New popping noises and static -- video driver was main issue

This started a few days ago. Pops or brief static noises a few times a minute while playing, only after playing for about 15 - 20 minutes. It stops when I take a break and start again without unloading the organ. It happens with any sampleset.

Hauptwerk 8
Windows 11 22H2, Dell XPS 8950 i9-12900K 128GB
MOTU 24Ao

During the problem, using Caen, CPU utilization is 5 - 10% (normal), Hauptwerk's CPU meter highwater mark is 2 - 3 green bars, plenty of free memory, no disk activity, network off, services and apps disabled where possible. Core temperature stay moderate. No recent updates.

I tested each speaker in mono playing full organ and fonds using Caen through one speaker for a while on each, and heard no noise. So I think it's not a dying amp or faulty cable.

I suspect my audio interface is giving up at 7+ years old, but I'm not sure what else to test to prove or disprove that. I hate to order a new one for $1000 and then find out I didn't think of something to check.

Any other testing ideas? Thank you.

[mdyde]

Hello Mark,

[You mentioned re-loading the organ; simply restarting audio/MIDI (which restarts the audio driver) via the Engine menu may achieve the same result.]

Some tests you could try (stick to just using St. Anne's, as a consistent and simple reference point):

- See whether Hauptwerk's built-in audio recorder records it.
- Try the ASIO4All driver with you audio interface.
- Try the ASIO4All driver with the PC's built-in audio output, if any.
- Try not running the Hauptwerk process at real-time priority, in case it's interfering with a driver, for example.
- Try disabling the wind model.
- Try binding vs. not binding to CPU cores.
- Try running LatencyMon to see whether it reports any problems when the noises occur.
- Check audio cable connections and grounding/interference.

Thank you!

Since it's a new problem the past few days, I went to LatencyMon first, and it found two problems:

1. There were some 15 ms ISR to process latencies, which I tracked down to a stupid mistake. A week or so ago, I connected to the Internet to get iLok to recognize a new license, and slapped airplane mode back on without remembering to disable the wifi adapter again. Disabling that cleared up the long ISR latencies.

2. There is, and has always been, 2 - 3 ms peak DPC latency due to a NVIDIA driver, which I can't disable since I need video.

Anyway, I need to play longer to retest it, but it looks like I got bitten by 1. and possibly 2. only after I switched back to binding threads in HW a couple days ago. Unbound does make it more resiliant, and I don't know any way to fix the DPC latency in 2 on this machine.

As a side note, with bound threads, LatencyMon itself interferes with HW to the point of making the CPU meter fully red all the time. Unbound, HW has 0 - 1 green bar with LatencyMon running. Both realtime.

[mdyde]

Thanks, Mark. Hope your changes have resolved it.

Seems to be good now. Thank you.

I spoke too soon. While the popping noise is more muted, it's still there at times. The longer episodes of static are gone, but a quieter pop happens a few times per minute. Playing with full fonds and some bass makes it more likely (but chorus reeds might hide the noise). Playing more quietly doesn't seem to do it. Any sampleset makes noises, but high demand ones like Nancy and San Francisco do it more often.

A Hauptwerk audio recording, taken while noises were audible, plays back clean on another device, listening loud on good headphones.

The noise is audible playing Hauptwerk through speakers and on headphones, both of which are through a MOTU 24Ao.

Martin, I didn't try substituting ASIO4ALL, as you suggested, because I'm not sure I can put the manufacturer's driver back again. The MOTU driver is from Jan 2023, but I installed it about a month ago. Possibly that could be the problem. I'll see if I can get the previous driver back in case that has something to do with it.

When the popping noise happens, LatencyMon and the Hauptwerk CPU meter don't show any changes, and Task Manager doesn't show a spike on any of the cores. Running LatencyMon does seem to make the pops more frequent.

Binding threads makes it worse, as does 96k sample rate or a smaller buffer size.

Is it possible for buffer deadlines to be missed but the CPU meter doesn't show more than 2 - 3 green bars? I guess it only estimates the time available periodically so missed buffers can sneak by.

LatencyMon shows long DPC latencies for the NVIDIA driver, almost 2 ms, but that happened previously when there was no noise, from the beginning with this PC. The DirectX graphics driver approaches 1 ms, again nothing new. Longest interrupt to process time is a little over 2 ms. Almost all ISRs and DPCs are on core 0.

Still testing. I'll try playing back some CD tracks on the computer through the audio interface, reverting the driver, and seeing if St. Anne's makes the noise (I don't have it configured).

[mdyde]

Hello Mark,

I'd suggest sticking to just using St. Anne's for now, for testing:

- Try the ASIO4All driver with you audio interface.
- Try the ASIO4All driver with the PC's built-in audio output, if any.
- Try a different audio interface, if you have one to hand.
- Try not running the Hauptwerk process at real-time priority, in case it's interfering with a driver, for example.
- Make sure the MOTU firmware is up to date.
- Try reverting to the older MOTU driver.

There shouldn't be any risk in installing ASIO4All -- it shouldn't affect any other driver. After installing and configuring it, you can safely switch back and forth between it and the native MOTU driver (or any other driver) via the "General settings | Audio device ..." screen in Hauptwerk. Doing so won't lose any settings on any other screens.

Either way, on that screen also make sure that "Number of buffers" = 1.

Is it possible for buffer deadlines to be missed but the CPU meter doesn't show more than 2 - 3 green bars? I guess it only estimates the time available periodically so missed buffers can sneak by.

It's technically possible that a spike could be missed if it was extremely short, but even then it's very unlikely. If buffer under-runs were occurring regularly (as yours are) then the chance of none of them being visible as spikes on Hauptwerk's CPU meter is essentially zero.

Given that Hauptwerk audio recordings don't exhibit the problems, and Hauptwerk's CPU meter is consistently shwoing only green, I think the problem must be occurring after the audio has been successfully streamed from Hauptwerk to the driver, I.e. the problem is probably occurring within the audio interface, or its driver, or beyond that (cables, amplifiers, speakers). However, given that you mentioned that restarting audio/MIDI (which restarts the audio driver) resolves it for a while, it's probably within the driver or audio interface, rather than the audio cabling, amplifiers, or speakers.

Since the popping occurs with headphones, the only cable that's in common between headphones and speakers is the USB to the 24Ao. That also rules out speakers and amps.

I didn't know what Hauptwerk's audio recording being clean meant, but with that and the consistently green CPU meter, it looks like I can stop worrying about other CPU loads possibly interfering with HW and focus on the audio interface and its driver next.

Thank you!

[mdyde]

Thanks, Mark.

I tested with St. Anne's, and the noises happen there, too. Rolling back the MOTU driver didn't fix it, so I reinstalled the current driver. The firmware is up-to-date as well. I haven't tried ASIO4ALL yet -- I'm having trouble talking myself into trying a 3rd party driver when MOTU's has done the job for over 7 years, but I'll get to it.

Not running realtime doesn't seem to change the problem unless LatencyMon is running -- then it's worse.

I bypassed the MOTU 24Ao to route output through the RME UFX to my headphones and played for an hour or so with three organs. No noise.

Next I'm planning to use Windows Media Player to play some CD tracks through the MOTU and see if popping happens without Hauptwerk or the ASIO driver involved. Will test a different USB cable and port.

Thank you.

ASIO4ALL also makes popping noises through the MOTU 24Ao, so it isn't specifically the MOTU driver.

Windows Media Player doesn't show the MOTU channels as output options, so I can't figure out how to try a playback from the PC to the 24Ao without running Hauptwerk. Maybe I can route a tablet's headphone port to the RME UFX and then ADAT to the MOTU to test playback standalone. (The 24Ao doesn't have any analog inputs to plug in a sound source.)

That should finally prove if the noise source is the 24Ao and not the computer. I'm suspicious of both based on the evidence so far, but leaning towards the audio interface because the PC and HW drive the RME UFX cleanly.

Unfortunately, I can't keep ignoring our house guests right now!

I figured out where the MOTU was disabled for Windows sound (stupid user trick), tested a CD track with Windows Media Player, and the playback had the same popping noise as Hauptwerk. Some progress at least.

I also tried moving interrupt affinities to distribute DPCs better -- almost all were on core 0 -- but the big offender NVIDEA ignores the mask, and moving MOTU PRO Audio to other P cores 8 10 12 14 worked but doesn't change the noise.

Still need to recable a bit so a tablet can play into the audio interface and eliminate either the audio interface or the computer as the cause, but I need daylight for that. Not enough light around the gear rack.

[mdyde]

Thanks, Mark.

Both ASIO4All and Windows Media Player would call some driver components from MOTU driver (they just wouldn't call the MOTU ASIO driver itself), so the fact that they both exhibit the problem too doesn't necessarily eliminate MOTU's driver components entirely. I.e. it could still be a problem in (or related to) the MOTU software, rather than hardware. Your test of trying feeding audio into it via ADAT is a good idea for helping to determine which, I think.

Success, so far! It turns out not to be hardware or MOTU software related.

A recent Windows update included a new NVIDIA driver, 546.17, which made the really bad DPC latency on that driver get even worse, apparently enough to tip my audio over the edge into frequent glitches. All my interrupts and almost all DPCs were executing on core 0, so the extra load interfered with the audio driver enough to be heard.

Two things improved this, subject to a lot more organ playing over the next few days to make sure it's reliable:

1. I used GoInterruptPolicy.exe (a revision of the old MS Interrupt Affinity Tool that works on Windows 11) to assign the NVIDIA drivers to another P core and rebooted. This moved about 1/4 of the interrupts and all the really long DPCs to another core with Hauptwerk playing. The audio popping noises stopped, but high latency was still reported for that driver, just on a different core. This bandaid will have to be redone after any driver updates. I also moved the MOTU PRO Audio driver to a different set of cores, which moved the DPCs but not the ISRs. Anyway, spreading the DPCs around did help.

2. In the NVIDEA control panel, under 3D Global settings, I set Low Latency Mode to Ultra and Power Management Mode to Prefer max performance and rebooted. Reported latency for that driver dropped from 2 - 4 ms down to less than 0.5 ms. From reading, I'll probably have to keep redoing this fix, too.

NOTE: In 2., it may suffice to set Low Latency Mode to Ultra and leave Power Management at Optimal. The GPU temperature goes high otherwise, and it worked at low latency without the power setting. The setting does "wear off" at random and need to be reapplied. Not reliably persistent.

There's a note that driver 512.95 or .96 was free of this problem, and anything from 516.1 upward will have it. I didn't try the older driver, but that would be my next step if this resurfaces, assuming 512.x supports my graphics card.

This is all from searching for "NVIDIA DPC latency".

It's too bad Dell disabled the Intel built-in graphics port when they forced me to buy a graphics card. I can't find a way to restore that and turn off the NVIDIA thing.

It's too bad Dell disabled the Intel built-in graphics port when they forced me to buy a graphics card. I can't find a way to restore that and turn off the NVIDIA thing.

Mark,

Thank you for posting your troubleshooting steps and resolution here on the forum. I too use all MOTU equipment for my VPO business and was getting concerned regarding the 24Ao. For new VPOs I assemble a computer with specified components. I get case fans, CPU cooler, and power supply all rated super quiet. MOTBs I use have 2 on-board video ports and are capable of 128GB of RAM. For budget or very basic instruments I can get by with a refurbished Dell i7 Optiplex maxed out on RAM. These are very smartly built and I've never had an issue with them in the field, meaning they boot and auto-start Hauptwerk everytime!

Regarding MOTU, the new UltraLite MK5 is an amazing unit! I think the sound production is step above (but not by much) the older units like the 24Ao. Its ready to play like in 2 seconds of powering it on, so I can eliminate the HW audio delay completely!

Danny B.