Announcement

Collapse
No announcement yet.

Fixing slight audio drift

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Fixing slight audio drift

    Hey there. In our multi-cam setup we record two separate audio tracks into a Tascam DR-70D... one for the teacher (lav mic), and the other for room mic (pzm). We take a line out of the recorder and run it into our main camera. The Tascam generates a slate tone which I use to manually sync the recorded audio from the Tascam with the video from camera one in the edit. The other two cameras are just recording audio with onboard mics... although I'm considering using a DA to send signals to each of the cameras for faster manual syncing. Each of the cameras are about 20 - 25 feet from the teacher.

    These classroom sessions are long recordings... up to 2.5 hours. I don't stop the cameras or the Tascam during the break so that I only have to go through syncing the media once... thinking that will save me time in the edit. However, I've noticed that the further into the recordings we get, there is audio drift between the cameras and the Tascam audio.

    I tested this by placing video from camera one onto the timeline (this is the camera taking a direct feed from the Tascam as it's recording). I then took the wav file from the Tascam and placed it in the timeline then matched them by the slate audio wave form/sound. During playback, it takes about 6 minutes to start hearing phasing... and then the further it goes the more noticeable the drift becomes. At an hour in it sounds like they are about a frame or more apart. This is strange because, again, the audio source for the camera is the live output of the Tascam. Why would the two be different?

    The cameras (Sony AX53) - and Tascam are each recording at 48k (although different bit rates, which shouldn't effect timing).

    So, I'm wondering...
    1. Any ideas what is causing this?
    2. Any ideas what I can do about it?
    3. Any ideas about changing my workflow to avoid or minimize this problem - either at the recording stage or in editing?

    Thanks for your help!
    Mobile setup: Edius X - Laptop running i7-10875H CPU @ 2.30GHz 2.30 GHz (8 cores/16 threads) with 64GB RAM and NVIDIA GeForce RTX 2080 Super with Max-Q Design w/8GB VRAM.
    Desktop: Edius X - AMD 5950x installed in Asus ProArt Creator B550. 64GB Ram, 2 x 1TB m.2, 6TB HHD. EVGA 3080ti w/12GB vram. Running Windows 10 Pro

  • #2
    The first question is, What seems to be drifting? The audio recorded on the camera or the Tascam? I would suspect it is the Tascam. The audio out of the tascam feeding the camera should just be an analog feed, while the tascam is digitizing the audio and recording it on it's card.

    If what I am suspecting is correct, I would say that the oscillator in the Tascam is not sampling at exactly 48kHz. This is not uncommon when using a double system like this and doing long recordings.

    I personally use a Zoom H4n Pro to do similar things when I am doing event videography and have never noticed drift that quickly. I am often letting the cameras and zoom run for 60 to 90 minutes and everything stays in sync. Perhaps the clock in the zoom is more accurate than the Tascam.

    I would recommend putting something like a Beachtek DXA-Micro Pro or DXA-Micro Pro+ on the camera so you can feed both audio sources, (Lav and Room mic) to the camera, then you won't need to do any A/V syncing. The main difference between them for your purposes is that the Pro takes a 9v and the Pro+ has an internal rechargable battery
    https://beachtek.com/product/dxa-micro-pro/
    https://beachtek.com/product/dxa-micro-pro-plus/

    You can still record on the tascam if you wish as a safety and pass that same output to the camera, but only resort to it if you have to. The beachtek simply allows you to deal with 2 audio sources independently and also allows you some addition level controls to make sure you are not clipping the input on the camera.
    Edius WG 9.55.7761, various 3rd party plugins, VisTitle 2.9.6.0, Win 7 Ultimate SP1, i7-4790K @ 4GHz with HD4600 GPU embedded, MSI Z97 Gaming 7 Motherboard, 32GB Kingston HyperX RAM, nVidia GTX680 4GB GPU, Matrox MX02 Mini MAX, Corsair 750W PSU, Corsair H110i GT Water Cooler, Corsair C70 case, 8TB Internal RAID 0/stripe (2x4TB Seagate SATAIII HDD's, Win7 Software stripe), 1TB Crucial MX500 SSD, Pioneer BDR-207D, Dual 1920x1080 monitors (one on GTX680 and one on Intel HD4600).

    Comment


    • #3
      Typical drift between even cameras. It is the clock drift in the recorders so that a realtime feed would be fine but the recording output would drift. That is why I use Vegas to ensure audio is in sync when I need audio from several sources. In Vegas you can squeeze or stretch audio on one track to fit the reference track without pitch shift. Usually the closeup camera is reference so that lip sync is maintained. With DV or I think some consumer AVCHD cameras audio can drift up to a frame from video and still be within spec. Don't have the problem these days as I record to the GH5 two quality tracks and the others cameras are just for sync guide. In the past it was a problem with DV and HDV though and not as good a set of equipment as I have now.

      BernH has given you the solution I think. Don't know how you set up the cameras but AX53 does have timecode so you could set all of them and then start them all at the same time with a single remote. I used to start all my Sony's this way . Just use the GH5, GH5S and the AX100 now so they are different setup.
      Ron Evans

      Threadripper 1920 stock clock 3.7, Gigabyte Designare X399 MB, 32G G.Skill 3200CL14, 500G M.2 NVME OS, 500G EVO 850 temp. 1T EVO 850 render, 6T Source, 2 x 1T NVME, MSI 1080Ti 11G , EVGA 850 G2, LG BLuray Burner, BM IP4K, WIN10 Pro, Shuttle Pro2

      ASUS PB328 monitor, BenQ BL2711U 4K preview monitor, EDIUS X, 9.5 WG, Vegas 18, Resolve Studio 18


      Cameras: GH5S, GH6, FDR-AX100, FDR-AX53, DJI OSMO Pocket, Atomos Ninja V x 2

      Comment


      • #4
        Thanks for the input. Really appreciate it.

        BernH: Not sure how I would tell whether it's the camera or recorder that's drifting. Any ideas on that? I bring the Tascam audio into a DAW for prep before importing it in Edius. My preference is to use a dedicated audio recorder which gives me a limiter, more visible level control, etc.

        Ron Evans: For clarity, are you saying you've seen this drift with the cameras? Also, do you happen to remember what remote you were using for your Sonys? I'm not having much success locating one that doesn't require a IR cable to the camera... which would prevent starting all three at the same time.

        Hope I don't need to invest much more to get the audio tightened up. I have a bunch of different audio tools. I'll also check and see if any of them can squeeze/stretch audio without pitch issues.
        Mobile setup: Edius X - Laptop running i7-10875H CPU @ 2.30GHz 2.30 GHz (8 cores/16 threads) with 64GB RAM and NVIDIA GeForce RTX 2080 Super with Max-Q Design w/8GB VRAM.
        Desktop: Edius X - AMD 5950x installed in Asus ProArt Creator B550. 64GB Ram, 2 x 1TB m.2, 6TB HHD. EVGA 3080ti w/12GB vram. Running Windows 10 Pro

        Comment


        • #5
          Just checked with my AX53 and the RMT 835 which will start recording but will not stop ! Not all controls work either. The AX53 can be controlled with a Smart phone though too. I have not tried that as my use of the AX53 is for handheld shooting. Will find some more remotes to see if they work and report back if I find one.

          When I first started shooting video as a hobby I had a range of cameras, all tape of course, and they drifted. The specs for lots of consumer cameras did not have locked audio to video clocks so the audio will drift from video mostly within a frame. This became much less of an issue when tape was replaced by memory cards. But I can still see a difference between the audio on the AX100 and the GH5's over two hours. Very minor but would be heard. External recorders will almost certainly drift compared to video cameras unless they are real pro units with genlock sync. This is why it is best to record your audio to the main camera if possible as BernH mentioned. I have the XLR unit for the GH5 and that is the main audio for my shoots one channel for a shotgun mic on the camera and the other a house feed ( which is also an offset as the mic is closer to the stage for the house feed so has to be corrected anyway ). The GH5S has a Rode Stereo mic and so does the AX100 for room and audience audio.
          Last edited by Ron Evans; 05-13-2021, 02:04 PM.
          Ron Evans

          Threadripper 1920 stock clock 3.7, Gigabyte Designare X399 MB, 32G G.Skill 3200CL14, 500G M.2 NVME OS, 500G EVO 850 temp. 1T EVO 850 render, 6T Source, 2 x 1T NVME, MSI 1080Ti 11G , EVGA 850 G2, LG BLuray Burner, BM IP4K, WIN10 Pro, Shuttle Pro2

          ASUS PB328 monitor, BenQ BL2711U 4K preview monitor, EDIUS X, 9.5 WG, Vegas 18, Resolve Studio 18


          Cameras: GH5S, GH6, FDR-AX100, FDR-AX53, DJI OSMO Pocket, Atomos Ninja V x 2

          Comment


          • #6
            Not sure how I would tell whether it's the camera or recorder that's drifting.
            Drift likely caused by slightly different clock frequency due to tolerance of the crystal oscillators involved. If you have an accurate frequency meter you could measure them, or perhaps compare time codes referenced to a known accurate source, such as the bureau of standards broadcast time signals..

            As a point of reference the accuracy for a good crystal is typically on the order of +/- 10 PPM, and is subject to some aging and temperature effects as well. 10 ppm is roughly 1 frame per hour, so a 1 frame per hour difference in clocks is not surprising. That is why studios use genlock to provide a common clock for all gear.

            A 1 frame difference in audio is equivalent to about 33 feet (for NTSC) in camera position relative to the talent.

            Comment


            • #7
              Originally posted by tosok View Post
              Thanks for the input. Really appreciate it.

              BernH: Not sure how I would tell whether it's the camera or recorder that's drifting. Any ideas on that? I bring the Tascam audio into a DAW for prep before importing it in Edius. My preference is to use a dedicated audio recorder which gives me a limiter, more visible level control, etc.

              Ron Evans: For clarity, are you saying you've seen this drift with the cameras? Also, do you happen to remember what remote you were using for your Sonys? I'm not having much success locating one that doesn't require a IR cable to the camera... which would prevent starting all three at the same time.

              Hope I don't need to invest much more to get the audio tightened up. I have a bunch of different audio tools. I'll also check and see if any of them can squeeze/stretch audio without pitch issues.
              To answer the question about drift the easiest solution is to look at the sync long enough into a recording where you have seen drift, lets say about 10 minutes in. If the audio on the camera that is taking the feed from the tascam is in sync, but you then mute the camera audio and listen to the tascam audio and it is out of sync that is the culprit.

              As I said, the best solution is to record audio to the camera via the beachtek adaptors so it is always in sync with the video. You may see a slight timing difference between the lav and the room mic depending on placement distance due to the speed of sound reaching the 2 mics compared to the speed of light reaching the camera, but this shouldn't be a lot unless you room mic is a long distance away from the lav mic.

              Using the beachtek box is a pretty small expense to solve your problem. The Pro one is $149 USD and the Pro+ is $229 USD. You only need one of them to put on your main camera.
              Last edited by BernH; 05-13-2021, 09:11 PM.
              Edius WG 9.55.7761, various 3rd party plugins, VisTitle 2.9.6.0, Win 7 Ultimate SP1, i7-4790K @ 4GHz with HD4600 GPU embedded, MSI Z97 Gaming 7 Motherboard, 32GB Kingston HyperX RAM, nVidia GTX680 4GB GPU, Matrox MX02 Mini MAX, Corsair 750W PSU, Corsair H110i GT Water Cooler, Corsair C70 case, 8TB Internal RAID 0/stripe (2x4TB Seagate SATAIII HDD's, Win7 Software stripe), 1TB Crucial MX500 SSD, Pioneer BDR-207D, Dual 1920x1080 monitors (one on GTX680 and one on Intel HD4600).

              Comment


              • #8
                OK. I'll have to consider the Beachtek solution. Couple of more things here... dealing with I've already recorded.

                1. My audio software gives me the option to render a file in 48k Pulldown... which is actually 47952k. Since I'm shooting in 29.97 and the Tascam is recording in 48k, does it make sense to bring audio saved at 47952k into the timeline rather than the 48k version? I understand that this will not entirely fix drift issues, but will that compensate for some of the drift over the long recordings? I'm not clear in my mind if it's better to use the pulldown sample rate to match the video frame rate... especially since the camera itself records audio at 48k.
                2. Is my best solution with what I have already recorded just to cut the audio track at points where the drift is noticeable and advance/delay that audio clip with the audio offset feature? It seems like a simple enough solution, but want to run it past you who understand the issue.

                Thanks!
                Mobile setup: Edius X - Laptop running i7-10875H CPU @ 2.30GHz 2.30 GHz (8 cores/16 threads) with 64GB RAM and NVIDIA GeForce RTX 2080 Super with Max-Q Design w/8GB VRAM.
                Desktop: Edius X - AMD 5950x installed in Asus ProArt Creator B550. 64GB Ram, 2 x 1TB m.2, 6TB HHD. EVGA 3080ti w/12GB vram. Running Windows 10 Pro

                Comment


                • #9
                  Originally posted by tosok View Post
                  OK. I'll have to consider the Beachtek solution. Couple of more things here... dealing with I've already recorded.

                  1. My audio software gives me the option to render a file in 48k Pulldown... which is actually 47952k. Since I'm shooting in 29.97 and the Tascam is recording in 48k, does it make sense to bring audio saved at 47952k into the timeline rather than the 48k version? I understand that this will not entirely fix drift issues, but will that compensate for some of the drift over the long recordings? I'm not clear in my mind if it's better to use the pulldown sample rate to match the video frame rate... especially since the camera itself records audio at 48k.
                  2. Is my best solution with what I have already recorded just to cut the audio track at points where the drift is noticeable and advance/delay that audio clip with the audio offset feature? It seems like a simple enough solution, but want to run it past you who understand the issue.

                  Thanks!
                  The pulldown solution is designed to implement a 0.1% overall speed change to make something that was recorded at 30fps or 24fps match 29.97fps or 23.976fps (30fps to 29.97fps is 0.1% and 24fps to 23.976 is 0.1%). There are larger standard pulldown/pullup amounts in the 4% when going 23.976 or 24 to or from PAL. If your drift as about that amount it may pull things closer, but without knowing which direction or how much you are drifting, I can't really say if it will work. Here is a fairly descent blog post on audio pulldown: http://javierzumer.com/blog/2019/4/2...io-pull-updown

                  I would tend to suspect that if you can't determine how much the drift is and consequently if the pulldown/pullup will work, you would be better off cutting the audio track periodically and resyncing. You should try to do this in places where there is silence do that you are not cutting off words and you should apply a very short audio cross fade at those points to avoid any clicks or pops that a hard cut may introduce.
                  Edius WG 9.55.7761, various 3rd party plugins, VisTitle 2.9.6.0, Win 7 Ultimate SP1, i7-4790K @ 4GHz with HD4600 GPU embedded, MSI Z97 Gaming 7 Motherboard, 32GB Kingston HyperX RAM, nVidia GTX680 4GB GPU, Matrox MX02 Mini MAX, Corsair 750W PSU, Corsair H110i GT Water Cooler, Corsair C70 case, 8TB Internal RAID 0/stripe (2x4TB Seagate SATAIII HDD's, Win7 Software stripe), 1TB Crucial MX500 SSD, Pioneer BDR-207D, Dual 1920x1080 monitors (one on GTX680 and one on Intel HD4600).

                  Comment


                  • #10
                    I think you will find that the drift of the Tascam is not a fixed amount but varies over time. So it is not a straight offset in time like a mic distance would be between mics at different distance from source. You are still better recording the audio with the main video then you have no issues. Personally I would not use the Tascam at all. Just feed the cameras. The cameras I find have reasonable clocks and should sync together fine for video and just use audio from the main camera. In this case the worse you will have is an offset between the tracks for left and right which even EDIUS can fix.
                    Ron Evans

                    Threadripper 1920 stock clock 3.7, Gigabyte Designare X399 MB, 32G G.Skill 3200CL14, 500G M.2 NVME OS, 500G EVO 850 temp. 1T EVO 850 render, 6T Source, 2 x 1T NVME, MSI 1080Ti 11G , EVGA 850 G2, LG BLuray Burner, BM IP4K, WIN10 Pro, Shuttle Pro2

                    ASUS PB328 monitor, BenQ BL2711U 4K preview monitor, EDIUS X, 9.5 WG, Vegas 18, Resolve Studio 18


                    Cameras: GH5S, GH6, FDR-AX100, FDR-AX53, DJI OSMO Pocket, Atomos Ninja V x 2

                    Comment


                    • #11
                      Helpful info. Thanks so much for your input. Ron, I'm interested in what you discover for wireless remotes that work with the AX53. Even just starting them together - if that's all it could do - would meet the pressing need.
                      Mobile setup: Edius X - Laptop running i7-10875H CPU @ 2.30GHz 2.30 GHz (8 cores/16 threads) with 64GB RAM and NVIDIA GeForce RTX 2080 Super with Max-Q Design w/8GB VRAM.
                      Desktop: Edius X - AMD 5950x installed in Asus ProArt Creator B550. 64GB Ram, 2 x 1TB m.2, 6TB HHD. EVGA 3080ti w/12GB vram. Running Windows 10 Pro

                      Comment


                      • #12
                        At an hour in it sounds like they are about a frame or more apart.
                        One frame per hour, That is one part in 108,000 (for NTSC) or about
                        10 parts per million, 0.001% - inline with the accuracy of a crystal, especially considering the recorders/camcorders are from different manufacturers. Thus You may hear the phasing change if you are mixing long runs of sound of the same performance from the different recorders.

                        As a point of reference common 13.5 MHz crystals are offered with tolerances as good as +/- 10 PPM and similar thermal stability specs.

                        When mixing audio sourced from different recorders, and if precise phasing is an issue, I would use an audio workstation with resolution and the ability to edit essentially down to the individual sample level. Video editing typically gives resolution down to the frame level which apparently does not meet your needs for audio. Trimming/inserting samples during quiet sections and near zero crossings is a way to get precise matches in audio. I've done it to match audio from one live performance with video from another.
                        Last edited by dpalomaki; 05-14-2021, 02:29 AM.

                        Comment


                        • #13
                          I have tried all the remotes I have and it is only the RMT 835 that only starts recording and not anything else! Will continue to look and see if there are any older remotes around that I have not found so far.
                          Ron Evans

                          Threadripper 1920 stock clock 3.7, Gigabyte Designare X399 MB, 32G G.Skill 3200CL14, 500G M.2 NVME OS, 500G EVO 850 temp. 1T EVO 850 render, 6T Source, 2 x 1T NVME, MSI 1080Ti 11G , EVGA 850 G2, LG BLuray Burner, BM IP4K, WIN10 Pro, Shuttle Pro2

                          ASUS PB328 monitor, BenQ BL2711U 4K preview monitor, EDIUS X, 9.5 WG, Vegas 18, Resolve Studio 18


                          Cameras: GH5S, GH6, FDR-AX100, FDR-AX53, DJI OSMO Pocket, Atomos Ninja V x 2

                          Comment

                          Working...
                          X