Disruption of HD Sound

[Bryn]

Listening to the On Demand (Listen Again) version of the HD Sound offering for last night's extended Bruckner performance I was struck once again by occasional but annoying disruptions to the audio. Here is a link to a 21 second clip (48kHz 16 bit WAV) taken from the first movement. The sort of disruption I refer to can be heard about 12 seconds into the clip. It is more prominent in the left channel. Any of you audio techies care to suggest the cause? It does not appear to be an encoding artifact. Maybe a mic. or other audio line connection problem? There are many other examples to be found in the recording.

http://we.tl/ea3M9OPfEA

The clip will be deleted from WeTransfer.com on March 8th.

[makropulos]

Listening to the On Demand (Listen Again) version of the HD Sound offering for last night's extended Bruckner performance I was struck once again by occasional but annoying disruptions to the audio. Here is a link to a 21 second clip (48kHz 16 bit WAV) taken from the first movement. The sort of disruption I refer to can be heard about 12 seconds into the clip. It is more prominent in the left channel. Any of you audio techies care to suggest the cause? It does not appear to be an encoding artifact. Maybe a mic. or other audio line connection problem? There are many other examples to be found in the recording.

http://we.tl/ea3M9OPfEA

The clip will be deleted from WeTransfer.com on March 8th.

I noticed that too (and unfortunately haven't the faintest idea what caused it). It's an odd sort of problem. I'd assumed it was some sort of dodgy microphone connection? But perhaps it's something more complicated.

[OldTechie]

I've been chasing a similar effect on the HLS live feed, not present on the shoutcast feed or the web browser feed on a PC. I have been blaming it on different players and I initially thought it was a problem re-assembling the segments, but I decided it sometimes happens too frequently to be the tied to the segments in HLS. So I was blaming it on network issues - and indeed I found some that were upsetting some HLS players. However, I was starting to think it was transmitted.

I was surprised when I tried to listen to the Bruckner - it suffered from the very first few notes and I decided to abandon it as too broken for listening with pleasure.

It made me start looking at the network traffic on play again in case that had changed to HDS. However it is definitely still RTMP, not HDS or HLS. The live feed through the browser is definitely HDS and does not seem to suffer with the issue.

Maybe they are creating the play-again feed from the HLS feed. I caught the end of the Bruckner last night live and immediately went to the play again feed and was surprised that it was already there so they must have changed the production methods. At the time the schedule page was also a bit broken - it showed Jazz Line-Up as live when Between the Ears was actually on air.

I've just tried this morning's James Jolly programme and it also seem to have the fault. For instance I hear one reliably on the speech after the opening Elgar Imperial March. He says "The Imperial March of eighteen ni^nety-seven..." where the ^ is a disturbance.

I wonder whether they may be trying out one of the copyright protection schemes that implants "inaudible" watermarking into the stream.

[johnb]

Intrigued by what OT posted I've just tried Listen Again with this morning's James Jolly and am currently listening to last night's Bruckner 8.

On my system (Squeezebox playing the on-demand RTMP/AAC-LC stream) there is no disturbance in the "eighteen ninety-seven" and the start of the Bruckner is playing without any glitches.

I wonder whether Triode's Squeezebox plugin picks up a different feed to that used in the BBC web player.

Andrew Scott (of the BBC) has said they are looking at DRM, but to be used for downloads rather than for streaming.

(Now about 15mins into the Bruckner - no glitches so far.)

[Bryn]

What I was referring to are not glitches in the usual sense. Although they sound a bit like individual 'plops' such a one hears in poor reception areas with DAB (sometimes known as the boiling mud effect), when one examines the waveform there is nothing particularly out of the ordinary to be seen. Have you checked out the short .wav file I liked to? If you open that in an audio editor you will see and hear what I mean. My first thought was encoding error, but on looking at the waveform I put that interpretation to one side. Note also that in the sample I uploaded to WeTransfer.com, there is a clear bias to the left. I suspect some mishaps prior to encoding.

[jayne lee wilson]

Yes, I've been hearing odd noises like this since the Rattle Beethoven cycle, on the rare occasions I've taken a relay. As you pointed out Bryn, it occurred during the Barbican Sibelius too. Muffled little thuds, knocks and bumps, sounds like "noises off" (stage, mic etc) but must be something more, I think. It was obvious again during the RSNO/Norrington Usher Hall concert last Friday, though it didn't disturb my listening too much (excellent concert too, with a wonderfully audacious Pastoral Symphony).

I wonder - is it more frequent during busier passages? I think it might be...

[johnb]

Sorry Bryn, I haven't listened to your clip. When I click on the link my browser (Firefox) reports the https:// site it goes to as having an invalid security certificate. Although it is probably OK, I decided to play safe.

However, I did listen to the whole of the first movement of the Bruckner on my system (Squeezebox + Triode's plugin) and there was nothing that sounded untoward.

I've just listened to James Jolly after the Elgar and to the start of the Bruckner, both with the BBC's own website player, and didn't hear anything that sounded at all odd. So perhaps I have duff ears (very likely) or perhaps the problem has been resolved.

[Bryn]

Sorry Bryn, I haven't listened to your clip. When I click on the link my browser (Firefox) reports the https:// site it goes to as having an invalid security certificate. Although it is probably OK, I decided to play safe.

However, I did listen to the whole of the first movement of the Bruckner on my system (Squeezebox + Triode's plugin) and there was nothing that sounded untoward.

I've just listened to James Jolly after the Elgar and to the start of the Bruckner, both with the BBC's own website player, and didn't hear anything that sounded at all odd. So perhaps I have duff ears (very likely) or perhaps the problem has been resolved.

Re. the download, very strange. The method I used was to send it to myself via WeTransfer.com. By sending it to myself I get notification of the relevant URL for the download without having to send it to anyone else's email address. I also use Firefox, and have no trouble downloading from WeTransfer.com, i.e. no "invalid security certificate" notices.

I will try accessing last night's concert again. They may indeed have sorted at least the truncation by now.

[No, LiC, Between the Ears, and Hear and Now are all still truncated. I think a few jumps through the iPlayer's "Contact Us" system is called for.]

[OldTechie]

I've been trying to describe the noises to myself without much success. Sometimes it could have been a plucked cello string at an inappropriate moment. Sometimes I think there me be a moment clipped out. I've just heard quite a loud plop about 18 minutes into the Bruckner (which I am playing through Triode's plugin on an LMS server I've run up. That is going through squeezeplay and I'm grabbing it on the way to the WASAPI interface which means it has gone through the Windows 8 mixer and been up-sampled. I'll try to compare this with Bryn's clip This may take a while. Now I've started listening I can't stop - the performance has grabbed me and the plop-rate is not too bad now its going.

Triodes plugin accesses the same server group from limelight as the web browser when I look at the network traffic. There is a difference in that the Triode plugin uses the original specification RTMP on port 1935, whereas the BBC browser player uses the same protocol wrapped in http on port 80.

I could hear my slight plop on the speech I mentioned before when playing through squeezeplay/LMS+triode.

[PJPJ]

It's a similar "plop" to one got by chopping a twentieth of a second or so out of the stream. I wondered whether it had something to do with the wiring at the Barbican but the fault seems to be elsewhere, too.

Iplayer LA has lots of issues this weekend - the team seems aware of the truncated R3 programmes, and the non-appearance of much in the last 48 hours on R4Extra.

[johnb]

Triodes plugin accesses the same server group from limelight as the web browser when I look at the network traffic. There is a difference in that the Triode plugin uses the original specification RTMP on port 1935, whereas the BBC browser player uses the same protocol wrapped in http on port 80

That's curious as when I tested which CDN the BBC web player was using for the James Jolly and the Bruckner it was Akamai in both cases (this was using a 'sniffer'). I couldn't tell which CDN Triode's plugin was accessing.

[OldTechie]

That's curious as when I tested which CDN the BBC web player was using for the James Jolly and the Bruckner it was Akamai in both cases (this was using a 'sniffer'). I couldn't tell which CDN Triode's plugin was accessing.

I was using Wireshark to watch the network traffic. Triode used fcds173.lcy.llnw.net [95.140.227.99] and the web browser used fcds249.lcy.llnw.net [95.140.227.237]. However testing again later I found triode's plugin connected to 195.59.58.143 which is an Akamai address with no DNS entry.

[OldTechie]

I've been doing more tests...

I was wrong to say that the live web feed using HDS is devoid of the plops that annoy me. I simply had not listened to it much. (I use the shoutcast stream through foobar2000 so I don't accidentally kill it by closing the browser.) Now listening I'm hearing plenty of them.

I've been checking through listen again clips. They all seem to suffer the fault starting with 01:00 Through the Night on 14th February. The previous item (World on 3, starting 23:00 13th Feb and finishing 01:00 14th Feb) is perfect and so are all clips previous to that point that I have checked.

I guess this fault is on all items process by the Audio Factory. I think it is difficult to hear against a high audio level, so it is Radio 3 that is particularly prone to an audible fault.

[Bryn]

I'm away from my main storage at the moment, so can't immediately check, but was that perhaps when the sample rate was changed from 44.1kHz to 48kHz? Com to think of it, I think I mentioned the date of the change somewhere here.

[Ah, no, that was a couple of weeks before the disruption appeared.]

[OldTechie]

Now I'm hearing the disturbances on the shoutcast feed, which has changed from 44.1kHz to 48kHz. If I switch over to the HLS feed (which is delayed by more) I hear the same disturbance at the same point in the music.

Later... Maybe I was mistaken. This evening's concert from Glasgow seemed fine on the shoutcast feed, but switching to HLS I did hear the odd disturbance. Perhaps the fault I heard earlier was just on the source material and would have been present even on FM.