My New Build

[MMP]

Though I knew about the 4gig file limit, I recently had this issue on a client’s project.

I was recording a long audio book sporadically over a number of weeks, and kept continuing on the same track, ultimately reaching over 4 gigs and having the resulting corruption of the file.

Once the file is over 4gigs, though the data is still intact, the pointers to it are corrupted. So, what you were hearing is most likely some random data on your hard drive playing back in a corrupted manner. Even though a file is deleted, the data stays on until that disk sector is overwritten. The noisy audio would just be the misinterpretation of the old data.

[John Ludlow]

The recorded files can only be limited to around 4 gig... in a live recording situation, you would need to start a new set of files to record past that limit.

In general, this is around 4 hours max.

You can track a timeline up just past 24 hours, but the max length of any file is around the 4 gig limit before the position pointer loops back afround to zero.

Bob L

That makes sense. It sounds like a 32 bit integer issue. I have no idea whether 64 bit C++ compilers have 64 bit integers available - but I think they would solve that problem. It's probably a relatively small thing though for most people if it's sticky for you.

[John Ludlow]

Successive tests today culminating in this one:

48/24.
1, 32 sample, buffer.
60 FOH stereo input channels.
60 instances of FreeverbToo (one on each channel).
2 Monitor mixes with 60 stereo input channels each.
8 channels out (4 on FOH, 2 each on Mon1 and Mon2).
60 stereo tracks recorded.
65% SAC Load.
1 hr and 10 minute run.

Still no sign of sweating hard. Going up 8 channels at a time.
Phenomenal: this is a fairly pedestrian computer except for the NVMe drive.

May be too busy building a prototype bass trap cage (thank-you Angie!) tomorrow to continue testing until sometime next week.

[John Ludlow]

Finished a 4'x2'x4" bass trap prototype cage (ala this guy: https://www.radford.edu/~shelm/acous...ass-traps.html) (thank-you Angie!). My friend Ryan, the guitarist and building builder, offered to help me. But, ultimately he just did it. I spent my working life building virtual things or the machines to house them - and I was good at it. But, watching someone who can build real things, like: in the physical world things - efficiently and with precision - is such an extraordinary experience for me. His pieces all fit together perfectly. Everything is exactly the right size rather than a curf too small. Things that are supposed to be 90 degrees - really are 90 degrees. He's got little tricks and tools to improve each step. Much respect.

"We're" (lol) going to knock these all out over the next week. The following week is about the monitor shootout.

[John Ludlow]

Time was left over today after all. We're up to 2 mics into 84 stereo input channels; 84 reverbs; 2 x 84 stereo channel monitor mixers; 84 stereo tracks recorded for 2 hours +. SAC load: 90%.

It's working hard now. Not much load left. Surely the end is near. I haven't wanted to open Task Manager to find out and possibly use processing power that it needs, but, I at least presume that 90% load requires more than 2 cores.

That it's processing the mixing of 504 separate distinct signals (between FOH, Mon1, and Mon2) is pretty amazing. But the 84 parallel VST reverbs is what astounds me the most.

[John Ludlow]

Well, I didn't exactly break it. It was still amplifying and still recording. But, the last 16 stereo channels I added caused the SAC Load to ascend to 99%. And, I did peek at the processor usage. It was not using much of any but core 1. But core 1 was pegged. I had begun to notice that there were very small ticks that seemed to happen only when I spoke. Listening back to the recording at 92 channels, during the long period of time in which I wasn't there - it was totally silent. But, when I'd speak into the mic it would make the very occasional, barely audible, ticking noise.

In consideration of the theory that it takes as much processing power to mix and record no sound as it does to record something, digitally (which I can also make a case for) I doubted this had something to do with the load. And, again, when I listened to the recording during times when nothing was happening there were also no ticks.

But, then I did this last test at 100 stereo channels and 100 reverbs, and yada, yada, yada. Yes, it was still silent when I wasn't around. But, in the recording playback, when I spoke into the mic those occasional ticks had turned into almost static that followed the dynamics of my voice. When I paused they weren't there, when I spoke - they were. I presume these must be dropped buffers. This leads me to believe that the theory has a hole in it somewhere.

So, yes it's still running. But, it isn't useful anymore - and hasn't been since at least 92. It might well still be if I didn't have that many reverb instances running though. That said, given that the increase in signal strength caused distortion, it's also quite possible that had I been recording something with more amplitude all the time they would have shown up sooner. Maybe I should do just one more test at 100, but after having turned off all the reverbs, and see whether, and when, I hear dropped buffers. That's going to require that I go through and manually close 1 VST in each of 100 channels, which will be a royal pain. Maybe... tomorrow.

[cgrafx]

No hole. For the purpose if this discussion the data values are illustrative not real.

The incoming audio stream is filling a buffer with numbers 0-255, if the incoming audio doesn't change then the incoming buffer is filling with the same number.

So if the audio is effectively silent than the buffer is filling with a constant stream of 0 0 0 0 0 0 0 0 0 and thus the output stream is 0 0 0 0 0 0 0 0 0

If you drop a buffer or two or three or even 30 or 40 because the input stream isn't changing from 0 0 0 0 0 0 0 0 0 the output stream is still going to be 0 0 0 0 0 0 0 0 0. Even with chunks missing out of it.

However, when you talk into the mic, the input stream is now 0 10 20 30 40 50 60 70 80 90 100, etc etc. So if you drop three buffers the output stream is 0 10 50 60 70 .....

That gap where it jumps from 10 to 50 instead of the smooth transition from 10 to 20 is where you hear the tic.

Its not that the audio engine is processing more data or working harder its simply that you can now hear the anomaly audibly when it drops the buffer.

[cgrafx]

That's going to require that I go through and manually close 1 VST in each of 100 channels, which will be a royal pain. Maybe... tomorrow.

you can select all 100 channels and turn off or bypass the VST in a single click, you don't have to do it one channel at a time.

[John Ludlow]

you can select all 100 channels and turn off or bypass the VST in a single click, you don't have to do it one channel at a time.

Doh!!! Too late. That makes perfect sense. I don't know why I didn't realize it.

[John Ludlow]

OK, I've finished my last stress test. I manually removed all 100 reverbs and then ran the same test with just 100 stereo FOH SAC inputs and 2x100 stereo monitor inputs - and 100 stereo tracks recorded. I used 1, 32 sample, buffer and recorded in 48/24. Load dropped to 40% and it was clear as a bell still after an hour and a half.

So, I upped it to 119 for everything. I didn't use channel or track 120 - because I set that aside in SAC for playback from SAW. It doesn't go higher than that. Load stayed at 40%.

Nuthin' but net. Everything that was supposed to be there was and nothing that wasn't.

I did have a temporary problem when I went from 100 reverbs to none. Somehow tracks 93-100 got screwed up - or at least SAW would tell me that they were not in a format it was familiar with every time I tried to record over them. I had to start over on the SAW side. But... when I did - it worked perfectly. Not a single tick or crackle that I could discern. This time I played some PFunk into the microphones from the old system. Perfect. I don't know why anyone would ever need to record 120 distinct stereo tracks of anything, but if you ever did need to: this system could do it while polishing it's nails. There's seemingly a lot of capacity left that it's not using. It's still astonishing to me. Very impressive products. RML is back! I think the whole world should know about SAC and SAWStudio 64 and I'm planning to do my small part in trying to accomplish that.

We've been working feverishly on building bass trap cages for the past few days. The 24 slabs of 2" 703 arrive from GK Acoustics on Friday. The fabric is ready for them and it's looking as if the cages will be too.

Then on Monday the monitors in the shootout arrive along with their stands and cabling and do-dads. Just getting them downstairs safely is going to require young people (let alone getting them back upstairs). Two of the monitor sets are very heavy. The stands are very heavy too. We plan to record a sweep before and after the traps are deployed to prove our hard work meant something. Then I'll start listening to recorded music and re-mixing old projects through them after moving them over from 32 bit.

They're all powered monitors so I'll just be able to switch between them by using SAC outputs. Since I'm using each monitor set in a separate monitor out, I should be able to further tune it's output channel for the monitors connected to them (and the room). One of those sets, the Genelec 8341a's, have their own acoustics setup program and a special microphone to tune the monitors for the room. The other two I'll just have to use my ears and sweeps for. Very exciting time. I need to clear out a lot of boxes and bags and manuals before it really gets crazy Monday. I'm not known for my neatness - but it's gotten bad, even for me, in the past month with all the new gear.