Another thing I found out is that size does matter for coupling capacitors, particularly with 32Ω headphones. Yes, you can compensate with the EQ, but it would be preferable if users didn't have to have electronics knowledge to get the desired bass response. )

Yes, it is indeed an earth loop.

Just for a test, I swapped the shield and return wires back at the case end of the cable so that, without the bridge-to-be-added, the return wire completes the circuit…at the cost of leaving the shield floating, but (EDIT: It's not that the floating shield did anything after all.) So long as the connectors don't touch the PC chassis, the headphones stay quiet aside from whatever signal is being played back at the time, but the moment the connectors do make direct contact with the case, the noise is back. Does that problem sound familiar to you? And, yes, I could add a spare 12Ω 0.25W (though 0.125W or even lower power would be fine here, as it's dissipating something like 1mW at most) carbon film resistor between the connectors and case without adding any detectable noise back, so I'm on the right track with the values. Though further calculations on my part suggest that 0.47µF would actually be a better capacitor value; that would of course require an X7R type, but this position really doesn't care about linearity. Perhaps in conjunction with a second smaller capacitor, a 1nF C0G type, just to be extra bullet-proof against ultra-high frequencies.

Curiously, after further research it does appear that the "unused" pin I suggested using for the shield here is, in fact, a duplicate earth under AC'97 specification (with my old AOpen case only being wired for AC'97). Only, AOpen didn't bother connecting it on their version of the connector PCB (despite that there would have been no cost in doing so). The "remaining" pin after that is a +5V supply, which few if any cases actually use.

I'll christen this problem "Earth Loop II", where Earth Loop I is the cause of mains-frequency hum when interconnecting multiple analog audio devices with their signal circuitry hard-earthed through the mains.

So you can hear it for yourself…

Doesn't have any relation to the problems described above but it's worth mentioning anyway: For some reason, running the Realtek ALC892 (and maybe some others, but I haven't tried any) with the output rate set to anything below 192kHz results in a weird echo-ey effect. Mind you, I'm not a believer in ultra-high sample rates (this article explains it in detail); it seems to me to be caused by a design shortcoming. At 192kHz it works fine, so I'll just keep using it that way (and not using exclusive mode). (I don't actually have any audio sampled at 192kHz, anyway.)

LATE EDIT: What actually happened there was that the "Headphone Virtualization" thing was set on, but for some reason it doesn't function in 192kHz mode. This is the thing that was generating the echoes, and they sound horrible. Now that I've unticked the box for that shit-sucker, changing the output rate to 96kHz gives a noticeably cleaner sound (without the treble pollution from 192kHz mode), and the lower rates seem fine too. Changing the rate might not have exactly the same effect with every similar audio codec, but given that the 192kHz rate is just silly anyway, why use it?

Yeah, "nice" trick there, Realtek. (Not that it isn't a decent codec, although it doesn't have the most impressive specifications.) Why do they even include such gimmicky "features", anyway? It would be far more useful to feature a simple, no-nonsense cross-feed.