This little Canon 720p camcorder looks nice!

[Hulk]

http://www.usa.canon.com/templatedat...70221_tx1.html

[Hulk]

Not for serious shooting but for record keeping shots anyway.

I hope it shows some decent quality.

[Jerry Jones]

I read somewhere recently that camcorder sales are slowing relative to still image cameras.

Is it possible this means more and more manufacturers are going to build enhanced video recording capability into still image cameras as a means of boosting demand for those types of cameras?

I suspect the answer is "yes."

Canon usually delivers great image quality in decent light so I would expect the 720p video from this camera to be OK, but the huge question -- in my mind -- is whether a 30 frames per second 720p can deliver enough quality when the networks generally rely on a 60 frames per second 720p.

Sony has apparently achieved some sort of breakthrough with 60 frames per second.

Link to Sony PDF About New Sensor: http://tinyurl.com/2rlx2k

Link to Sony Web site news release: http://tinyurl.com/3b8u8c

Jerry Jones

http://www.jonesgroup.net

[Jerry Jones]

Here is a related PC Magazine article: http://tinyurl.com/yrlh5y

"26 minutes of 720p on a 4gb card."

Jerry Jones

http://www.jonesgroup.net

[Jerry Jones]

Mark,

Please check my math; you are an engineer.

The Canon specifications page here...

http://tinyurl.com/yv4h8b

...suggests the file size -- at the highest quality 720p setting -- will be 4,480KB/second.

That should translate to 35.84Mbps, right?

If so, the 720p pictures from this tiny camera should be quite good.

I can't wait to see the samples.

Because we're talking about Motion JPEG compression here.

The conversion formula I used came from this Web page:

http://tinyurl.com/dkzt5

Do you get 35.84Mbps, too?

Or are my calculations off?

Anxious to hear back from you,

Jerry Jones

http://www.jonesgroup.net

[Hulk]

Jerry,

I think you switched a number somewhere.

Doing the unit conversions..

(4GB/26min) x (1024MB/1GB) x (1min/60sec) x (8Mb/1MB) = 21.005Mb/sec

What we initially know is our data rate of 4GB per 26 minutes.

First conversion makes "GB" cancel so we have GB/min units

Second coversion makes minutes cancel so we have GB/sec units

Third conversion makes "MB" cancel so we are left with units of Mb/sec, sometimes written Mbps for "Mega bits per second."

It's harder to see when you can write out top over bottom fractions on paper but the idea is to put down what you start with and then put in conversion factors so the right unit cancels. In the first one my conversion if 1GB is equal to 1024MB so I have to either write "1GB/1024MB" (read 1GB per 1024MB's) or "1024MB/1GB" (read 1024MB's per 1GB). As you can see both are valid but only the latter "1024MB/1GB" will put "GB" in the denominator so it cancels with the "GB" unit in the "4GB/26min" so that only MB are left.

I used to spend and entire year trying to teach my physics and chemistry students this unit conversion method and unfortunately only a few of them get it. There's really no one formula for it but putting down your starting units and then the conversion factors in the right way so all units but what you want to be left with are remaining. Then all you have to do is do the arithmetic with the numbers to get the numerical answer.

The other science teachers would give the kids page after page of equations to memorize when they really only needed to get the theory for this. I thought it was worth the time to teach the method as it would serve them for the rest of their lives. But admittedly I never really found a great way to get the idea across. It just takes lots of practice.

Generally you don't write the "1" in front of the units as it is implied but I wanted to be as clear as possible.

Anyway I hope I'm right on the answer but I think I am.

That does seem like a quite high data rate for such a little camera. It bodes well for the image quality.

Here's another test for the method:

(13GB/1 hour) x (1 hour/60 min) x (1 min/60 sec) x (1024MB/1 GB) = 3.679 MB/sec

or doing one more conversion

(3.679MB/sec) x (8Mb/1MB) = 29.58Mb/sec, which is the about the data rate for DV, keeping in mind that includes the uncompressed audio stream data as well.

[Jon P. Inghram]

Going by Canon's stated KB/s rate, it's either 36.7 Mbit/s (if they used KB as 1024 bytes, "KiB") or 35.84 Mbit/s (if they used KB as 1000 bytes.) And it's using MJPEG which is less efficient than DV due to the lack of intraframe compression.

[Hulk]

Well now I'm confused. I am quite sure my calculation is correct if we assume 26 minutes recording time and a 4GB card.

Let's see...

(4480KB/sec) x (MB/1024KB) x (8Mb/MB) = 35Mb/sec

Checking Canon's time for 512MB card using their stated data rate:

(512MB) x (sec/4480KB) x (1024KB/MB) = 117 seconds, or 1 minute 57 seconds

If we use 1000KB per MB then it comes out to 114 seconds or 1 minute 54 seconds

Same calcuation on 32MB card gives 7.14 seconds using 1000KB=1MB

Canon's calcs seem wrong to me.

Okay working backwards to find data rate:

(512MB/102 sec) X (1024KB/MB) = 5140KB/sec

or using 1000KB/MB conversion we get 5019.6KB/sec

Using Canon's data table for how long recording time you get with each card size, the calculated data rate varies from around 5000KB/sec to 5400KB/sec, which is over 40Mbps.

I am quite confident my calculations are correct. All you have to do is look at Canon's table and you can see it's wacky.

128MB is 4 times more than 32MB right? But the recording time goes to 26 seconds, not 6x4=24 seconds.

And the 512MB card is 16 times larger than the 32MB card so recording time should be seconds x 16 = 96 seconds or 1 minute 36 seconds, not 1 minute 42 seconds.

Now I realize the larger the card the more accurate the average bitrate should be, assuming this camera is using VBR recording but I would assume Canon would use the same average bitrate to calculate recording time for each card size? Did they actually record and fill up the card in real world usage and that's the variation?

Finally, a 4GB card hold 8 times more than a 512MB card. Multiplying 1 minute 42 seconds times 8 = 13 minutes 36 seconds, or about half of 26 minutes.

So my initial calculation of 21mpbs is correct. I believe the problem is PC magazine quoted a 4GB recording time for the LP recording setting.

So that solves one riddle.

But still if you use Canon's quoted recording time for a 512MB card you come up with a recording data rate average of about 40mpbs.

My recording experience with MJPEG is PICVideo. I usually capture at quality 19 which is generally (depending on content) a bit over 5MB/sec. 720p is about 2.66 times more pixels than 480i so to equal PICVideo Q19 quality it would need about 12MB/sec. Not the 5MB/sec I'm calculating based on recording times on Canon's site.

Still PICVideo quality 19 is VERY high quality, almost indistinguishable from lossless compression. If Canon is using a good compression algorithm, which I bet they are, and they are using a good lens block and CCD, which they generally go, and if they are using a good image processor, I think their Digic is the best, then the quality could be very good indeed.

And MJPEG is VERY robust when it comes to editing. Robust and fast.

This little guy could be promising.







Jon,

I wasn't implying the new Canon camera uses DV compression. That was just a calculation example.

And I thought both MJPEG and DV compression were both intraframe? i.e. one frame does not need any others to be decoded? I believe they both use the same basic FFT for compression.

[Jon P. Inghram]

Sorry, you're right about DV: I misread intra as inter and thought DV used interframe compression.

Anyway, the file bitrate (which includes audio and video data) which Canon shows in their spec pages are accurate, at least with my S3 IS which happens to show the same bitrate as the TX1 with 30 fps 640x480: around 1950 KB/s. For example, one video averaged 14.5 Mbit/s video with CD bitrate audio at 1411 kbit/s.

[Hulk]

There is something wrong since they aren't consistent from one card size to the next.

I need to subtract the audio date from my figures above. Looks like it's probably 48KHz stereo PCM. That's about 192KBps.

So I'm still calculating about 4800KBps based on their recording time vs. card size.

They are saying 4480KB/sec.

If you see where I've made a mistake please correct me.

If they were going to pick an average bitrate
1. Why wouldn't it be consistent from 32 to 128 to 512MB cards in regards to recording time.

2. Why wouldn't the average bitrate shown actually give the correct recording time for the storage capacity of the device listed.

From my calculations they got the bitrate wrong and it varies a bit for each card size!

[Jerry Jones]

Mark,

I suspect the "26 minutes per 4gb card" business is a statement from PC Magazine that may begin with a tricky assumption, namely, use of the "LP" 720p mode. Might that explain the confusion?

Let's instead focus on the official CANON specifications page here:

http://tinyurl.com/yv4h8b

Here are my steps:

1. (per this conversion page: http://tinyurl.com/dkzt5). According to the conversion page, "Because there are 8 bits in a Byte, to get a bit-rate (speed) from values given in Bytes, you must multiply the total number of Bytes by 8." So... 4,480KB/second (4,480 kilobytes per second) x 1000 = 4,480,000 bytes (one kilobyte is about one thousand bytes).

2. 4,480,000 bytes per second x 8 = 35,840,000 bits per second

3. 35,840,000 bits per second divided by 1,000,000 = 35.84Mbps (Mega-bits per second).

Isn't this correct?

Also, the CANON table shows something curious... that is a bit confusing.

It seems to be saying that a 512MB card can hold approximately 1 minute and 42 seconds of "non LP" 720p (and 3 minutes and 17 seconds for "LP" 720p).

At least, that's what the table seems to be saying.

Jerry Jones

http://www.jonesgroup.net

[Hulk]

Jerry,

Yes your calculation is correct.

But if you assume the recording time is correct in the table and compute the data rate from that you get about 40mbps, or 38 taking out the PCM audio.

36 or 38 I guess we're close enough.

Should be high enough for good quality.

[Jerry Jones]

Yes!

I was hoping this would be the case.

I'm interested in this camera.

It apparently records 44KHz audio with a built-in microphone, which is okay, but I'd use a Sony MiniDisc recorder for audio acquisition to get around the lack of a dedicated microphone input terminal.

Now it's just a matter of seeing some sample clips.

Jerry Jones

http://www.jonesgroup.net

[Jerry Jones]

But this little camcorder would really need to support 8gb memory cards to be truly practical.

Is there any reason why it couldn't support an 8gb card?

Jerry Jones

http://www.jonesgroup.net

[Hulk]

One would think so. But the fact that Canon does not list that size might mean it does not or they haven't tested it with one.

Okay so now they just need to make some changes to the HV20. Add 4 card slots in place of the tape transport. Give us the option of 720p and MJPEG recording format with data rates up to 100mbps.