Peal in two places
Unlike most wacky ideas, I was not in the pub when this one presented itself. A work conference in April included a presentation entitled ‘LOLA - LOw LAtency Audio Visual Streaming System’ and began with a live performance of Beethoven’s Sonata in A Major Op. 69. However, only the cello player was in the room, the pianist was 250 miles away in Edinburgh, which includes the computer setup needed). Within a few seconds I knew a use for this.
The LOLA project had started in 2005 and the first UK performance was in September 2012, so was very new. Retiring for the evening, I contacted ringers at home to find others suitably excited, so I sought out one of the presenters the next day.
Gill Davies is a researcher in distributed music over high speed research networks at Edinburgh Napier University. After a discussion, Gill could find time in July to help us. There were only three systems capable of running LOLA in the UK and as one was at Edinburgh Napier, so the plan was to find a couple of Scottish ringers to participate. Initial enquiries for a vague 'handbell peal with a difference' met with a positive response, but we could not find a suitable date. Then my wife reminded me I'd be in Scotland for the Cumberland Country Meeting and that I should ring. I enquired with janet (the provider of the educational computer network in the UK) to see if one of the other two system could be booked to come to Derby the week before we rang. Emma Smith confirmed the kit was available and so the date was set.
Far more non-ringers knew about the peal than ringers. My concern was that, once other ringers knew it was possible, the race would be on to ring the first peal!
Not being a very accomplished handbell ringer, I thought that if I could use LOLA, then most other people could. Two enthusiastic people in Derby were found, but I still needed one more for Edinburgh. Ideally a conductor in both locations would be best and then I realised that Emma Southerington would be at the meeting - sorted!
Next we had to decide which pairs to ring, where they should be and which set of bells to ring. We thought it best to have a 1-4 and 5-8 split. Now, which bells? We had access to the Derby D A set that once belonged to Sir A P Heywood (size 15). These have round clappers and are quite loud. I also have a new ‘pocket’ set of Whitechapel bells, size 10, which are much quieter. I decided to split and take both sets. It’s a long way to go to find the bells were not suitable.
Having used Hawkear for the 12 bell competition over the past few years, we knew that errors of 50 milliseconds are easily detectable by the human ear. We needed to be well below this to make the system work. LOLA promised, with the right setup, delays of as low as 14ms between Derby and Edinburgh. A 5ms delay is introduced by the software, the rest is dictated by the network equipment and distance between the 2 locations. The system has heavy network requirements. Most people have ‘up to 8Mbps’ to their home. A growing number have up to 80Mbps. This system, using colour at 60 frames a second mode needs over 500Mbps. To further speed up the system in Derby, it was plugged directly into the Internet, bypassing firewalls and other systems that protect the innocent user from the ravages of the Internet.
The last consideration was around the room type. Ringing handbells in the bathroom is completely different to the lounge. Fortunately we ended up in music studios in both locations, so the bells sounded fine.
On the 5th of July, the system arrived in Derby. To cut a long story short, I left for Scotland on Wednesday without testing the connection to Edinburgh. The technical details are below.
On the day of the peal, once the Derby contingent had got into the building, we tried both sets of bells and decided the quieter ones were best. The delay was 24ms. The sound was ace! It sounded like 4 ringers in one room ringing 8 handbells ringing Plain Bob, we could hear Simon conducting as if he was in the same room. Initially there was a perceived gap between the locations and the first attempt at plain hunt took some getting used to. Something similar to perceptual adaptation took over during the ringing and our brains recalibrated to take account of delay, see Science Club - playing tricks on the mind. Listening to the youtube video of the end of the peal, ( https://goo.gl/tzdY2S ) the delays didn’t sound so obvious, though, which rather points to them being a subjective effect.
Society of Royal Cumberland Youths Derby & Edinburgh, England & Scotland University of Derby & Edinburgh Napier University Sunday, 14 July 2013 in 2h29 (10) 5024 Plain Bob Major Composed by Simon Humphrey 1–2 Anthony P Cotton 3–4 Emma J Southerington 5–6 Simon Humphrey (C) 7–8 Christian M Peckham
Ringers of 1-2 & 3-4 were in Edinburgh, 5-6 & 7-8 were in Derby.
With thanks to:
Gill Davies, Paul Ferguson and Niall Sinclair from Edinburgh Napier University
Claudio Allocchio and the lola project team
Emma Smith from Janet
Roger Browne from EMMAN
Mark Randell and Tony Iveson from University of Derby
This gives you an idea of the setup in Derby, 5-6 in the foreground:
Ringing in Derby, 5-6 in the foreground
Count down to the peal:
T-12 days - Visit the studio, which is a 10 minute bus ride away, to configure and test the Internet connection, which doesn't work. They must have configured it wrong. Luckily they had which was easy to rectify.
T-11 days - Visit studio again and the Internet is working. Using speedtest.net, I’m getting 760Mbps. Happy with that. This later improved to 876Mpbs.
T-9 days - The equipment arrives in 2 large crates and I set it up next to my desk. Testing went well until I started the software, which brought an error. HELP! Although the equipment was tested before shipping, it was not working. Using Skype, I was able to contact the developer in Italy, however the system had no microphone, so we ended up talking on the phone! By home time, one issue had been resolved, but another appeared. The camera did not display an image. Give up and go home...
T-8 days - Pop into work on Saturday, plug everything back together again, still not working. Plug in spare camera, not working. Plug back the original camera and switch on - it works! Time to go home, before anything else breaks!
T-6 days - Transfer the system to the studio on the bus. As I get off the bus, a reversing car bumps into the crate and knocks it off the trolley - fortunately it is well padded. Set up the equipment with Mark Randall, audio technician from the University, who provided microphones and speakers. First proper test to Italy. It’s all working. Simon Humphrey arrives to see the installation. However, all is not well. The video in Italy is stuttering. It’s fine in Derby... Many hours later, come up with two things to try - swap the PC to see if this is the fault and test to Napier to see if this works.
T-5 days - I take my work PC apart, which was almost identical to the LOLA system, and swap the cards over. Switch on and it works. Connect to Italy and it works for a while, but then fails with the same issue as the day before. Unfortunately, testing with Edinburgh is not possible as they are moving buildings and Paul Ferguson is on holiday.
Later in the evening, Claudio in Italy confirms that he has been looking at the data they have been getting over the network. It transpires that he was getting all the right packets, not necessarily in the right order. I look at the network monitoring and find that there is a large amount of errors (around 1 million!) between our sites in Derby. Network fibre has a pair of cables, one for sending and one for receiving. The errors were on the fibre that sends the traffic to Italy. The graph shows a blue line to Italy, and a flat green line from Italy.
T-4 days - Replacing the fibre is not an option without taking down the University internet connection - I would not be popular! The resilient link between our sites was in the process of being setup, so that was not an option. Around 10am, EMMAN suggested using the University's resilient link. This has a longer path via Mansfield and Lincoln to Nottingham (the primary route goes direct to Nottingham) which will introduce more latency (delay). Let’s try it! By 11.30am it was working and there was no stuttering in Italy! As I would not be in Derby for the peal, I set up the room with the speakers, projector, etc and left for Scotland without doing a test to Napier - fingers crossed!
Unfinished communications cabinet |
Network traffic during the peal |
T-3 days - Finish ringing a peal to find that Paul Ferguson from Napier had returned from holiday and tried to contact me. Left a message with him to see if I could visit tomorrow.
T-2 days - After a peal in the morning, I had the afternoon off, so visited Napier to try a test with Mark Randell in Derby. After an initial problem with the sound (which meant communicating via written messages held up to the camera as there was no mobile phone signal), all was working - we’re on! As the picture shows, the facility was still being put together. Our computer was the only one plugged in!
T-1 day - I though I'd better check with the SRCY officers that they were happy with us ringing it for the Cumberlands.
D Day arrived. At the Edinburgh end, Paul, Gill and Niall Sinclair (music student) where there to help us. Gill and Niall sat through the whole peal. At the allotted time, there was no sign of anyone from Derby. It transpired that the security team had not got the message to let them in the building! Once they were in, again we had sound issues, which we were able to resolve quickly this time.
Emma, in Edinburgh, takes up the story:
The equipment:
- we had microphones right by our bells - mine was about waist height (when sitting) and a tiny bit to the left.
- there was a speaker where 5-6 would 'sit' and a speaker where 7-8 would 'sit'
- in between the speakers was the largest TV I have ever seen - on a very low hanging stand, so that the visual effect was that we were sitting at the same level and we were about the right size too, but a bit more squashed - the camera adds at least 10lbs width-ways!
- there was a camera fractionally above the TV filming us
The sound:
- there wasn't any detectable delay if you ring by ears: when I closed my eyes to ring, we could have been in the same room
- the sound was ace. it sounded like 4 ringers in one room ringing 8 handbells ringing plain bob! :) we were in a brand new music centre at Edinburgh Napier. It's not open to the Students til October; and so all the equipment, computers and speakers and things were pretty whizzy and shiny.
- Anthony brought two 1,2,3,4s to Edinburgh and the Derby lot had two 5,6,7,8s: we tried both sets of bells, and the lighter set which had dampeners on the clapper pegs produced a clearer sound
- we could hear Simon (c)-ing as if he was in the same room
The visuals:
- there was a tiny tiny fractional delay on the visual side of things: when you look to pick out coursing order, the bells didn't move when you expect them to: the smallest tiniest delay is seen, a bit like ringing with a competent but nervous less experienced ringer who's over-thinking
- we started slowly: 2h52 speed for the first quarter. Everything was easier when we sped up to 2h14 speed: I didn't particularly notice the visual delay when we were ringing quicker.
- there wasn't really a limiting factor for speed: we started slowly as this was a new experience. If four people are sitting in a circle, there is the vertical movement of handbells, but also an amount of horizontal movement - ie the other ringers' bells move out towards the middle of the circle. on a screen, this "outwards" movement isn't there, and so its a bit like ringing with a very tidy ringer. When ringers have quite a flamboyant visual style, its just a bit different to what you would normally expect, and so we started very very steadily.
- the cameras were set so that if Simon and I grinned at each other in the 3456 rollups, we could catch each other's eyes, as if he was sitting next to me.
Other stuff:
- the only "speed worry" I thought we had, was whether the connection would last perfectly in both audio and visual for the duration of the peal: it did! (I was also a bit worried that the 2 research students in Edinburgh might just say "HOW much longer?! I thought this was only going to be a couple of hours..." (they didn't).
- I asked about ringing over distances (ie ringing with people in India or the USA) and the latency increases substantially; I can't remember what the audio detection limit is (40ms?) or the latency drop for distance; but its so many milliseconds per km. In basic-for-Emma-terms: once you get to Europe distance, then things are OK, but above Europe distance, then things aren't (sorry to bring bad news!) and an audio delay is detectable. We had 19ms latency at the start; and so (other than the musical notes) the sound from Anthony's pair were indistinguishable from Christian's and Simon's.
Simon's account from Derby:
Subjectively at least my experience was a little different from the Edinburgh accounts.
I’ve not been in a professional recording studio before. Very impressive.
The sound was a bit of a problem, Edinburgh initially said they could hardly hear us. We turned up the gain on our microphones but then got audio feedback and had to turn down the gain on our speakers. Then we could hardly hear them! After a bit of fiddling about with the gains we did manage to end up being able to hear them, albeit rather quietly, with Edinburgh able to hear us well. But better that way round than the other, since they would need to hear the bobs being called.
I thought the delay in the Edinburgh bell strikes were quite perceptible, to the extent I thought that we might have to give up, at least to begin with. The video action looked spot on but there seemed to be a definite time lag in their bells. It felt as though they were deliberately trying to slow down, as I’ve witnessed before (and probably been guilty of) when brains aren’t keeping up with the speed of ringing, which I can safely say was not the case in this instance!
We rang very slowly to begin with which I found helped a lot. Maybe when the delay was less than some proportion of the average inter-bell gap it became less apparent. However, we believed we had a deadline of 5pm to get out of the Derby building, or risk being locked in till Monday, so I several times exhorted the band to ring faster. From my perspective when we did eventually speed up the striking sounded worse, probably because the delays became more noticeable, though eventually we did manage to find a compromise speed. And we weren't locked in afterwards.
Listening to the youtube video the delays don’t sound so obvious, though, which rather points to them being a subjective effect – real to me at the time, nonetheless!
T+1 - pack away all the equipment in Derby: