Surround sound card with good ADC?

[tvih]

For noise/leakage/loop and all that, there's really no meaningfully audible background noise problem except with the line input in play. Although for an unknown reason the noise even with the line input is drastically (like, 90%) lower today than it was the first two days. Win10 ran some updates last night (old installation that hadn't been upgraded for some time, device drivers didn't change) but no changes to connections etc.

The DSP computer is an i5 4690k with MSI Z97M-G43 motherboard, single SSD, etc. I don't know the actual wattage it is using because I don't have a meter for that. I could use an even older desktop computer (i5-750 based) but at least it wouldn't be any more energy efficient.

The sound card arrives tomorrow so then testing will resume.

 

[tvih]

Initial test results... well, what audible background noise/interference remains with the sound card would appear to be from the ground loop rather than the line input itself. If I feed line input a signal from my mobile phone for example, there's no interference. But when feeding it from my USB DAC or via the AVR, there is interference. This really is the first use case where I ever have this problem. I guess a computer line input is more susceptible to the noise from a ground loop than the other gear involved. What a pain in the rear. I should've guessed it wasn't going to be this easy.

Mind you, it's technically "usable" as is now. But then again it was technically "usable" without DSP, too. Bleh. On the upside if I'm simply listening to music I could do that from the DSP computer itself and thus wouldn't have to worry about the loop as long as I turn off the line input volume. Downside is having to directly use the DSP computer rather than just leaving it be as was the intention

Would still prefer to get rid of the interference regardless. But I'm really not familiar with this stuff given lack of prior encounters with the issue. I can't afford the mixer right now assuming that would indeed help, even a cheap one. And adding yet another device into the chain is quite a drag to consider.

As for the sound card itself... well, there's the consideration of keeping it or not. It's certainly better than the ACL892 and has no audible background noise of its own. As such I suppose no other reasonably priced sound card would likely do much better, so it's more whether to stick with this approach or try something else (the only solution with similar budget that applies to all audio sources and that I can think of basically being getting a miniDSP 2x4 for the sub while using the limited PEQ in the AVR for the main channels). Though there's also some concern about whether or not ASUS is ever going to finish the Win11 drivers for it (they've been pushed back about a million times by now apparently), but at least for now the DSP computer is running Win10.

If I had gotten the Focusrite 2i2 and a Topping PA5, would a balanced connection between the two have broken the loop? As said, I really don't know much about this stuff.

EDIT: I just realized I had forgotten to point out that the electrical sockets in this crappy apartment aren't in fact even grounded except in the kitchen and bathroom. Since they're pretty much hidden away and I just deal with power strips, I didn't think of that before now. How does this affect dealing with the issue?

 

[AnalogSteph]

EDIT: I just realized I had forgotten to point out that the electrical sockets in this crappy apartment aren't in fact even grounded except in the kitchen and bathroom. Since they're pretty much hidden away and I just deal with power strips, I didn't think of that before now. How does this affect dealing with the issue?

Jeez, how old is this electrical installation? In the US, grounded outlets have been mandatory in new buildings since the early 1970s (and that was later than e.g. in the UK and Germany). One would think that the last 50 years should have been enough time to bring things up to code.

Technically, you aren't even supposed to operate IEC Class I equipment without a proper ground. Electrical safety is only ensured when the wall outlet is replaced with a GFCI-equipped one (do you have those at least?), and even then multiple devices on one outlet strip in particular may still provide enough mains leakage to give you a noticeable tingle when touching a metal case.

The article Cheater plug is a good read in this context.

This situation may be one step forward and two steps back. You may be able to avoid "hard" ground loops (by using multiple outlet strips) but may have fun with "soft" ones, plus the latent safety issues. If device ground is floating near half the mains voltage, open inputs also have a tendency to hum pretty bad (more so when touched) as you're always getting minimal capacitive coupling to the environment which is at or near ground potential. My JBL 104s in the office are like that (speakers with built-in SMPS complete with the requisite mains filtering, not earthed) - no hum at all when plugged in, but unplug them and touch the plug (or have it touch the grounded PC case) and all hell breaks loose, even at a modest volume setting. At least it means that the unbalanced input is actually usable.

The DSP computer is an i5 4690k with MSI Z97M-G43 motherboard, single SSD, etc. I don't know the actual wattage it is using because I don't have a meter for that.

A basic Kill-A-Watt is like 30 bucks. This stuff is not easy when you're flying blind. Package Power indicated by monitoring software is a good guide but not the complete picture.

On the software side, you can check processor C-states and clocks e.g. in ThrottleStop, HWMonitor (Windows) or via powertop (Linux). Package C7 should be reached on a regular basis.

I've had a look through the board manual, and going by that (there may have been changes to settings with BIOS updates) I would check the following options in the OC section:
CPU setting - all on default
DRAM setting - all on default (check whether XMP interferes with CPU C-states, disable if needed)
Voltage setting -
CPU Core/Ring/GT Voltage mode: Auto, Adaptive, or Adaptive + Offset for undervolting
Other setting -
CPU Features:
Intel C-State: Enabled or Auto
C1E Support: Enabled
LakeTiny Feature: Enabled (I think this is the same as the customary Aggressive SATA LPM setting)
Package C State Limit: Auto or C7s

Long Duration Power Limit (a.k.a. PL1) and Short Duration Power Limit (a.k.a. PL2) can probably be left on Auto unless a check in ThrottleStop (TPL button) indicates higher than CPU spec values. Those should be 88 W and 110 W, respectively. You may actually want to set lower values here e.g. to accommodate a PicoPSU or cooling solution.

Adjusting Turbo Boost clocks would have to be done within the OS.

I did not see any settings relating to PCIe ASPM. There generally are OS / kernel setings for this (e.g. in the current power plan in Windows), but if you can't get beyond package C3, this may be your culprit. This board doesn't seem to have any extra onboard chips that could screw things up, so that's good at least.

There is an EuP 2013 setting in Power Management Setup that I would test out, but it probably disables Wake on LAN along the way as it's designed to minimize power-off standby power draw. This is handy for all machines that spend a lot of time being turned off and are being turned on manually, and generally saves a watt or so when off.

 

[tvih]

It's a '70s building, with just normal "European" two-prong plugs (other than said kitchen & bathroom). New constructions without grounded plugs were forbidden in 1997 here (Finland) from what I read, but existing ones weren't required to convert. Something do with insulating floors and whatnot. Not much I can do about it regardless. And yeah, the watt meters don't cost all that much, but I don't generally need one, so curiosity alone wasn't enough to get around to buying one. Even in this case it's not like I need to optimize down to the last watt.

Anyway, I did kinda get the noise low enough, especially if using a separate amp for the rears so I can bring their level down - otherwise the problem was having to run +10dB fronts and -10dB rears from the AVR to match them with low enough line input level on the soundcard to not cause audible noise (unless you have your ear to the speaker). The boost didn't seem to cause any problems, but it's still not exactly optimal either.

Ran into a new unexpected problem yesterday when finally doing surround audio stuff, though. Well, at least unexpected in this magnitude. Delay. I knew there'd be some, but damn if it isn't way higher than anticipated. The bigger part of the problem seems to have something to do with the ASUS card. I measured with REW's acoustic time-alignment and got a 173.5ms delay for the ASUS compared to the rear channels that were plugged directly into the AVR, and 83ms delay for the ALC892. This was sending the signal from my main PC through the full "chain" of AVR, DSP PC, amp, speaker. The latter delay I would've not been very surprised about, but the former... yeah.

I then routed the rears through the DSP computer as well, and between various ways of wiring them the delays kept going kind of all over the place. There might be some errors there where the connections didn't go "right" but overall the delay kept changing when switching plugs around and then back like they were. Subsequent measurements without changing the connections yielded the same results like you'd expect, but then switch the cables to another config and back and bam, different. I don't get it.

But at least one thing was clear, the ASUS has a much bigger delay regardless. It was the laggier one in all possible wiring variations. Also worth noting that the extensive EQ curve I have on the fronts only impacted the delay by less than 1 ms! So it's definitely the rest of the signal path rather than the EQ being a slowpoke in both cases.

This presents a problem because the AVR's internal delay can only manage to compensate for 30ms, despite the distance you can set suggesting up to 60ms. Though even if I route the rears through the DSP PC's ALC892 to match the delay manually and accept the degraded signal quality for them, the overall delay of course remains high. (And speaking of the odd variances... upon setting up said rear delay I'm suddenly only having a 62ms disparity between the front and rear, despite it previously being over 90ms? Go figure, and who knows if it stays that way!) It's not so bad as to completely ruin lip sync, but even so far from optimal. I tried to look up ways to reduce it, but with the included "low latency" drivers already installed, all the optional processing from the ASUS driver turned off etc. I'm not seeing a whole lot of options. Obviously can't use something like ASIO/WASAPI either since that kills the EQ. By comparison miniDSP devices should cause less than 15ms delay in all use cases, much lower if not using Dirac.

So I guess it circles back to "why did I expect it to be easy/cheap"...

 

[AnalogSteph]

It's a '70s building, with just normal "European" two-prong plugs (other than said kitchen & bathroom). New constructions without grounded plugs were forbidden in 1997 here (Finland) from what I read, but existing ones weren't required to convert.

Ah yes, Finland. One of those countries that were a bit late to the party. How were Finns even supposed to be hooking up an ordinary PC before 1997 (PCs, monitors, laser printers, all traditionally IEC Class I)? Many people would have been on their second one by this point, perhaps even the third. It's not like Japan with that separate earth lead on the plug that you're expected to attach somewhere (not too clear on the specifics).

I would check the Asus for enabled APOs, which may be causing the added delay..Try some different modes and turn off unnecessary frills.

What are you trying to use as a DSP - CamillaDSP perhaps? The best shot at getting delay down in that seems to be the chunksize parameter, alongside switching to WASAPI exclusive mode.

So I guess it circles back to "why did I expect it to be easy/cheap"...

Yeah. It's not mainstream enough, so right now you can only get either easy or cheap(ish) but not both.

 

[tvih]

They would've used the computers etc. the same way I'm using them now? This isn't uncommon at all in these older buildings and it sure doesn't stop anyone. That said it's not like most new buildings wouldn't have been made with grounded plugs long before 1997, and many other converted, despite it not being required by law.

For the DSP/EQ I'm using EqualizerAPO. I doubt others would make much difference, since it seems it's the internal processing of the sound card hogging up most of the delay time anyway. If it was just the software pipeline, the ASUS would have a delay more in line with the ALC892. Similarly it seems unlikely any other sound card would have significantly less delay than the ALC in a fool I/O loop, meaning I'd need to DSP both fronts and rears anyway.

So as I see it, my options are more less:
1) Keep the ASUS sound card and just live with the delay (which seems to have come down slightly to 156 ms... for now)
2) Return soundcard and use AVR's PEQ. It's a bit limited and won't allow using the subwoofer because it can't be EQ'd separately and it's way out of control without EQ in this crappy concrete box of an apartment. But now that i've used it properly integrated, i want it. Would only have it for PC sources in this scenario though.
3) Return soundcard, try another sound card. Seems like a coin toss at best at this point, and as said seems likely to need to process both speaker pairs anyway. But technically could get the total delay to be shorter. Doesn't really feel like worth the potential hassle, though.
4) Return soundcard, get a miniDSP 2x4 or the HD version. Either use it just for the sub and use the AVR PEQ for the speakers, or if it isn't detrimental to the quality (user experiences vary) use it for the mains too. The latter would most certainly require the HD and thus cost 2.5x as much as the ASUS card. But at least it'd be low-latency.

Since none of these are meant to be a permanent solution, it's another reason to try and keep the price from ballooning. But the financial time-frame for a more permanent (and thus expensive) solution would be in about a year from the look of things, so i'd rather not be entirely without an interim solution. Though if the 2x4 HD would be good enough I guess it might not even be necessary to get a more expensive solution, at least in any hurry. But it's impossible to know without trying, and it's also "conveniently" out of stock for the near future at the only reasonably priced domestic retailer carrying it - well, except for the kit version, but no case for a savings of just 19€ is a bit of a hassle, and would make selling it later if need be more difficult. What a mess.

 

[tvih]

Oh well. Caved in and ordered the 2x4 HD kit. I'll be stuck with it even if it sucks, so I sure hope it doesn't. And I'll have to either design a 3D-printable case for it, or get crafty with other materials.

 

[tvih]

As a closing note the 2x4 HD seems to be performing up to par, so thankfully this is now "solved".