[Dave Haynie]

The other really sort of jumpy pan you can run into, which I didn’t mention, is due to an interlacing inversion. You’ve shot and rendered progressive, so this shouldn’t be an issue.

But basically, what happens in interlaced video is of course that one field is temporally ahead of the other. If this gets inverted, what you see is video with unwatchably horrible judder in any motion, as you’re jumping backwards and forwards in time constantly.

Should be impossible rendering from 24p, regardless of the target format, but I just figured I’d mention it.

-Dave

[Dave Haynie]

Sounds like you’re doing the right thing, pretty much.

For the “jumpy” pans, check that’s not the original video itself. At 24p, a fast pan can simply be jumpy. One thing you can do to minimize this is follow the typical cinematographer rules for shooting, such as using the right shutter speed. A typical film is shot with a 180 degree shutter, which would be 1/48th second for 24p. If you shoot with a much faster shutter speed, you’ll eliminate the natural motion blur of a pan. So you’ll have nice, static frames (in theory… read below), which “jump” one to the next.

That might not happen, but not for a good reason, but due to the AVC compression on the 60D. Interframe compression CODECs like AVC and MPEG don’t deal well with any situation that has too much information changing frame to frame too quickly. That’s precisely what a pan does, and the faster the pan, the worse it’s going to look. If you need a fast pan, the best you can do it shoot it at a faster rate, 60p maybe, then add some motion blur when you downconvert to 24p. You might try applying

Having studied a little of how professional mastering is done on DVDs, I recommend looking at any similar section of a DVD or BD, frame by frame. A mastering engineer can tweak individual GOPs when creating a DVD, changing frame rate and tweaking the video on the fly. What they typically do for a high motion scene, to avoid macrocell blocking, is to bump up the encoding rate, but also add a low-pass filter (eg, a blur) to minimize the high frequency information in the whole picture, thus lowering the macrocell to macrocell changes even before the MPEG encoding is done. You can do a little of this in Vegas by pre-blurring segments with high motion in them, but that’s not going to help if you’re already.

You can try a motion blur envelope on the jumpy pan, see if that helps. I use Boris Continuum Complete, which has a number of additional motion blurs, but the built-in is enough to make some progress. If that’s not enough, try a tiny bit of panning. Basically, use pan/zoom to zoom in just a bit on your “jumpy” pan. Set the start and end keyframes of your pan to have the pan move in the direction of the original motion. Adding a supersampling envelope may make this look better… you can mess around with this stuff, see what works in your particular case.

-Dave

[Dave Haynie]

As a general rule, you should NEVER use the in-camera mic if you’re trying to get good sound. For weddings, I use multiple off-camera recorders. The lapel mic on the groom is what you want. I usually have at least one shotgun on the ceremony as well, and if the setup permits, I put a digital recorder on the podium, in a plant, etc… something near the ceremony, but in a different direction. Also good to have at least one mic on the PA, or a digital recorder off the PA mixer if possible.

I don’t like the idea of switching… I want all of these things recorded at once. Back in the day, I used MD recorders, these days, I have a couple of digital recorders; one is small enough to clip on to the groom or anyone else I want to mic directly — no worries about cabling or radio. His lapel is often good for micing the bride, too, but not always.

Part of this is maximizing quality. If I have 3-6 different audio positions, I will have better audio for every speaker: bride, groom, cleric, invited guests, singers, anyone who speaks at the ceremony. I started out as a sound guy — I’d really like a lapel mic on every speaker, but that’s not remotely practical.

Part of this is redundancy. I was shooting a wedding back in 2006 in which the PA system failed entirely. I did have a mic on the PA, but that was completely worthless… everything I recorded came in via shotgun mics on my two cameras. I made it work, but it wasn’t the best audio I could have managed.

-Dave

[Dave Haynie]

Derivatives of Mel-Frequency Cepstral Coefficients, or just the FFT, with cross-correlation? I had to solve a somewhat similar problem a few years back (software defined radio implementation of NTSC television, naturally had to run in realtime, had to reliably find the H/V syncs, also the color clock… FFT on that made it easy to find the phase offset). I had one decoder running on a single Intel Core 2 core, an MPEG-4 encoder on a second core… eight cores per rack. Kind of cool, and yet, a stupid way to avoid using a $2.50 chip 🙂 Using the SDR wasn’t my idea….

-Dave

[Dave Haynie]

First of all, assuming that the task manager is counting both threads per CPU in its formulation for full CPU percentage, you’ll probably never get to 100%, at least as if it’s being calculated the same way you would for a real 8-core machine. The idea of the hyperthreading system is to fill in pipeline delays (eg, read/writes to/from cache or memory) by swapping the CPU register set to the alternate thread. That’s going to utilize the CPU’s basic hardware better than a straight four-core chip, usually (basically, as long as you don’t start thrashing on-chip cache), but it’s still not going to match that of an 8-core. So that may be your first bottleneck.

Disc speed matters more, obviously, as your CPU increases in performance. You’re probably fine with a single file in and out to the same drive, but in general, writing back to the primary media source drive will case increased seeking — and thus, very decreased HDD performance. You can test this by writing the output to a different drive… the C: drive is fine (you have no seek issues with an SSD, though of course, you’d like to minimize writing to them, they don’t have as many write cycles as a magnetic HDD), or any external drive. At least, as long as you’re not rendering any crazy formats, something MPEG-ish is good.

And I hope you didn’t mean actual RAW, eg, uncompressed video when you say “Disk 2: Raw video”… Raw HD video is going to make most any single HDD a bottleneck.

-Dave

[Dave Haynie]

Assuming they’re using a real quartz crystal, and not some cheap-ass ceramic resonator (which is more like 0.1% to 1% accurate), the run-of-the-mill super cheap ($0.50) PC-class crystals have an accuracy of +/- 100ppm or better, and usually with only marginal temperature compensation at best. So that’s worst-case 0.01% accurate, or at 48KHz, that’s up to 17,280 samples per hour.. or a drift of about 1/3 second. Of course, if you have two of these running at the same time, you could actually be 2/3 of a second apart at the end of an hour (16 frames at 24fps, 40 frames at 60fps).

So in short, with cheap crystals, your WooWave is going to have potential trouble with clips over 1/2 hour. HOPEFULLY no one’s using cheap crystals like these. The super cheap ($0.50) 32kHz crystals used in PCs and other consumery devices for the TOD clock are usually around 30ppm.

I found a video of the Zoom H4 (older model) drifting 6-7 frames versus a Canon 7D over a ten minute shot… even given a frame rate of 30fps, that’s actually three times worse than my 100ppm worst-case. But the H4 and earlier Zoom recorders are known to have awful crystal stability, about 0.05%, or 500ppm — possibly even worse. Not sure about the 7D. Rumor has it the H4n has a clock accuracy of 50ppm; Zoom/Samson themselves say it’s accurate for shorter clips, and suggest software handles the issues in longer clips. Of course, they’re poised as the go-to field recorder for DSLR videographers, most of whom use Canon, and Canon’s still inflicting the EU’s issue with camcorder taxes on all of their DSLR users… so you can’t record more than 29’59” continuous on any Canon DSLR (many others go longer, but they’re a minority in the business). That would mean about 4 frames, worst case, at 24fps, with an equally reliable timebase in the camera.

I looked up the Marantz recorder John’s using, and several new ones, but there’s no indication anywhere of the timebase accuracy… so much for “professional”. Though the higher end unit, the PDM-671, has some kind of add-on for external sync. And it fixes a known problem with the pre-amps in the PDM-660, not sure what that is. I’m assuming the audio’s clear enough to be a good sync source.

For some of the digital radio stuff I’ve done over the few years, I use a 1.5ppm crystal as a timebase for the frequency synthesizer. This is a TCXO (Temperature Compensated Xtal Oscillator), too, so it’s pretty good over temperature. That’s going to mean 260 samples per hour, or a maximum device to device drift of 520 samples per hour… 10.8 microseconds. Not a practical problem in the least.. even good musicians can only hear about a millisecond of timing drift. But I had to pay a whole $15 each for those crystals (well, my company did).

-Dave

[Dave Haynie]

[Steve Rhoden] “Well abandon or not Dave, There are always tons of options.”

There are… just not ideal options. I’d love something like DVD-Lab for Blu-ray. I’m certainly not planning to drop a few grand on a high-end application just to author better looking BDs.

[Steve Rhoden] “If Apple could abandon such a powerful and world renowned Compositor such as Shake in its prime and an industrial
standard at the time….Then what says DVD Architect…lol”

Apple’s a special case. They’re clearly motivated toward dropping their high-end… that’s the end of Shake, Compositor, DVD Studio Pro, etc… the ancient Mac Pro, the end of the server and storage server, etc. The Appification of Final Cut Pro, without regard to professional users (no project import, etc). That’s what happens when the entire Mac world is only 15% of Apple’s business, and nearly all their growth is from consumers.

Apple still thinks they’re a hardware company. The acquisition of all that media content creation software was a smart strategic move back in the days of a 1.5% Mac share of the PC business, mostly media content pros. They needed that to shore up the platform against other suppliers dropping it, and they succeeded beyond their wildest dreams. But little of that’s important to Apple now, apparently, because iOS has done so much better. There are even rumors (I’m very skeptical, but they persist) of Apple kicking x86 out of the Mac in a few years and replacing it with a higher-end ARM of their own creation (in fact, I posted on a few EE Tech sites just how crazy this sounds.. but if you restrict “Mac” to basically just iMac and laptop users.. increase the performance of their A6-core processor (don’t know if they have a name for it yet) by 2.0x-2.5x, put in 4-8 cores, and you’re well into i5 territory, if not better.

In fact, that’s kind of a strategy. They can’t be Intel, building a bunch of different CPU microarchitectures to fill out a full PC line of product. Even AMD has trouble doing that, with a much larger volume than Apple. But take the ARM approach of modular systems, and they could build mobile and desktop processors with the same basic tech. They might also find that ARM on a power supply can run really fast.. that’s not something that’s been done, given the emphasis on mobility. But one of Apple’s two IC company acquisitions is Intrinsity, a company that’s basic mission has been making stuff go faster (pre-Apple, they worked with Samsung on the “Hummingbird” core processor, taking ARM’s 600-ish-MHz A8 and making it go twice as fast — that’s also the CPU core in the A4 SOC from Apple).

-Dave

[Dave Haynie]

2GB per core is a general rule of thumb, not really specific to Vegas. Of course, other things can influence this… I went from 8GB for my six core system to 16GB in 2011 because I absolutely needed it. Something about editing 20-50 shot panoramas from 18Mpixel RAW photos. Never needed more than 8GB for video.

-Dave

[Dave Haynie]

[John Rofrano] “The bigger problem is that your processor is woefully under powered. You have to understand that laptop processors come in classes and the T4500 is at the bottom of the list. Above it is the T5000 series, T6000 series, T7000 series, T8000 series, and the T9000 series. I wouldn’t attempt to edit HD video on anything less that a T8000 series.”

Absolutely agree. When you look at rendering, there are a bunch of things basically in a pipeline… CPU, FPU, GPU (if you have a good one), main memory, storage (HDD, etc). It’s not that difficult to have enough RAM, enough HDDs, etc. to make the CPU subsystem the weakest link in the pipeline.

The things I suggested are what you’d need to look at that might be weaker links in that pipeline. Certainly you have too little RAM, and when you consider that RAM is usually at least 1000x faster than a hard drive, if you’re paging during a render, you’re going to make things run very, very slow compared to what you might achieve. You can pretty much tell just by popping up the Task Manager and looking at your CPU and memory use. If you see less than 90-95% CPU in use during a render, or see your total memory nearing its limit (heavy use of paged, aka, virtual memory), you’re being slowed by things more easily controlled.

But you have a weak CPU, no two ways around that… the Pentium T4500 was a low-end CPU in 2009. Looking it up, I find a CPU Mark (https://www.cpubenchmark.net) average of 1402. The CPU in my desktop, not currenty by any stretch, is an AMD 1090T six core, rated at 5618. It would be reasonable to expect my computer to run at least 4x faster… I’ll bet in practice it’s more like 8x-10x, since newer processors have media-specific optimizations not always brought out in general benchmarks. And the latest Intel i7s (no, I can’t afford one either) averages out 14,763… over 10x the performance on this benchmark suite (so expect media performance to be even greater).
https://www.cpubenchmark.net/mid_range_cpus.html

The CPU in the fairly low-end laptop my son Sean’s getting for Christmas (shhh…), an AMD A8-4500, does about 2651 on this suite. And that’s in a $400 laptop (shhh… and before you judge, he got a custom-made-by-me i7-based desktop for his birthday). You’ll usually get much faster performance from a desktop PC for the money… in both cases, you have to shop around.

-Dave

[Dave Haynie]

I do this all the time. I generally use one digital audio recorder as the effective timebase, since it will be run through the entire shoot, while I may be starting and stopping cameras. No obvious reason it shouldn’t work the other way.

The audio/video sync shouldn’t be any worse than two camcorders, unless you have a poor audio device (early Zooms were good for less than 20 minutes, typically, but they fixed that in the H4n and other recent units). One hour might get you a slight de-sync, but I’d deal with that in a second pass.

Of course, make sure your video segments are split by time on capture. A single clip that’s got a time discontinuity will be unsyncable. I haven’t used the “events in order” switch. Not sure what else could go wrong, it has always pretty much just worked for me.

-Dave