Convergent Designs Flash XDR

[Mike Schell]

Hi John-
Thanks for the explanation! It makes good sense to me. I think the histogram offers a more qualitative view of how compression affects the quality of the images. It certainly collaborates our earlier findings, that 50Mbps MPEG2 outperforms DVCProHD, at least on a still image.

A local EX1 user, Jim Arthurs, sent me a Photoshop file of the compressed images, as individual layers. By turning on one layer in normal mode and a second layer in difference mode, you can easily see the effects of compression. Again the results show the superior results of the 50 Mbps rate and near ideal results with 100 Mbps (you see almost no differences to the uncompressed).

[David Heath]

Originally Posted by John Richard

A summary of a histogram in Photoshop is that it basically shows the distribution of all the pixels captured by the camera.

The horizontal scale goes from 0 on the left where total black is, to 255 on the right where total white is. ...........

In the 4 histograms provided for each of the respective formats, they all have gaps due to their 8 bit structure.

Yes, to the first bit of that, but an 8 bit structure inherently gives 2 to the power 8 levels - 256 - hence the 8 bit aspect doesn't explain the gaps.

If I import 8 bit DSLR images into Photoshop, the originals show a smooth histogram structure. Manipulating those images by such as gamma, black level or colour correction then produces histogram "gaps", as the process favours some levels over others. (Start with a 10 bit original, and this will obviously be much diminished.)

.......a Photoshop file of the compressed images, as individual layers. By turning on one layer in normal mode and a second layer in difference mode, you can easily see the effects of compression......

I'd recommend this as an excellent way of quantifying the differences, and treating Y,U,V separately can help to show what's going on. Normally the compressed signals are subtracted from the corresponding uncompressed ones, and mid grey added. Perfect compression will show up as a plain grey, the more detail that can be seen, the more errors the compression is introducing.

It can also show up compression effects due to motion, if the filmstrip feature within Photoshop is used.

[John Richard]

David:

Please excuse my math ignorance. I surely would appreciate a little more help understanding your explanation:

"Yes, to the first bit of that, but an 8 bit structure inherently gives 2 to the power 8 levels - 256 - hence the 8 bit aspect doesn't explain the gaps."

My thoughts on thinking the gaps were due to an 8 bit photo were due to their inclusion in the standard Photoshop 24 bit histogram.

Maybe I should do some of my own screen grabs of HDV material from our Canon H1. Could it be that the gaps are created by the fact that even though the H1 is putting out a 10 bit signal from the HD-SDI, 2 of those 10 bits actually have no data reportedly?

[David Heath]

Originally Posted by John Richard

My thoughts on thinking the gaps were due to an 8 bit photo were due to their inclusion in the standard Photoshop 24 bit histogram.

When you talk of Photoshop being 24 bit, that refers to 8 bits each for red, green and blue channels, hence each channel is 8 bits, or 256 levels. For video, it's different in so far as it is normally recorded as 3 component signals: luminance (Y), and difference signals (U and V), which show the difference between luminance and blue and red. From the three, red, green and blue can be reconstructed.

But each of Y,U,and V is recorded as an 8 bit signal, so video could also be described as 24 bit if you looked at the total.

It's conceivable that the gaps are due to a Y,U,V=>R,G,B conversion within Photoshop. It may be more valid to look directly within Photoshop at the Y,U,V histograms.

[Piotr Wozniacki]

Mike,

Am I right assuming that with Flash XDR, having the 2 audio tracks embedded in the HD-SDI output from the EX1, by also using the Flash XDR's XLR audio inputs one can achieve 4 channels of audio at the ouput?

This is quite important when considering between the XDR and the nano versions, while only having a camera with 2 audio channels.

[Dan Keaton]

Dear Piotr,

Your question is very interesting.

Mike will be able to give you the proper answer.

But, in the meantime, I want to post the following observations for discussion purposes.

We believe that most camera audio circuits will not be as high in quality as the audio circuits in the Flash XDR.

So for the highest quality audio, one should choose the Flash XDR external audio inputs as opposed to routing the audio to the camera first.

(The audio outputs from the Flash XDR can then be routed to your cameras inputs, if you so desire.)

Few cameras support 24 bit audio, but the HD-SDI specification does include both 16 bit and 24 bit audio.

So, in use with most cameras, the embedded audio in the HD-SDI signal will normally be 16 bit.

Whenever possible, I highly recommend using 24 bit audio. 16 bit audio can give great sound, but 24 bit audio gives you the luxury of recording great audio, even if the levels are not optimum.

A 16 bit audio recording, with the audio levels set low, leaves something to be desired. But, a 24 bit audio recording, with low levels, can be optimized in post while maintaining very high quality. Recording in 16 bit audio, with the audio levels set for the expected audio level, will not be able to handle unexpectedly loud sounds.

It is possible to record in the Flash XDR at 16 bits. This would match the audio format from the camera. It would be nice if the Flash XDR combined both sources to give you four channels of embedded audio, as you suggested.

At this time, I do not know if the HD-SDI specifications allows for two channels at 16 bit and two channels at 24 bit. We will need to determine if this is the case.

Please note that if

1. You send an HD-SDI signal, without embedded audio, to the Flash XDR, and
2. You send two channels of audio to the Flash XDR via the external audio inputs, the Flash XDR HD-SDI outputs will have the audio embedded in the signal.

In other words, the Flash XDR acts as an audio embedder. I feel that this is an important feature. The same applies to the external timecode input.

I am certain that Mike will answer your original question as soon as possible.

[Mike Schell]

Originally Posted by Piotr Wozniacki

Mike,

Am I right assuming that with Flash XDR, having the 2 audio tracks embedded in the HD-SDI output from the EX1, by also using the Flash XDR's XLR audio inputs one can achieve 4 channels of audio at the ouput?

This is quite important when considering between the XDR and the nano versions, while only having a camera with 2 audio channels.

Hi Piotr-
I just spoke to the hardware designer, who wondered when this question would arise. Yes, you can take 2 channels of embedded audio and mix with 2 channels of analog audio on the XDR. Of course this is not possible on the nanoFlash.

[Piotr Wozniacki]

Dan & Mike,

Thanks for this answer, as well as the interesting considerations re 24 vs 16 bit audio.

Any news on proposed XDR mounting options with the EX1?

[David Schmerin]

Originally Posted by Mike Schell

Hi Thomas-
Good suggestions. As soon as we have done some basic verifications of our MXF file format, we will post example files for download and test. We have the MXF code finished, but we need to complete the FAT32 code first, which is very close.

Just curious if there were any updates on the FAT32 code? The still images are fun but I am a video guy and I need to see these .mxf files work in FCP. This may be the greatest recorder in the world but it is a paper weight unless NLE's can read the video files.

David Schmerin

[Noah Yuan-Vogel]

one thing i think that might be important to keep in mind when comparing codecs especially through histograms is that the XDR is probably the only codec that isnt scaling the input*. seems pretty likely that the differences in the histogram might be related to the fact that different pixel values are being created through the combining/filtering/averaging of surrounding pixel values. that might be all you are seeing in the histograms. of course that is a wonderful thing, since scaling introduces various kinds of artifacts and of course softening itself, but im not sure those histograms tell us much about the compression codecs and how they compare. perhaps you could throw in another codec for comparison that supports 1080p without scaling, such as 50 or 35mbps xdcam or even something more competative like hdcamsr or cineform (especially relevant if cineform ever comes out with their similar solid state recording device)?

*for reference, dvcprohd scales the image to 1280x1080 and hdv scales it to 1440x1080, whereas xdr maintains the full 1920x1080 frame

[Craig Irving]

Yeah I'm bothered by the fact that Cineform haven't given an ETA on when their device might be coming to market, which is why I'll probably end up buying this NanoFlash XDR. For me it's kinda, whichever comes first. Codec isn't TOO important with me, because they're both 4:2:2, perform inverse-telecine, and offer very high bitrate/quality.

[Thomas Horton]

I really like the specs on the Flash XDR... One very important factor for me... Will I be able to start/stop the unit via the built in start/stop rec. button (handgrip) on the camera?

If yes, how will this be achieved (time code movement or firewire out to trigger start/stop)?

Thanks,
Thomas

[Dan Keaton]

Dear Thomas,

If you set your camera to "Record Run" timecode, the Flash XDR will detect when the timecode starts (via the timecode embedded in the HD-SDI signal or the External Timecode Input) and then automatically start recording.

[Mike Schell]

Originally Posted by David Schmerin

Just curious if there were any updates on the FAT32 code? The still images are fun but I am a video guy and I need to see these .mxf files work in FCP. This may be the greatest recorder in the world but it is a paper weight unless NLE's can read the video files.

David Schmerin

Hi David-
We're very close on the FAT32 and MXF file formats. We have the MXF code working on the MAC, just need to transfer the code to the MCU inside XDR. We will post files on our website ASAP and I'll post a message on this forum when they are avaialble.

[Mike Schell]

Originally Posted by Piotr Wozniacki

Dan & Mike,

Thanks for this answer, as well as the interesting considerations re 24 vs 16 bit audio.

Any news on proposed XDR mounting options with the EX1?

Hi Piotr-
Yes, we have some updates on the mounting options. I will try to post next week.