HD Bitrate is under 5 Mb/s for most channels - is this correct?

[sofakng]

I've been checking the bitrate of some of the HD channels of Dish Network (using the simple [file size] / [total seconds] method) and most HD channels are coming in under 5 Mb/s.

Does this seem accurate? I'm looking at some older posts on the forum and it looks like they used to broadcast around 8 Mb/s but I guess they are squeezing more HD onto the transponders?

 

[Scott Greczkowski]

You can really measure like that... if it was uncompressed 5mb/s or even 8mb/s would be really low, but DISH is using 8PSK and MPEG4 so that a 5 MB/s stream is really equal to a 12 mb/s uncompressed stream. (These numbers are not exact but is meant to show you the benefits of MPEG4.

 

[MikeD-C05]

You can really measure like that... if it was uncompressed 5mb/s or even 8mb/s would be really low, but DISH is using 8PSK and MPEG4 so that a 5 MB/s stream is really equal to a 12 mb/s uncompressed stream. (These numbers are not exact but is meant to show you the benefits of MPEG4.

So how soon till DISH goes mpeg 6 or mpeg 8? I know they want to be able to add more channels in the same space ,without damaging picture quality.

 

[Scott Greczkowski]

There is no such thing as MPEG6 or 8 yet. They haven't even fully rolled out MPEG4 yet.

In addition the mpeg4 encoders keep getting better and better.

If you looked at MPEG4 from 2 years ago until now you would see some very nice improvement.

 

[Juan]

So how soon till DISH goes mpeg 6 or mpeg 8? I know they want to be able to add more channels in the same space ,without damaging picture quality.

more like MPEG-H

Moving Picture Experts Group - Wikipedia, the free encyclopedia

 

[pendragon]

You can really measure like that... if it was uncompressed 5mb/s or even 8mb/s would be really low, but DISH is using 8PSK and MPEG4 so that a 5 MB/s stream is really equal to a 12 mb/s uncompressed stream. (These numbers are not exact but is meant to show you the benefits of MPEG4.

A few minor corrections. An uncompressed 1080i HD video stream actually runs at about 1.5 Gbits/s. An uncompressed 5.1 HD audio stream could easily be 12 Mbit/s. There are a variety of compression formats used to take these exorbitant rates down to something reasonable for current distribution technologies.

For HD video DN exclusively uses H.264 compression (sometimes ambiguously referred to here as MPEG-4, as there is more than one MPEG-4 video compression format). H.264 is about 2X more efficient than MPEG-2 for the same video quality. OTA uses MPEG-2 exclusively and DN previously used this format for HD video. I haven't surveyed audio compression on DN's HD channels, but what I've seen has been Dolby Digital (DD) at 384 kbit/s.

In terms of bit rates in the US, the FCC allows an OTA channel to use a maximum of around 19 Mbits/s. Many share this across multiple subchannels, but some stations provide a single HD channel at this rate. With DD audio and padding this easily leaves 18 Mbit/s for MPEG-2 video. That's kind of a reference standard, but higher rates are available. Blu-ray can reach around 40 Mbit/s (not talking 3D) and some satellite backhauls I've seen hit about 80 Mbit/s for a single HD video stream. OTA maximum video rates look pretty good, but the two latter formats are quite a bit better, with the last being spectacular.

The 2X rule of thumb I quoted earlier is generally accepted in the industry. However I've done some very careful and tedious comparisons of the latest H.264 and MPEG-2 encoders and believe the true number to be less than 1.8X. Regardless, an H.264 HD video stream needs to be at least 9 Mbits/s to match the OTA single HD channel reference. Many of the premium channels are fed to DN and other providers at around this rate for H.264, or double for MPEG-2.

Subtracting out the audio data rates, most of the DN HD channels clock in less than 4 Mbit/s for the video stream. However these rates are averages only. DN multiplexes several HD channels per transponder, and their compressors can dynamically allocate higher or lower rates for each channel based on moment to moment requirements. A static scene on one channel would require far less than a high action scene on another. Still the data rates do not appear to change drastically and the average rate does appear to be a reasonable predictor of video quality.

Furthermore DN reduces the resolution of a number of their HD channels from 1920x1080 to 1440x1080. This leads to a softer picture more amenable to higher compression. Regardless, a 4 Mbit/s HD video rate is a joke when compared to the distribution feed of 9-10 Mbit/s for H.264, or 15-18 Mbit/s for MPEG-2.

 

[vurbano]

You can really measure like that... if it was uncompressed 5mb/s or even 8mb/s would be really low, but DISH is using 8PSK and MPEG4 so that a 5 MB/s stream is really equal to a 12 mb/s uncompressed stream. (These numbers are not exact but is meant to show you the benefits of MPEG4.

But high quality is 9 mb/s Mpeg4 and 19 mb/s mpeg2 for OTA. Heck even FOX which frequently looks like crap was at 16 Mb/s mpeg2 at one time. 12 mb/s mpeg2 is absolute GARBAGE even at 720p and if DISH is at its equivalent? ew.

 

[mike123abc]

It is an old story. Dish (DIRECTV and cable too) have figured out that the number of HD channels in the marketing brochure sells, not the perceived quality of the picture when you actually get your system. They will compress the channels as much as they can get away with to carry as many as possible. Everyone wants the channels they watch in HD, and even HD lite is preferable to SD.

 

[3HaloODST]

Each transponder has about 42Mbps capacity. DISH places about 8 HD channels per transponder. That leaves in a worst-case scenario 5.25Mbps, but keep in mind that the channels are VBR, and static scenes can end up using less than that. When another channel is not using as much bandwidth, a channel that needs more bandwidth (high amounts of action) can "borrow" the unused bandwidth. Using the calculations you used, you will gather the average bitrate, but that is not indicative of overall quality. The only way you will see the higher bitrate that is used, is to record a scene with a lot of action for about a minute, then run calculations. Otherwise, you're just getting the ABR, which can include many still scenes.

 

[vurbano]

I would agree that you just can not look at average bit rate. You need to monitor an entire broadcast of a fast action to see what peaks they are passing.

 

[3HaloODST]

I would agree that you just can not look at average bit rate. You need to monitor an entire broadcast of a fast action to see what peaks they are passing.

That would be the most accurate way, as even calculating an action scene for a minute still gives you an ABR.

 

[pendragon]

Each transponder has about 42Mbps capacity. DISH places about 8 HD channels per transponder. That leaves in a worst-case scenario 5.25Mbps, but keep in mind that the channels are VBR, and static scenes can end up using less than that. When another channel is not using as much bandwidth, a channel that needs more bandwidth (high amounts of action) can "borrow" the unused bandwidth. Using the calculations you used, you will gather the average bitrate, but that is not indicative of overall quality. The only way you will see the higher bitrate that is used, is to record a scene with a lot of action for about a minute, then run calculations. Otherwise, you're just getting the ABR, which can include many still scenes.

I believe I addressed this as:

... Subtracting out the audio data rates, most of the DN HD channels clock in less than 4 Mbit/s for the video stream. However these rates are averages only. DN multiplexes several HD channels per transponder, and their compressors can dynamically allocate higher or lower rates for each channel based on moment to moment requirements. A static scene on one channel would require far less than a high action scene on another. Still the data rates do not appear to change drastically and the average rate does appear to be a reasonable predictor of video quality...

However, if you want graphs, I've got graphs. I've attached an instantaneous bit rate graph for a Dish Network HD channel and a comparable HD channel on Shaw Direct, a Canadian DBS provider, both measured from the original transport streams.

A few notes are in order: the Dish numbers are for a H.264 video stream, but the format was misinterpreted by the tool as 1080p30 instead of 1080i60. Times should be halved and bit rates should be doubled for this graph. Shaw uses MPEG-2, which requires about twice the bit rate for the same quality; the Shaw graph numbers are correct as shown.

The Dish graph certainly shows some peaks, but the rates are dominated by the average - the maximum rate is 7.5 Mpbs, but the average is a pathetic 3.2 Mbit/s, far below the "worst case" 5.25 Mbit/s value referred to by 3HaloODST. To compare with Shaw, the "MPEG-2" equivalent rates for Dish would be a peak of 15 Mbit/s (very nice) and an average of 6.4 Mbit/s (horrible). As one can see from the graph, the peak rates are only present for very short periods of time.

The Shaw graph demonstrates quite a different profile. The average rate is 10.8 Mbit/s (pretty good) and the peak is 19.3 Mbit/s (excellent). Unlike Dish, Shaw allows the peak rates to linger for significant lengths of time.

I've chosen these examples from a number of recent curves I've run off Dish and Shaw. There are some variations from channel to channel and source to source, but these graphs are pretty typical. Putting the numbers aside, the subjective Shaw HD picture quality makes Dish's look pedantic. The difference is clear for action scenes and static shots.

 

[3HaloODST]

Okay first off your attachments don't work. Second off my "worst case" still stands as regardless of your claims the transponders have 42Mbps capacity. You seem to be implying that the channels are using less bandwidth than the TP can supply. For all of your graphs claiming how "pathetic" the bitrates are, there are other channels on that same TP using the extra bandwidth they need for action scenes. Just because a channel has a low average bitrate and shorter peaks, does not necessarily mean it's going to result in bad PQ. There is a good chance that during that period it does not require that high of a bitrate.

 

[pendragon]

Look at the thumbnails at the bottom. For some reason there was a spurious link generated at the top of the original post which I have since deleted.

I agree DN is sending out around 42 Mbit/s on the HD transponders. However there are a lot of nulls in the streams. The nominal aggregate rate of non-null video packets is more like 25 Mbit/s, with around 3 Mbit/s of audio. They hold back the rest for peaks, which are handed out in a very stingy fashion.

I have accumulated 4+ years of transport stream recordings off of Dish and 18 months from Shaw. These are typical examples that correlate very accurately to my perceived video quality of the two. You're welcome to speculate/fantasize a different reality, but until you offer some objective evidence, there's not much to discuss.

 

[harshness]

There is no such thing as MPEG6 or 8 yet.

While the specification isn't done yet, there is groundwork for MPEG7.

Video FAQ

 

[3HaloODST]

Look at the thumbnails at the bottom. For some reason there was a spurious link generated at the top of the original post which I have since deleted.

I agree DN is sending out around 42 Mbit/s on the HD transponders. However there are a lot of nulls in the streams. The nominal aggregate rate of non-null video packets is more like 25 Mbit/s, with around 3 Mbit/s of audio. They hold back the rest for peaks, which are handed out in a very stingy fashion.

I have accumulated 4+ years of transport stream recordings off of Dish and 18 months from Shaw. These are typical examples that correlate very accurately to my perceived video quality of the two. You're welcome to speculate/fantasize a different reality, but until you offer some objective evidence, there's not much to discuss.

So now your claim is that they are stingy with handing out bandwidth. You're saying they waste 15Mbps. Another unproven claim. You're right, not a whole lot to discuss.

 

[CK SatGuy]

I don't know if this will help or not, but I'll post it. After the uplink activity that happened this week...

From DBSTalk.com:

Good news ... bad news ...

The bad news ... DISH now has three (of 19) transponders on Western Arc that have 9 HD channels on them (not counting mirrors). (There are six 9 HD channel transponders on Eastern Arc and one 10 HD transponder.)

The good news ... on both arcs DISH has removed non-sports channels from the transponders that carry part-time RSNs. This may be a step toward lighting up 24/7 RSNs - or adding more "capacity" so more than 12 channels can be aired at the same time.

Uplink Activity for the Week of 5-16-11 - DBSTalk.Com

 

[3HaloODST]

Yikes... 9 Channels... EA 10 channels? Guess I'm glad I'm on WA but I hope this isn't a continuing trend...

 

[pendragon]

So now your claim is that they are stingy with handing out bandwidth. You're saying they waste 15Mbps. Another unproven claim. You're right, not a whole lot to discuss.

I am making no claims of proof, as that would require resources in which I am unwilling to invest. However I am offering semi-objective data, from which I can make reasonable conjectures. You are merely speculating.

One of my conjectures is Dish appears to idle their channels at a base rate, leaving a "pool" of available bits that can be dynamically allocated. At the time I studied this particular transponder, the base rate was about 25 + 3 = 28 Mbit/s. The total transponder data rate was 41.2 Mbit/s, including nulls, meaning about 13 Mbit/s was available in the pool. If Dish allocated all of this dynamically, the average video rates would have been about 1.5 Mbit/s higher per channel.

With 8 (or 10!) channels to split up the remaining 13 Mbit/s, one can see the need for caution. On this particular transponder I have seen them allocate up to about 10 Mbit/s to a single channel, although that is very, very rare and only for an instant. More typically they may go up to 7.5 Mbit/s, which they could do on 3 of the 8 HD channels at any given time.

Shaw on the other hand may be more aggressive in using their pool. With only 3-4 HD channels in a 39.3 Mbit/s stream, their compressors have it easier in terms of deciding where to allocate the bits, and harder in terms of fewer statistical opportunities to make room for peaks. But they are willing to allocate, in the case given, virtually half of their transponder to a single channel for not insignificant lengths of time.

Historically Dish has gotten more and more miserly with their HD rate allocations over the 4.5 years I have monitored them. Some here drink Dish's kool-aid and believe their compressors are getting much better, ergo they are able to maintain the same quality levels at lower bit rates. Unfortunately this is not borne out by comparing recordings made of the same source material on the same channel over the time frame. The HD PQ was not outstanding on Dish in 2007, but every year since picture softness and blurring has been steadily increasing. I don't see much hope for this trend to be reversed.

 

[MikeD-C05]

While the specification isn't done yet, there is groundwork for MPEG7.

Video FAQ

So there is another mpeg # coming. I was being silly in suggesting mpeg 6 or 8 but 7 ? Cool.