Nimbus Records on Facebook Nimbus Records on Twitter Nimbus Records on YouTube

Newsletter

 

Nimbus DVD Music

Nimbus Records has pioneered Surround Sound since 1972. All Nimbus recordings have been made using Ambisonics, a system we prefer because it can capture the acoustic quality of each hall and recreate it in a domestic environment with great fidelity. Until now users have needed a specific Ambisonic decoder to be able to enjoy Nimbus recordings the way they were intended. As only a handful of manufacturers have made these, the full effect of our recordings has only been available to a dedicated minority of listeners.

Now, DVD makes it possible to deliver Nimbus Surround Sound recordings to home listeners with standard, readily available equipment - all that is required is a DVD Video player connected to a DTS capable Surround Sound system, or a DVD Audio player with multi-channel playback. We believe that Surround Sound will enhance your listening experience more dramatically than all other technical innovations combined.

What's On the Disc

The Surround yourself with ... series of discs present up to 90 minutes of music (up to 180 minutes on double sided discs) in three different formats. You should choose the format that is appropriate for your home playback system.

Stream 1 - 2 channel, UHJ encoded, PCM, 48 kHz, 24 bit (DVD-Video players)

This stream is the most appropriate if you only have two channel playback or if you have your own Ambisonic decoder set up for your specific playback system. The audio quality is a close equivalent to what you would hear from a standard CD.

Stream 2 - 4 channel, DTS compressed, 48 kHz, 24 bit (DVD-Video players)

This stream is the one to use if you have a DVD-Video player connected to a surround sound amplifier with a DTS decoder. DTS compression is a lossy compression system (see paragraph below), but does give very good audio performance. DTS decoders are built into the vast majority of domestic surround sound amplifiers now available

For more information on DTS please try http://www.dtsonline.com/

Stream 3 - 4 channel, MLP losslessly compressed (see paragraph below), 88.2 kHz, 20 bit (DVD-Audio players)

This stream is playable only by DVD-Audio players. All DVD-Audio players are capable of decoding MLP data streams. The four channels will be delivered from the analogue multi-channel outputs of your DVD-Audio player - these must be connected to the multi-channel inputs on your surround sound amplifier.

For more information on MLP please try http://www.meridian-audio.com/p_mlp_in.htm

Audio Compression Systems

Audio compression systems fall into two categories, referred to as lossless and lossy. When using a lossless compression system, the audio data you get from the decoded stream is identical to the information you put in - the compression reduces the size of the data without any impact on the data itself. When using a lossy compression system, the size of the data can be reduced further but the data you get out is not identical to the original. Different lossy compression systems offer different compromises between data size reduction and data accuracy.

The Speaker Arrangement

The four channels delivered by both the DTS and MLP streams are four speaker feeds - front left and right, back (or surround) left and right. When decoding Ambisonic material it is possible to decode for a specific number and layout of speakers - these four channels were decoded assuming four matched speakers arranged in a square with the listener(s) in the centre. While this arrangement may not match exactly many domestic playback systems, after extensive tests we found it to be the most flexible target layout. It works well for systems where the listener is closer to the back speakers and this has been compensated for with gain adjustments on the amplifier. It also works particularly well for the standard ITU speaker layout, which is most likely to be recommended in surround sound amplifier manuals.

The decision was made to decode for a target layout of four rather than five speakers for two reasons:

For a centre speaker to work well as part of a music playback system it must be well matched to the other speakers in terms of its sonic character and its relative level must be set correctly. The majority of domestic surround sound playback systems are used for movie playback, where the centre channel is used primarily for on-screen dialogue. For movies it is much less critical if the centre speaker does not match the others, or if the level is set incorrectly. It is likely that many of the centre speakers in domestic playback systems are not appropriate for music reproduction for these reasons. The conclusion was that excluding the centre speaker would reduce the danger of bad surround playback. Ambisonics is fundamentally best decoded for regular arrangements - that is to say four speakers in a square, five in a pentagon, six in a hexagon etc. The centre speaker in the standard five speaker layout biases the sound towards the front, making Ambisonic decoding more difficult and requiring more compromises than for the four speaker arrangement we have chosen.

For these reasons the front speaker is intentionally silent.

Similarly, there is no LFE (low frequency effects, normally sent to the system subwoofer) channel data. The reason for this is more straightforward. There is no low frequency information in classical music recordings which cannot be reproduced satisfactorily by a normal, full range speaker - even the cannon in Tchaikovsky's 1812 Overture. If you have a system with smaller speakers and a subwoofer, then your surround sound amplifier's bass management system will automatically redistribute the low frequencies that your main speakers cannot handle to the subwoofer. There is no need to include any of this information in the dedicated LFE channel on the disc.

Ambisonics

The Ambisonic system used by Nimbus Records was developed in the early 1970s. It comprises methods for recording and reproducing Surround Sound with great fidelity, and remains unchallenged as the most coherent and effective Surround Sound system.

Ambisonics can be helpfully divided into four areas - microphone, post-production, transmission, replay.

In its simplest form an Ambisonic Surround Sound recording is made using a Soundfield microphone - this is a microphone with four cardioid capsules in a tetrahedral arrangement. The outputs of these four capsules can be processed to give a B-format signal. (The raw output of these four capsules is sometimes referred to as A-format). B-format is the core of the Ambisonic system. It consists of four channels best described in terms of four coincident microphones - three figure-of-eight patterns each aligned with one of the three axes (channels X, Y and Z) of a three dimensional graph, and one omnidirectional (channel W). These four signals together describe a full, three dimensional soundfield.

Nimbus uses a slightly different microphone arrangement developed by our Reseach Director, Dr. Jonathan Halliday, which consists of two figure-of-eight microphones, one directed along the front / back axis, the other along the left / right axis, and an omnidirectional microphone, all held togther as closely as possible in a frame. The result from this arrangement is single plane B-format, comprising channels X, Y and W, without the Z (height) information.

The post-production part of the sequence, although not relevant to the way Nimbus makes recordings, allows multitrack recordings to be mixed to produce an Ambisonic result. There is both analogue equipment and now plugins for various audio processing software that allow the mixing engineer to manipulate sources recorded in more conventional ways, either mono or stereo, as part of a B-format soundfield. These include tools to select location and movement within the soundfield in three dimensions.

The finished, Ambisonic recording can be delivered to the user in a number of ways. B-format itself is a very efficient transmission format, conveying full three dimensional sound in only four channels. For single plane recordings, three channel B-format can be used. It is also possible to decode the B-format to a common, regular speaker arrangement for transmission as speaker feeds. The most basic method for transmission of an Ambisonic signal uses a system called UHJ. UHJ is a matrixing system developed to allow Ambisonic material to be transmitted via only two channel media. The two channels are stereo compatible - that is to say you can listen to them as a normal stereo recording on standard equipment - but with a UHJ decoder you can decode them to your chosen multichannel speaker layout.

An important aspect of Ambisonic transmission and replay is that nothing in the production chain to this point has made any assumptions about the number or arrangement of playback speakers. Ambisonic material can be decoded to any number of speakers. Decoding works best for regular speaker arrangements, as mentioned above. Decoding for irregular layouts is possible, but it becomes more difficult to recreate a convincing soundfield. Even using speaker feeds as a method of transmission, if the target arrangement of speakers is known a decoder can reconstruct the original B-format from the feeds and then redecode to any other chosen speaker layout.

For much more information on Ambisonics please try http://www.ambisonic.net

The Production Process

When Nimbus started to make digital recordings in around 1980 until the late 1990s all of our master tapes were two channel, UHJ encoded, 44.1 kHz, 16 bit - exactly the same format as we deliver on CD. Before that we recorded to four track analogue tape, in either A-format or B-format. More recent recordings have been made using three or four channel magneto-optical discs in B-format, 48 kHz, 20 bit. The significance of all this is that each of these master formats requires different treatment to derive the four speaker feeds on the DVDs. These DVDs were all taken from two channel, UHJ, 44.1 kHz, 16 bit masters.

The process of establishing exactly how best to treat these various master formats was started some time ago, at the prompting of Richard Lowe (Periphonix), a long time supporter of Ambisonics. His initial work creating test material and his collaboration during the development of the production process has been a crucial part of the project.

The material on the DVD-Video part of the disc must be presented at 48 kHz - this is part of the DVD-Video specification - so the material was sample rate converted for those streams using a software solution. The Ambisonic decoding was undertaken using software written specifically for this project and supplied by Meridian Audio.

The DVD-Video streams are presented at 24 bits. It can be demonstrated that correctly dithered 16 bit audio contains information well below its expected noise floor. Analysis of our earliest digital recordings indicates that there is at least 18 bits worth of information available, and later recordings made using more sophisticated A to D converters have even more. The DVD-Video specification allows for digital audio streams of 16, 20 or 24 bits. The 20 bit option would give enough capacity for the data available in the recordings, however support for 20 bit audio files is not implemented correctly in the majority of DVD Authoring systems, so the decision was taken to use a 24 bit stream, even though the bottom four or more bits may have no useful data. 20 bits is used for the DVD-Audio stream, as the MLP encoder does handle 20 bit streams correctly.

The DVD-Audio stream is presented at 88.2 kHz. While there is little obvious benefit to be had from such a high sampling frequency when the original recording was made at 44.1 kHz, it has been observed that D to A converters sound better when working faster. The choice of this sampling frequency has no impact on the total playing time of the disc, as the MLP lossless compression system is very good at not encoding information that is not there - a stream at 88.2 kHz, 20 bit compresses more efficiently than the same material at 48 kHz, 24 bit. 88.2 kHz was chosen over 96 kHz as it is a simple multiple of the original 44.1 kHz, and therefore avoids the complex 44.1 to 48 kHz sample rate conversion step.