FAQ:The MPEG-4 Recorder
From The Neuros Technology Wiki
NOTE:
This is the FAQ for issues common to all versions of the MPEG-4 Recorder. For questions specific to the Recorder 2 see the Recorder 2 FAQ
[edit] Which Version of the Recorder do I need?
See Here
[edit] Will a "full screen" dvd maintain its aspect ratio, or be stretched to 16:9?
Answer This is a somewhat complicated question, but the short answer is that again the recorder copies what would be seen on a normal 4:3 TV set. Typically this means black bars at the top and bottom of the file (when 320x240 or 640x480 is chosen). The good news is that the PSP has a zoom function which allows you to zoom in on the image and fill the screen. In addition, the recorder offers a "WQVGA" resolution setting that is closer to the aspect ratio of the PSP. The trouble with this setting is that the PSP won't allow zooming on it for some reason. For more information look Here
[edit] Can the MPEG-4 Recorders record DVDs?
Answer You can basically record anything with the recorder that you can watch on a TV set. The recorder does not attempt to inforce copyright laws, that is up to our users to do. We do not condone or encourage copyright infringment and we hope that our users will not use the MPEG-4 Recorders to break the law, but the devices cannot distinguish between fair use copying and infringment, so we do not cripple the devices on the assumption that our users are criminals. In other words, the device does not seek out legacy copyright protection mechanisms to determine if recording should be allowed.
In addition, we don't believe that "crippling the analog hole" does anything to prevent piracy when pirates have access to much more direct ways of breaking encryption.
See Here for more information on Neuros' position on your digital rights.
[edit] Does that mean the Recorders are not affected by Macrovision?
(Macrovision refers to a legacy copy-prevention mechanism that inserts garbage into the non-visible portions of an NTSC signal in an attempt to confuse home VCRs.)
Yes. The Neuros Recorder is not like a VCR and so Macrovision does not cause mistracking of the heads (the recorder has no heads). Further, we have not added circuitry or logic to artificially make the Recorder simulate the behavior of a VCR in response to Macrovision.
[edit] Any way to record movies with subtitles to be viewed on a psp?
Answer Most Subtitles are not yet supported. I'm not sure if subtitles are supported with the .mp4 wrapper or not, and I'm not sure if the PSP supports subtitles on .mp4, so I'm not sure if that's even a possibility. This requires more research.
why not? what you see is what you get. if you play movie with subtitles and record the video at same time, then you have subtitles.
[edit] Any way to record audio only? (Like from a mic?)
Answer: Not yet.
[edit] Any way to use battery power for mobile recording?
Answer: Nothing official yet. Some of the helmetcam folks I believe have hacked something together, if anyone has links, please post.
Olson Engineering has added a battery pack.
[edit] Is there signal degredation when passing through video on the OSD (R3) ?
No, in the R3 we designed in additional circuitry to be used as a pass through mode, where the incoming composite video gets piped directly to the output without any distortion or degradation.
[edit] Additional Engineering Q&A on Video Quality
Since the A/D D/A ICs for video encode/decode (which I believe are from TI) have some inherent quality loss isn't it in Neuros' best interest to use ICs that minimize this issue because the essence of the product is recording and recording requires A/D D/A conversion. Therefore, switching to another video encode/decode supplier my be in store. Is it true that A/D -> D/A conversion from video encode/decode suppliers (other than TI) do not have any noticeable degradation. There is some level of quality loss when using a composite audio and video in/out based on the simple laws of physics. How much is that loss from composite video?
A. First on must acknowledge that composite video is a less than ideal format to manipulate.
Yes, having composite video and audio together, due to laws of physics, have degradation. But this degradation is minimal enough that I doubt that it can be measured or detected. This should not be a source of worry.
Of course the quality of the ADC and DAC used have an impact on the overall quality of the video. However, the ADC and DAC used by the R3 (and R2) have enough resolution and sample rate to create a stunning NTSC or PAL signal. That is not the problem.
The degradation is caused by the definition of video itself. The statement "We have seen A/D -> D/A conversion from video encode/decode suppliers (other than TI) that do not have any noticeable degradation" is certainly the result of either some trickery much like the one we are doing on the R3, a TV bad enough that the degradation is not perceptible or eyes not trained to see the type of degradation expected. Chances are it is some combination of them.
The biggest quality degradation, by very far, comes from the necessary separation of luminance and chrominance prior to decoding. Unfortunately this is a necessary step. A video generator (like a camera or telecine) will start with RGB signals and encode into NTSC (or PAL, which is the same for the purpose of this discussion) and send the signal to a video sink downstream (such as a TV). There the video will be opened and decoded into RGB. However, because the encoding and decoding processes are lossy (by definition), information gets lost and the SNR gets lower.
The "rub" is that when the NTSC and PAL standards were created, a lot of attention was given to fit all the information in a single channel. The result is that luminance and chrominance signals occupy an overlapping band of frequencies, and even the best filters in the world cannot perfectly separate them. The net result is that certain parts of the luminance end up as noise in the chrominance, and certain parts of the chrominance end up as noise in the luminance.
The biggest challenge is to receive the composite video and come up with separate luminance and chrominance signals. TV manufacturers have been improving this process for years, especially for the big screen TVs where intrinsic weaknesses of the NTSC/PAL standards can be readily observed.
Some buzzwords that may come into mind are "comb filter" and "3-D Y/C filter". These are examples of convolutional processors that do a better job in the separation process than a flat BP filter. Of course there are many others, but they all have their own advantages and disadvantages. Of course, once Y and C signals are spectrally mixed, it is truly impossible to separate them ever again.
Every time there is a processing step there will be a degradation in quality. In the TV studio world this is called "generation loss". You can get a good feel for it if you have ever tried to copy a VHS tape by playing with one VCR and recording into another. The degradation in quality is quite evident. By the time the 4th or 5th generation is made it becomes almost un-watchable.
To get around the generation loss issue (before the all-digital video studio existed) manufacturers of VCRs created first the S-Video format, with the philosophy of never mixing luminance and chrominance to start with. This is a HUGE improvement in quality, and to this day S-Video source are fairly common in the consumer space (S-Video all but died in the studio).
The R3 will have an S-Video input, so if the user has a source of S-video he doesn't need to separate Y and C.
The method for the R3 to avoid degrading composite video is not to change it at all. In this pass-through mode the video is unchanged. There is no ADC and DAC (so another brand of converters will not make the video any better). There are no filters, no nothing. Just pass through.
Also keep in mind that even more important than the Y/C separation is the compression method and ratio. If not enough bits are allocated then the image is going to be degraded. If too many bits are allocated the memory requirements become impossible or impractical.
There is no "absolute right" number of bits. A stream that appears good on the LCD of a small cell phone may well be very poor when displayed on a 26" TV.
Conversely, something that plays well on a big TV have a big chance to "choke to death" the cell phone. To add insult to injury, the Microsoft player included with Windows CE could hardly be worse.
OK, The actual ADC/DAC discrete step has minimal loss. and I understand that the compression is the largest culprit of quality loss. Unfortunately, as you know for handheld device's purposes you must compress and downsize our video's highly in order to be compatible for playback. However, there are several ways to downsize, scale, and compress video. For instance bicubic or bilinear filters for downsizing can increase the picture quality significant compared to other basic filters where rows or columns are simply removed. This functionality is either handled in the cpu or decoder itself. My guess is the TI parts don't natively support some of these more advanced filters and algorithms compared to other vendors (however I could be wrong).
The other major source of quality loss as you stated is the Y/C separation. I assume some (if not all) of the color shifting I see when using R2 is because of the required Y/C separation unless there are some issues with converting color spaces. If this assumption is correct then isn't it in Neuros' best interest to compare other vendors filters/decoders (ie: comb, y/c).
Finally, the decision to go with composite video + audio vs component is obviously very significant. Is there quantitative evidence that composite audio + video is not noticeably different than component?
The first point you make is about resizing and conditioning the incoming video prior to compression. Your understanding about some methods being better than others are correct, but this is not the case with your guess that "TI parts don't natively support some of these more advanced filters and algorithms". In fact, the resizing is performed with a combination of hardware and software in the DM320, utilizing polyphase filters. This type of filter achieves the best possible results. Not a pixel is ever dropped.
One sidebar about these filter is that their quality is so good that in some cases, typically when the bit budget is tight, it is better to compress a smaller resolution video and use the filters to resize up the output than to compress with the desired resolution.
Your next point was the Y/C separation. The best general purpose Y/C separation method, as you stated below, is the comb filter. The video decoder part that is used on the R2 and the R3 actually have adaptive comb filtering for both Y and C paths. The TI part used leaves nothing to be desired in this regard. Of course, one can always use the S-Video input, where the Y and C paths do not need lto be separated to start with.
Your next paragraph questions the "decision to go with composite video + audio vs component". I am not very sure exactly what question you have in mind. As you know, composite video and component video are two different ways to carry a video signal. Perhaps you can clarify what you meant with (composite video + audio) in one side and (component) on the other, as the audio is clearly the same for both composite and component video formats.
My previous response assumed that you were questioning our choice of using a standard A/V jack, with the triple signaling of audio_L, audio_R and video, versus using a separate jack for audio. I thought that you were refering to the fact that the audio and video signals are in somewhat close proximity of each other, and the audio signals could cause interference in the video. Since the video signal is very low impedance (75 ohms, decreed by standard) and the capacitance of the shielded cable is pretty low, the cross talk is minimal.
Obviously I did not understand the question. Can you please elaborate?
Of course, the quality of component video is much higher than that of composite video, regardless of the audio. The next "big deal" operation after the Y/C separation is demodulation of the chrominance. In component video, the color signal is never modulated into the chrominance signal, and therefore it is not necessary to demodulate it in order to receive the video signal.
In a quick overview, the major steps of a video encoder are to take the RGB signal generated by the CCD or other sourece and feed it through a matrix. The result is a luminance signal compatible with the previously existing black and white signal, and 2 new "color components" called Cb and Cr, because they carry mostly blue and mostly red parts of the image. The luminance carries mostly (59%) green.
These Cb and Cr signals are then fed into a quadrature modulator, together with the 3.58 MHz color subcarrier signal. The result is the chrominance signal, or C.
The composite sync signal is then added to the luminance signal from the matrix to form the Y signal. Then the C signal and the Y signal are added together and become NTSC or PAL, depending on the time base and color modulator.
The receiver works in the reverse order. First Y and C are separated somwhow (comb filter here). The the C signal is fed to a quadrature demodulator and Cb and Cr are obtained. The Y, Cb and Cr are fed to a matrix which outputs RGB.
All that SVHS (S-Video) and component do is bypass aspects of the NTSC/PAL generation. S-Video bypasses the Y/C separation, by far the biggest offender of quality. Component bypasses the modulator/demodulator as well. The loss introduced by the matrix is pretty small. This is why there is no great push to use RGB, with component being the de-facto standard format in the studio and for connecting HDTV analog signals.
