Topic: ISO: A Roland R07 recording with clipping main track and a backup safety track

[TheJez]

Hi all,
I'm working on a piece of software that will automatically merge 'in a smart way' the main track and the safety track of an R07 in case of clipping on the main track.
Unfortunately I don't own an R07, so I have no real-life test tracks available to test my software. Right now I'm testing with main/safety tracks with a 20dB offset that I artificially created, but I'd really like test it with real tracks.

If you happen to have a set of unprocessed tracks available (or have the chance to create them) where the main track is clipping here and there, I'd really appreciate it if you'd be willing to share these with me.
In fact, the more clipping on the main track, the better  , as long as the 20dB softer safety track won't clip...

Thanks a bunch!

[aaronji]

Sound Devices patent for their method might be informative. Lots of formulas...

[Rairun]

I don't have a Roland R-07, but I actually think this software would be a really cool thing to have.

[TheJez]

Sound Devices patent for their method might be informative. Lots of formulas...

Thanks for pointing this out! Very useful reading. Here's also a link to a version that includes the images, which makes things more clear: https://patentimages.storage.googleapis.com/a7/6b/f5/77e31e68cca8b7/US9654134.pdf

In fact, while I was working on the software program, I started thinking things might not be so easy as I initially thought. So far the program calculates the gain delta between the two tracks for both left and right channels, but then it made me think:
- Gain will likely drift over time independantly for both main and safety tracks (e.g. due to temperature fluctuations)
- There will be a DC offset (no matter how small), which also will drift over time
I'm not sure how big the impact of the DC offset is, and how much gain and DC offset will drift in real life situations, but it likely means I'd have to keep track of 'the current gain delta' and 'current DC offset' of both tracks in order to be able to accurately convert samples from the safety track to the level of the main track. These are some of the things that are addressed by the techniques described in the SD patent.
Another thing that's tricky and I didn't realize so far, is that a 'clipping event' can have nasty effects on the waverform for a relative long period (e.g. hundreds of milliseconds) even after the input (and output!) has lowered below clipping level, all depending the amount and duration of clipping. Dealing with this effect is also covered by the patent. As I don't have an example set of tracks of the R07, I cannot say if this effect also applies to that recorder, but I guess it will... Finally there's the 'delay difference' between the two tracks. Apparently SD's also noticed a varying signal delay in the output of the two ADC, which they apparently correct for. I don't know if it's a practical issue or a theoretical issue, and if they really correct for it, but at least it's mentioned in the patent.

Now I don't want to copy-paste the methods of solving such issues from the patent (even if I could! I'm not a mathematician  ), as that would mean copyright infingement. And if I'd use more simple methods to solve these challenges in my program, I'd likely never reach the output quality of the SD devices.

I am curious how Zoom and Tascam merge the outputs of the two ADC's into a single file. Will they use the SD algorithms and pay some license fees (or not)? Have they developed their own algorithms, and how would these work? I guess nobody here knows, and those who do know are likely not allowed to tell...

[TheJez]

I don't have a Roland R-07, but I actually think this software would be a really cool thing to have.

Hi Rairun,
Thanks for your interest, I totally agree. However, things got a little more complex than I thought, and there doesn't seem to be too much interest for it. Although it's fun to work on this, it's not very motivating if nobody's going to use it. I even don't own an R07 myself. I guess this idea popped up a number of years too late, while the market is nowadays full of nice and affordable 32bit float recorders that can do the job well.

I might give it another go if some kind R07 owner is willing to make a nice clipping recording of his/her home hifi set and providing the resulting files to me. Until then, I think I will abondon this idea as 'being too complex to solve with sufficient quality in my limitted spare time'. I did learn a lot, though 

[adrianb]

TheJez - I have found the live recording with the clipping track, and associated safety track. I have cut out all the racist and bigoted chat between me and my mates before the performance (joking) so it’s 34 mins long. It was a short set.

I have my safety track currently set at 12dB, I assume it was the same at the time.

The problem is that I have no idea how to get two 524MB files to you.

[TheJez]

TheJez - I have found the live recording with the clipping track, and associated safety track. I have cut out all the racist and bigoted chat between me and my mates before the performance (joking) so it’s 34 mins long. It was a short set.

I have my safety track currently set at 12dB, I assume it was the same at the time.

The problem is that I have no idea how to get two 524MB files to you.

Ah, great! Don't worry, I will ignore the chatting  . The program-in-the-works already detects the level difference between the two tracks, so that's fine...
I will pm you on how the get the tracks to me. Thanks a lot...

[TheJez]

Adrian kindly provided me a set of main/safety track files where the main track is clipping a bit. Thanks a lot for these!
I guess we'll continue in the other thread discussing the options for auto-merging the main/safety track: https://taperssection.com/index.php?topic=205959.0