The results are finally in and EOSHD.com has found settings that give you the best possible quality from your GH1. There will also be improvements and updates still to come. Please read on!
The GH1 can now do 50Mbit 1080p MJPEG and the quality is astounding. It isn’t quite 100% robust yet, certain types of shot can halt recording. But it’s very usable as you can see from the video above.
At the temple shoot today, I was able to complete 70% of the shots using 50Mbit MJPEG 1080p,and fell back on 32Mbit AVCHD for the rest. The video above contains no footage from the out-of-the-box GH1. It’s all a mixture of 50Mbit MJPEG 1080/30p and the new 32Mbit AVCHD 1080p native 24p MTS files.
I have not yet had time to perfect my settings for reliability, only for sheer image quality. So this is where you come in. Try out the settings yourself, have fun shooting, and keep testing small adjustments in a systematic way to discover what works and what doesn’t.
The quickest way to see the bitrate of a MJPEG clip you’ve recorded is to open it in Quicktime and turn on Movie Inspector. My clips vary between 28MBit and 50Mbit. MJPEG is recorded in variable bitrate format on the GH1. Any higher than 52Mbit and the camera cuts off recording half way through a clip. We need to find a way to limit the 50Mbit variable bitrate MJPEG to a maximum of 50Mbit, no matter how much details is packed into the shot and how sharp the lens is. I am confident this can be done and I believe Tester13 is getting closer to enabling a true 24p framerate for MJPEG mode.
In Quicktime, I can see that the camera has recorded clips that hover around the 45 or 50Mbit mark just fine but a few clips which stop after 1 or 2 seconds are hitting the 70Mbit point!! This is too high for 1080p with this kind of dynamic range and detail from the GH1′s amazing sensor. The camera’s buffer and card IO controller just isn’t built to support this kind of data throughput.
The working 50Mbit footage is cleaner than even the 5D Mark II or 7D. The 5D Mark II records at 35Mbit per second. That’s 4.8MB/sec to the card, to get that 30MB/sec claim by Sandisk in perspective. I tested my Sandisk 30MB/sec edition. It does about 20MB/sec in real usage – i.e, copying a large file from a PC to the card. I’ve had clips off the GH1 which have been written to the card by the camera at over 8MB/sec, almost double the 5D Mark II’s rate.
For certain shots , usually always deep depth of field stuff with the lens stopped down, I got a card write error and had to fall back on the new 32Mbit AVCHD mode by Tester13 to get the shot.
I was able to use the Vari ND to get a shallower depth of field, and an image that isn’t quite as sharp. This helped limit bitrate so that it didn’t overflow the camera’s buffer. I used a class 6 as well as the class 10 Sandisk. You don’t actually need a super expensive Class 10 SDHC card. The card is not the bottleneck, even for 50bit footage – the camera’s buffer is. So we need to find that limiter setting.
Another plus is that the GH1 is doing a better job of scaling and processing data from the sensor than the 5D Mark II because it isn’t line-skipping when scanning the sensor. Instead it bins individual pixels – just like when you downsize an image in Photoshop. It has a lower resolution sensor by half, so it has CPU time left over to do more work on scaling and encoding. At these high bit-rates, it’s footage is beginning to look like what we were expecting from 2nd generation DSLRs that aren’t yet released. It will also give the AF100 a run for it’s money. Panasonic must not be too pleased about the hack, if they know!
MJPEG is not as modern or as efficient as the 5D Mark II’s H.264 with disk space, but as the name ‘JPEG’ would suggest, when JPEG goes to the movies it’s capable of producing just as good results as it can do with photographs. At high bitrates it shows hardly any noticeable compression or artefacts. H.264 on the 5D Mark II shows more compression. In fact, 50Mbit MJPEG on the GH1 is virtually in ProRes league.
Because of Microsoft’s ridiculous legacy file format FAT which is still in use all over the world, all cameras have a 2GB size limit on individual files. So when a clip reaches 2GB, it has to either span seamlessly onto a 2nd file or recording must stop. Like the 5D Mark II, the GH1′s high bitrate MJPEG has a 2GB limit on each clip, and you can record as many clips as you can fit on your card.
Such is the quality of the footage and the low compression and low efficiently of MJPEG, clips are limited to just 3 minutes each before they hit 2GB.
My GH1 reports a maximum clip length of 2m 39sec in 50Mbit 1080p MJPEG mode before recording begins. During tests I’ve seen as high as 8 minutes for reduced bitrates. But because 50Mbit clips are actually recorded with a variable bit rate depending on the shot, the actual running length of a clip may be higher or lower, depending on the amount of detail in the scene.
Personally speaking, a 3 minute clip length limit doesn’t usually affect my kind of shooting style, and it shouldn’t be a problem for most narrative filmmaking either.
It must be said that all this wonderful leap in quality is not going to be as noticable in a small window on a webpage. The 50Mbit MJPEG comes into it’s own on a large cinema screen, or decent Plasma TV. Yes, you notice the difference on a PC too (full screen) – it’s MUCH cleaner looking and more detailed. I suggest at the very least, to download the footage from Vimeo and play this on a TV.
Later in the week I will make available some of the raw footage so you can compare it directly to:
What we had before (17Mbps AVCHD) and crappy 720p MJPEG
The new native 32Mbit AVCHD 1080/24p
And we’ll compare the clips straight off the camera, the 50Mbit MJPEG and 32Mbit AVCHD directly uploaded to the web.
Until then I have completed the short film above and uploaded a sample of the raw 50Mbit MJPEG 1080p footage to Vimeo. Don’t view this on Vimeo, click the download link and you’ll have the original .MOV file straight from my GH1 and see for yourself that it’s true 50Mbit footage at 1080p!
Hopefully this article will give you some idea of what a great leap the GH1 has made. But wait until you see the head-to-head comparisons of the various formats – they’re mindblowing.
Summary, what is means for the none-techno-geek
Hugely better image quality. 50MBit MJPEG shows zero compression artefacts, zero mud and much more photo-like gradients, tones and textures
It’s better looking than the 5D Mark II’s H.264 and less compressed
MPEG 50Mbit shows low noise and the noise is of a finer grain, more film-like
The workflow improvements are immeasurable. The 50Mbit MJPEG, this can be edited directly, no transcoding required. As for AVCHD, thanks to Tester13 the native 24p can go straight into Log & Transfer without pulldown or deinterlacing work. Before I was waiting around 6 or 7 hours per project for Voltaic to transcode and deinterlace the GH1′s AVCHD to ProRes. Now the same amount of footage can be done in one step in Final Cut Pro Log & Transfer and takes 30 minutes.
Non-native 24p 32Mbit AVCHD (60i) can be played back in-camera. Focus assist works in both AVCHD and MJPEG mode. It’s expected playback of native 24p AVCHD can be fixed in the near future once a patch is released by Abed.
A JPEG is saved containing shot EXIF info, shutter speed, ISO, etc.
The image is smoother looking and better scaled from the 12MP CMOS compared to the 5D Mark II which has 22MP to deal with. Pixel binning is in action, not line skipping.
The Setting for PTools V3.35+
EOSHD 50Mbit 1080p MJPEG (30p)
MJPEG Size 1280M = Ticked
MJPEG Enco 1280M = Ticked
MJPEG E1 Quality = 400
MJPEG E1 Table = 103
MJPEG E2 Quality = 350
MJPEG E2 Table = 116
MJPEG E3 Quality = 250
MJPEG E3 Table = 120
MJPEG E4 Quality = 200
MJPEG E4 Table = 125
Tester13 32MBit 1080p AVCHD (24p)
Native 24p = Ticked
Video Buffer = Select 40,000,000
Video Bitrate = 40000000
Overall Bitrate = 50000000
Limiting Bitrate = 70000000
Version Change = Set appropriate increment (I’m on 64!)
Interface Language = Set prefferred language
I have ALL other options unticked
See this easy guide if you’re using PTools for the first time
This is a low risk thing to do. But not 100% without any risk at all.
You will need a fully charged battery.