Sunday, September 1, 2013

Filming a concert with multiple Panasonic m4/3 bodies and a pair of TASCAM personal audio recorders …

I’ve mentioned throughout the summer in various places (in the blog, on the forums, to anyone who would listen :-) that my son and his friends were planning to hold a concert for their friends and families late in the summer. Well, last Tuesday was the day and things went rather smoothly.

They rented a small venue called Club Saw, which is in the basement of the Saw Gallery, located near the heart of downtown Ottawa. Costs were quite reasonable and the venue provided both security and someone to run the soundboard and lights. They charged a very small entrance fee of four dollars to help cover costs, and the turnout was very strong. However, a bill is still expected for the soundman and the security so it is likely that there will be a slight cost overrun. Still, the venue was excellent and costs were quite reasonable.

Throughout the summer, I have been pondering exactly how to film a small concert. I knew that it would be necessary to run the cameras from a fixed position and in a single take. Else, I would be running around back and forth in front of the audience whenever there was an issue with batteries or changes to the set up or anything like that. So I ordered complete AC adapters for all four cameras from a source on eBay. They were very reasonable, the entire set costing me perhaps $50. And they worked perfectly.

The other necessity, especially considering that I had no idea how long they play for, was to get cards large enough to store the complete concert. Since I hacked all of the bodies with the exception of the G5, which is not hack-able at this time, I knew that my bit rates would be between 30 and 60 Mbps. This meant that, should I need to run the cameras for as much as two hours to cover everything I needed to, a 32 GB card would be touch and go. So I decided to purchase four 64 GB cards. I looked for a long time on various websites, and almost settled on a set of Patriot EP cards, which are reputed to be very fast. However, the new transcend cards can be had in 600X flavor, which is blazingly fast. but in the end, these were all just too expensive. What I ended up going with was the Polaroid PNY SDXC cards. these were showing as being quite fast, however my read speeds tend to average around 20 MB per second, which is pretty weak. But I had tested an old Class 6 Transcend card, and I knew that even 60 Mbps would write quite well on cards of that speed. So the somewhat lower speed of these cards Reedy was not much of a concern to me. And in fact I was right, since they ran perfectly through the entire concert with no glitches that I can detect.

Of course, I was also freaking out with respect to tripods and other accessories. I ended up over the last two months ordering for small tripods, all very inexpensive, to see which ones might be useful to hold up a camera for an hour or two. Of the four I bought, two were excellent. One made by SLIK and one made by CASON. The others are also acceptable, but I did not have to use them in the end. I had already replaced my large carbon fiber tripod with a much smaller and more pleasant to use carbon fiber tripod from Henry’s Canada. This turned out to be very useful to hold up the GH2 on the tall desk where the sound board was located. This affords a view over the audience and allows a very clean frontal view. I had, in fact, spent a great deal of time pondering where the camera should be.

I had settled on one camera with a long lens at the soundboard for the frontal view, one camera stage left, and one camera stage right. It turned out to almost work out that way but at the last minute the fellas moved the guitar chair and microphone set up to be parallel to the keyboard set up, thus ruining my stage left camera view. So I quickly pulled that up (I taped all the legs down using gaff tape) and move it closer to the guitar location. I zoomed it in a bit, thinking that I get a nice close view of the guitar and that is exactly what happened. However, it was a bit closer view than I had anticipated and Nick set a bit closer than I expected. So I really did not get a good view of his left hand on the neck. But the view of Dylan’s guitar playing is absolutely perfect.

I reserve the one camera to hold in my hands, which was the G5, but in the end none of the footage was very good. I had started with the 14-140 zoom lens, thinking that I would be able to zoom out or in at will, but I immediately noticed that the footage was very dark. So after a while I got very nervous and went to the back of the room, grabbed the 45 mm 1.8 Olympus lens and stood there filming over the audience. I absolutely loathe that footage. Lesson learned. And the slap in the face is that the footage I shot at the beginning was actually quite good. With a bit of grading it would have been just fine. So now I know that even slow lenses can do a decent job when the artists are lit.

Note that I shot only three images of the venue during the setup. The images are pretty crappy as I have very little time but at least you get a sense of the size of the venue and the amount of room I had to move around (as in not very much at all.)

The lighting was not particularly strong in this place, but there was more than enough light for the stage right camera, which was the GF three with the Sigma 19 mm 2.8. I experimented during rehearsals with having this camera in ape has a PASM mode, but in the end I felt I needed the high ISO mode so changed to scene mode and selected night portrait. This enables high ISO on Panasonic bodies. However, motion blur is not as nice because scene modes do not allow me to fix the shutter speed. This is most unfortunate, as all Panasonic needs to do to make these work really well is allow fixing of the shutter speed when in high ISO mode.

The guitar camera (shown here in its original stage left position) was the GX1 with the Panasonic 14-42 X VARIO PZ pancake zoom lens. I believe that I left this camera in a PASM mode and set the correct 180° shutter at 1/60s. The 180° shutter is very important to maintain believable and comfortable motion blur. And since this camera was focused on hands moving quickly over strings, that matters.

It is actually quite noticeable on the GF3 camera that the motion blur is to strong. But the point of that view is to be wider, so the lower quality of motion blur is not all that big a deal. However, the GF3 is known to be a very good video camera for close-ups but a rather poor video camera for longer shots. And in this case, the Sigma lens was a 38 mm effective focal length, which is wide enough to challenge the GF3.

And since I set the lens in manual focus mode and focused on the microphone, I causes the faces and hair to be very slightly blurred, further challenging the video acuity. This forced me to apply a Magic Bullet Quick Look that darkened the surroundings and applied a slight soft focus effect in order to bring all three cameras to look similar. I really like this look, but I would have liked to be able to show more of the venue. Of course, the venue is fairly drab and there were cords of odd colors all over the place, so it is just as well that I effectively sunk the background into black.

I tried the 14-140 Panasonic zoom lens on the GH2 at the rear of the venue. But I found that the ISO had to be fairly high in order to compensate for the slow lens. I check the focal length that I was using and realized that I could use a faster lens at 50 mm and achieve the same framing. I had had the foresight to bring the Pentax 50 mm 1.4 manual focus lens with me and this lens proved to record beautiful video at low ISO. However, I managed to not quite achieve critical focus on the stage so I was again very happy to apply the quick look.

One thing to note is that all of the cameras were unmanned, which means that their fields of view had to be wide enough to accommodate movement of the singers. What this led to, was a cropping of the video from the frontal view and the stage right view, leading to further edge degradation which is unfortunately quite visible in the final version of the video. However, I still feel that the video is quite watchable and so I consider it an acceptable compromise to achieve a better and more interesting set of shots rather than going for high acuity at the expense of a deeply boring visual experience.

Now, how about audio? Well, at some point in the summer I purchased a second TASCAM DR-05 personal digital recorder so that I would have at least two recorders running. I had learned from recording twice before in pubs that one position just was not enough to capture everything. In addition, I harbored the notion that I would get to tap into the soundboard and get a nice clean signal from the microphones. But this meant that I would not have a nice clean signal from the audience. So the second recorder was necessary.

And stunningly, it turned out to work exactly that way. After I had wandered about the venue in the hour before the show, taping down tripods and setting up AC adapters in getting the cameras running and so on, I went to the soundboard and asked if I could tap in. He asked what sort of chords I had, and I told him that I had brought RCA, 1/8 inch, and 1/4 inch. He said immediately that the 1/4 inch would work in the headphone jack for the soundboard. This made sense to me, since I wanted a complete mix of all of the guitars and microphones and of the two keyboards on stage.

I attached the TASCAM and got it running during rehearsals. I was able to note that the keyboards were very loud so I dropped the record levels to something around 15 or 20%. This is extremely low and would normally be a serious concern for his, however I was recording in wave at 24 bits and 48K sampling rate. I regret not going to 96K, but Sony Movie Studio simply does not support that level of audio. However, I do plan to upgrade to Sony Vegas Pro at some point and so I really should have set the recorders to 96K. Oh well, c’est la vie.

The point though, is that these settings have a naturally low noise floor according to what I have read. And I have to say that it really worked out to being the quietest recordings I have ever heard. By quietest I mean lowest noise, the signal is simply beautiful. After setting up the soundboard recorder, I walked to the front and attached the second recorder to the microphone stand facing the audience. I had mounted both recorders on small tripods, and the audience recorder was simply taped to the microphone stand using gaff tape around the pole and the tripod legs squished together. It worked perfectly. Unfortunately, it turned out that the keyboard on Chris’s first song is still too loud for the soundboard recorder and pretty much blew it out completely. So what you will notice on his first song is that the room opens up to create an enormous soundstage with a great deal of echo. This is because the only audio channels you are hearing are the audience might and the GH2, which is a decidedly challenging set up. I think it worked acceptably, but the audio improves markedly in his second and third songs simply because the soundboard channel was again available.

Aligning audio tracks and video tracks can be extremely painful. Crystals are not all made equal, and professional audio people know enough to sink them up with a constant time signal. But Mr. Hacker (a.k.a. me) has access to nothing like that. So I was faced with six separate recordings to be aligned.

Under normal circumstances, a sane person would simply purchase a professional editing tool like Sony Vegas Pro and the widely acclaimed plug-in called PluralEyes. This plug-in is used by professionals to perform semiautomatic to almost completely automatic alignment of many tracks, including automated compensation for timing drift. This is an absolute miracle timesaver. But alas, I cannot justify around $500 to make my life a little bit easier. So I did it the hard way in movie studio, aligning based on waveforms and cutting and shifting or stretching where necessary to account for drift. It is no worse than six simultaneous root canals so people should not whine about it. :-)

Luckily, it turned out that the two TASCAMs have identical timing crystals in them, leaving me with an absolutely trivial task to align the two primary audio tracks together without having to account for any drift at all. Of course, I was not so lucky with the three Panasonic bodies. Each of these had some drift and at least one of them had drift that seemed to get worse over time. As mentioned earlier, I used the hack and shift method and it turned out ok after a terrible first draft video. I would be surprised if anyone can detect where it goes slightly off in the final video, and I am not telling.

So the shoot is over and I have torn down the cameras in the recorders and pack it all away and now it is time to dump the cards to disk and see what I have. Well, it turns out that I ended up with 70 GB of material, and that took a couple of hours to copy to disk (this is why people purchase significantly faster cards, like those gorgeous Transcend 600X cards. Sigh.)

When I had the two audio tracks aligned, I locked them together. I was still able to adjust relative volume using envelopes, and I was still able to apply differing sound effects to each track. This allowed me to boost various frequencies when I thought it would work better.

Since I had three cameras with complete soundtracks, it made sense to bring them in one by one and align them in time. I started with the frontal view from the GH2 and found that it had significant drift. Once I had the beginning align, I tried selecting all of the individual events, and stretching the whole thing as a unit. This seemed to work, but later I trimmed off the excess at each end and found that it went completely out of alignment. My initial video shows this very clearly and frankly is a total mess. The audio is very dry and there is a huge sibilance in the version uploaded to YouTube. YouTube seems to aggravate any built-in sibilance, so you really need to be careful to not allow very much in. Another lesson learned.

For your interest, here is the first draft of the movie.

I actually created it because the boys wanted to have a version available so they could party and watch it this weekend. Hopefully, they will be watching the final version and not the draft because the final has squeezed out the timing errors and has three video angles with transitions in all the right places (mostly, anyway.)

In a project this large, the editing process is progressive. As I figure out what I want to do, I change the envelopes to expose the camera angle I would like, and ensure that the cut is either on a beat or just as the mouse is shifting to the microphone to begin singing. This seems to give a nice pacing that I find pleases the eye. In other words, I find that my eyes expect either no shift, or a shift just before something significant happens. Please let me know in the comments if you think I was wrong about that.

One error that I made was to get the frontal view not quite matching the side view. The frontal view has much stronger saturation for some reason and I will eventually have to go back and fix that for a third version. But as I have said, I quite like this version and find that I can enjoy watching it surprisingly often. I have listened to it on my computer with decent Sony headphones probably 10 times and to end. But hearing it on my 5.1 Paradigm Titan system on the big TV really brought home how nice it sounds most the time. But another lesson learned is that I have the base a little too high and I added a bit too much reverb to open the space. This very slightly D focuses the voices, which affects Nick quite a bit because he allowed the microphone to drift downward for his last three songs, thus putting significant competition into his songs from the instruments. He also sounds unfocused, because of the distance to the microphone. This is of course yet one more lesson learned.

So in conclusion, I had an absolute blast. Although I had absolutely nothing to do with the performances themselves, I did enjoy the pressure of trying to get decent video and audio tracks and I thoroughly enjoy manipulating raw material that I have created to get something out of it that is pleasant to watch and listen to. There is more work to do to tweak this into even better sound, and to make the video a little more consistent. There is also work to do to carve up the songs now into individual tracks for YouTube. But all in all, I am very satisfied with this final draft. Enjoy.