Hardware encoder to get professional quality mpeg2 video [Archive]

View Full Version : Hardware encoder to get professional quality mpeg2 video

astronob

11th March 2009, 03:46

I'd appreciate some recommendations from those professionals who author or have authored dvd's from any of the major studios or record companies. I want to encode mpeg2 to the same quality as on commercially available dvd's. I presume that means a hardware encoder, as no software encoder I've ever used has produced an encode as sharp as on a commercial dvd. If I want to stick to the windows platform, what are my choices? Can any hardware encoder encode say, avi directly to mpeg2 format, or are they limited to SDI input or the like? If they are limited to SDI, is there away around this obstacle? I don't want to have to buy a VTR with SDI output, it's beyond what I can afford. I have heard of software VTR's which can fool hardware encoders to think they are controlling a "real" VTR, but they're still very expensive (of course, nowhere near as expensive as the real hardware). Any suggestions are appreciated. Thank you.

neuron2

11th March 2009, 03:58

I presume that means a hardware encoder, as no software encoder I've ever used has produced an encode as sharp as on a commercial dvd. That's a false presumption. Software encoders can do a much better job.

What software encoders have you used?

Dark Shikari

11th March 2009, 04:06

I'd appreciate some recommendations from those professionals who author or have authored dvd's from any of the major studios or record companies. I want to encode mpeg2 to the same quality as on commercially available dvd's.You're setting your sights quite low then, given how commercial DVDs are possibly the worst examples of offline MPEG-2 encoding known to man.

setarip_old

11th March 2009, 21:50

@astronob

Hi!I presume that means a hardware encoder, as no software encoder I've ever used has produced an encode as sharp as on a commercial dvd. If I want to stick to the windows platform, what are my choices? Can any hardware encoder encode say, avi directly to mpeg2 format...I can assure you that no commercial DVDs are made using .AVIs as source material...

smok3

11th March 2009, 22:47

setarip_old: what's wrong with AVIs?

Dark Shikari

11th March 2009, 22:50

setarip_old: what's wrong with AVIs?Any sane authoring house would use the original film transfer or such, not an AVI, for authoring.

smok3

11th March 2009, 23:02

workflow links? (enlighten me)

reykroona

12th March 2009, 07:06

We use a CinemaCraft setup for both DVD and Blu-ray.

Our DVD hardware is an older system that uses a Digital Beta tape deck and SDI signal into a box that then compresses to MPEG2. However you have to run the movie several a few times, especially when removing pulldown. CinemaCraft has a software based version that works off AVI, but we've never used it.

You mentioned deck emulation. Avid can do this and I remember seeing a Mac program that can do this as well (Deck Link has Virtual VTR, but I think it was something else).

Our Blu-ray setup goes HD-SDI to a computer where we make a HUGE uncompressed AVI, MOV, or DPX sequence (around 750GB) and then use a software encoder to encode the H.264 files. So to respond to the AVI statement, as long as you use a good 4:2:2 10-bit uncompressed AVI, there is no difference between that and a Digital Beta, HD-Cam SR, or D5 tape. Though I think a lot of post houses are getting these files straight from the telecine place, so they aren't even using tapes anymore. We aren't so lucky.

If you are looking for other realtime options Digital-Rapids makes a pretty cool card, if you get a good enough computer you can run two encodes at the same time, depending on the card you get. Their MPEG2 encoder is made by MainStream.

As far as hardware vs software, I would think computers are fast enough these days that a several pass software encode (for SD) would actually be faster than running through the tape 3 or 4 times. That is if you don't have to make the uncompressed AVI/MOV and you care enough about quality to do more than 1 pass. HD is a different story. We found it typically takes 16-20 hours to compress a BD movie.

Lyris

13th March 2009, 06:46

@astronob

Hi!I can assure you that no commercial DVDs are made using .AVIs as source material...
Not necessarily true :)
I have used .AVIs - but you do have to be careful and make sure that the codec you're using is lossless (or as good as) and isn't doing anything funny with the levels or colorimetry.

Of course, as Rey points out...

So to respond to the AVI statement, as long as you use a good 4:2:2 10-bit uncompressed AVI, there is no difference between that and a Digital Beta, HD-Cam SR, or D5 tape.

The AVIs were created by capturing losslessly from DigiBeta :)

For the record, although I've not authored for major studios, I use Cinema Craft SP2. The quality of its output (especially with the low pass filter disabled) is ridiculously good when you look at what Hollywood is putting out (as Dark Shikari alluded to).

astronob

14th March 2009, 01:22

I think you may be missing my point. If I want to encode a high quality uncompressed avi (as reykroona pointed out) to mpeg2, to a quality standard as good as what is possible (commercially speaking), what is a good hardware encoder that can achieve this? As I said, I've yet to see a software encoder that can produce a video as sharp as the best examples on commercially available dvd's. I'm happy to be proven wrong, but I've tried pretty much all the software encoders out there (CCE, Tmpgenc, Procoder etc), but still can't quite get the sharpness of video I seek.

Dark Shikari

14th March 2009, 01:32

I think you may be missing my point. If I want to encode a high quality uncompressed avi (as reykroona pointed out) to mpeg2, to a quality standard as good as what is possible (commercially speaking), what is the best hardware encoder that can achieve this? As I said, I've yet to see a software encoder that can produce a video as sharp as the best examples on commercially available dvd's.
Have you tried HCEnc?

Additionally, have you ever considered that maybe those "sharp" DVDs were run through a sharpening filter before encoding? :p

Lyris

14th March 2009, 02:55

It's odd that you say the Hardware Encoders look sharper. More often than not, these come configured to Low-pass filter the video which removes detail. Most software encoders do not do this.

Unfortunately I can't help out any more with hardware encoders, I'm just saying that there's no good reason why the software ones should look softer.

astronob

14th March 2009, 07:57

All I'm saying as I have not been able to reproduce a video as sharp as on commercial standard dvd's with encoding software. There is always a (visible to the eye) loss of some sharpness in the video. Encoding an uncompressed avi (even say avi in HD format), to say xvid, will yied me a distinctly sharper video than mpeg2, irrespective of the mpeg2 encoder used. Yet I have seen mpeg2 much sharper, with more detail, than I can encode. So, what to do? How to achieve that quality? Does one need to go the next (more expensive) step and get the cinemacraft xtream to achieve the quality sought?

Dark Shikari

14th March 2009, 08:18

Lyris

14th March 2009, 10:09

Astronob: are you talking about spatial or temporal degradation? Is the lack of perceived clarity you mention in still frames or just in moving parts?

Also, since just about every commercial DVD is low-pass filtered - often quite heavily - you're sure you're not seeing the ringing around (what was once) high-frequency areas and seeing this as genuine detail?

astronob

14th March 2009, 10:37

Still frames or moving frames Lyris, it matters not. All I'm saying is basically, if I want to produce mpeg2 as good as what is commercially available (ie the best around), how expensive do I have to go? "Without" going the whole digibeta route. Is a mac and cce-mp (since it's based on the "supposedly" better xtream encoder) a better choice than windows and cce sp2. Is a hardware encoder Sonic Sd-2000, Sony Vizaro V701 or some other hardware encoder a better choice? All I'm saying is I've tried all the commercially available (reasonably affordable) software choices and I can't produce an mpeg2 encode as the best examples I've seen. Even if I use uncompressed HD avi as a source file.

Dark Shikari

14th March 2009, 10:41

neuron2

14th March 2009, 15:35

@astronob

Provide screenshots that demonstrate what you are claiming. Without those, you are just blowing hot air and wasting everyone's time.

I suggest that you post a screenshot of the putative high quality encode that you are trying to emulate, and a screenshot from an HCEnc encode together with the settings you used and your Avisynth script. The claimed difference will hopefully be apparent and we will be able to tell you how to change the HCEnc settings and/or Avisynth script to fix that, assuming that the difference is a real difference of quality and not just ringing or using the wrong color matrix, etc.

Be aware that using the wrong color matrix can make the picture look soft. Your error may be as simple as that.

Lyris

14th March 2009, 18:24

Is a mac and cce-mp (since it's based on the "supposedly" better xtream encoder) a better choice than windows and cce sp2.
I've heard that it is, but I don't think we've seen any real comparative testing done. From the looks of it though, it has less options to tweak. Unless I'm mistaken, you cannot assign different quantization characteristics to different regions of the video (which I like to do because I'm anal retentive about encode quality). In that case, it could make it worse in some cases.

I'd like to see screen shots too, though. Are you evaluating both encodes on exactly the same display/setup?

astronob

15th March 2009, 01:42

It will be a few weeks before I can provide screenshots (pc's not working, waiting on a part), but I'll have a go and post a few then. I haven't tried Hcenc, but no harm in giving it a go.

neuron2

15th March 2009, 01:52

@astronob

Do you know about color matrices?

drpaulng

16th March 2009, 01:55

One important trick before encoding is "dithering". It prevents video details from being grouped into "bandings" or "patches" during encoding. I think it is a key for "professional" encoding. I never do it because of it is too time-consumming for layman like me.

neuron2

16th March 2009, 02:23

But he's complaining about softness, not banding.

astronob

16th March 2009, 08:36

I don't know about color matrices neuron2, sorry. I'm also sending emails off to a few production houses, see what they say too.

smok3

16th March 2009, 09:13

neuron2

16th March 2009, 12:43

I don't know about color matrices neuron2, sorry. So you're ignorant of an important area of encoding that can cause exactly the issue you mention. Don't you think it behooves you to find out about it?

In any case we all await your samples, with which we will be able to quickly set you straight.

drpaulng

16th March 2009, 18:54

During the course of compression (data size reduction), near-colored pixels at certain regions may tend to be homogenized into a group of pixels that share the same picture values (banding - loss of details). By adding certain kind of noise to the original picture before encoding, banding can be avoided or reduced in amount and therefore preserves certain level of fidelity to the original picture. The higher the data compression rate, the more lossy is the final product. With the trick of dithering, we can fool the eyes into seeing more picture details at higher compression rate. Sharpness of picture in compressed form should never exceed the original, otherwise it is a sharpened picture through digital signal processing irrelevant to encoding (data size reduction). The sharpened picture filtered through DSP should occupy the same data size compared to the original without taking the next step of encoding. Some professional picture or video encoders provide with a dithering step before encoding. Use it at the best interest of preserving details while compressing data down to more than 10:1 of size. Or you may preserve a lot of details by compressing at 2:1 of size without caring much about dithering that adds "noise" to the original picture.

In short, it is an art of trading, a compromise between data size and details.

astronob

17th March 2009, 08:33

Thank you for that most informative explanation drpaulng. :D

scharfis_brain

17th March 2009, 08:58

unfortunately dither will be the first thing to be removed after quantisation within the MPEG encoder.
even at very high bitrates.

So one would have to apply noise (ie. extreme dither) to get it through the mpeg-decoder