head.gif - 13Kb

Like many audio enthusiasts I often listen to music on the radio and TV. I enjoy events like the ‘Proms’ on the BBC, and for this a good FM tuner has been a vital part of the audio systems I’ve used over the years. However I then bought a ‘Digital Terrestrial Television Receiver’ for use with the TV set and AV system in the living room...

Having enjoyed many Proms on BBC4 TV during 2004, it then occurred to me that the 2005 season would present an interesting opportunity for comparisons. The sound quality on BBC TV4 via DTTV seemed quite good to me, and lacked the background noise level of R3 FM. This set me wondering about two questions: “What are the relevant levels of audio performance of these broadcast ‘routes’?”, and “Would we be better off listening to such music on BBC TV4 via DTTV (or R3 DTTV) rather than R3 on FM or DAB?”

fig1.gif - 23Kb

Figure 1 shows that the various digital formats which are used for UK terrestrial TV and Radio broadcasts by the BBC. (I am ignoring satellite or cable TV simply because I have no experience with them.)

In practice, what we can receive at home will be a digital stream of audio (perhaps combined with video!) via the DTTV, DAB, or NICAM on PAL/analogue TV. Analogue TV also provides a monophonic analogue sound channel which is derived from the NICAM at the transmitter. Similarly, FM broadcasts are derived from NICAM.

Using the above we can draw up the following table of nominal ‘raw’ performances. Note that these values were correct at the time I wrote this article. But some of them may change in the future, and this affect the situation.

System Audio Bandwidth Dynamic Range Bitrate (kb/s) Bitrate Ratio
DTTV BBC4 TV 24 kHz 95 dB 256 0·166
DTTV R3 24 kHz 95 dB 192 0·125
DAB R3 24 kHz 95 dB 160 0·104
NICAM (PAL TV) 16 kHz 85 dB / 60 dB 896/646 0·583 / 0·421
NICAM derived FM 15 kHz < 75 dB / 60 dB 896/646 0·583 / 0·421

The values in the above table should not be taken too literally as they ignore various practical considerations. For example, what we receive when using an FM tuner will tend to be limited by the reception conditions. However the values do provide a simple starting point for considering what we might expect from the various systems.

The ‘Bitrate Ratio’ values in the right-hand column of the table simply show how the audio bitrates compare with the original value of 1532 kb/s emerging from the mixing desk at the hall. Ideally, if we were comparing similar systems throughout, it would be reasonable to expect the system with the highest value to be capable of the ‘best’ results. However we can only make comparisons with some caution because we are not always comparing ‘like with like’ throughout.

There is also a complication which I will ignore in what follows: Although BBC TV have essentially standardised on 48 ksample/sec for their audio, BBC R3 have internally standardised on 44·1 ksamples/sec – despite their DAB and DTTV output being at 48 ksamples/sec. This means that when signals are transmitted on DAB or DTTV by R3 they may have undergone some ‘additional’ rate-conversions which were not applied for DTTV BBC4 TV! It is hard to be certain of the situation here as I have not been able to get a clear explanation on this point from the BBC. In theory, any such rate conversions should be ‘audibly transparent’ for frequencies up to 20kHz, but alas practice does not always match theory. Here I will have to simply warn that differences may arise as a result, and if so, might be a factor that causes differences in the audible results.

Looking at the above values leads to an interesting first impression. This is that the DTTV and DAB systems nominally should be able to offer a wider frequency response and a greater dynamic range than either PAL/NICAM TV or FM. What is particularly interesting here, is that of these, the sound for BBC4 TV4 might be superior to R3 on either DAB or DTTV. This is because the sound for the BBC TV broadcasts uses MPEG-2 with a bitrate of 256 kb/sec which higher than the bitrate used for R3 on either DTTV or DAB. Thus we have the interesting implication that when we want to enjoy listening to a simulcast we may be best advised to listen to BBC TV, and not R3! This implication is an intriguing one, and combined with my subjective impressions when listening that the sound on DTTV BBC TV4 could be very good, prompted me to investigate further...

For the sake of example, here I will use one section of one concert. This is a performance of the Mozart Violin Concerto that was featured in the Prom given on the 22nd of July[1] and repeated on Radio 3 on the 26th of July. I recorded this in three ways:

I have been analysing the above recordings (and others) in various ways, but one of the simplest and most direct methods showed some interesting results.

fig2.gif - 22Kb

The input signals were divided into a series of 100 millisecond long ‘chunks’ and the peak power level on each channel during each chunk was noted. Then the number of times any given peak power level occurred was worked out. The black lines of the plots show what percentage of the chunks had a given peak power level. The plots are probability distribution histograms with a bin resolution of 1dB. The recordings from DTTV were via an SPDIF transfer, but the recordings from FM were via an ADC in a Pioneer PDR-509 recorder.

Taking the left hand channel of Figure 2 this shows for example that the highest peak power level that occurred at any point during the analysed recording was at about -9dB (referred to the nominal max recordable level). The most ‘common’ peak power level was at around -28dB. This tells us that the typical level was around 20 dB below the loudest instant in the recording. The right hand results can be seen to be fairly similar, as we’d tend to expect.

The red and blue lines in the Figure were obtained in a similar way. However the signals were filtered before the chunks were taken and had their peak level statistics collected. The Red lines are for the LF (low frequency) or bass part of the signal. The Blue lines are for the HF (high frequency) or treble part of the signal. The filters used were equivalent to 4th order Linkwitz-Riley with a crossover at 1 kHz, so mimic what we might get if filters of this type were used in a loudspeaker crossover system.

fig3.gif - 22Kb

Figure 3 shows the results of applying the same process to the recording of the same performance, but this time captured from the FM rebroadcast. (The tuner used was a Quad FM4.)

fig4.gif - 22Kb

Comparing Figures 2 and 3 we can easily see that the power distributions of the FM and BBC4 TV versions differ. However before commenting on that in detail we can also consider the plots in Figure 4. These show the results for the Radio 3 rebroadcast on DTTV at the same time as the FM version. To make comparison easier I have replotted the results in Figure 5. This just shows the unfiltered results, but with the three examples – FM, DTTV R3, and DTTV BBC4 – all on the same graphs. To make the comparison clearer the distributions have been gain adjusted to the same maximum peak levels. This corrects for any differences in the gains settings when the recordings were made.

fig5.gif - 23Kb

The above shows that that the DTTV R3 and BBC4 TV versions are very similar. They both have a a most common level (i.e. highest percentages of the time) which are around 20dB below the maximum peak level reached at any instant during the recordings, and the shapes of their curves match very well. Whereas the FM version looks like a level compression has been applied which tends to lift the typical sound levels by about 8dB. In effect, the FM distribution has been ‘squashed upwards’. This has the effect of reducing the dynamic range and causing loud passages to have less impact, In my experience another, more subtle effect, can be noticed on performances like solo piano recitals.

Percussive sounds like the notes from a piano consist of an initial ‘transient’ followed by a sustained tone that fades away slowly. The level compression can have the effect of bringing up the level of the sustained portion of the notes when compared with the brief peak levels of the transients. This may seem to give added ‘warmth’ to the sound, and extend the period the notes remain audible. The effect is a little like a subtle version of a ‘sustain’ effect pedal as used by electric guitarists.

In addition to the plots shown in Figure 5 I did a comparison for the filtered HF and LF sections, and the results were very similar to Figure 5. At first glance this might be taken to imply that the level compression should not alter the perceived tonal balance. However altering the transient/sustained relative levels might do so.

The results shown above make it clear that when comparing FM with DTTV we may not be comparing like with like. The sounds on FM may have been level compressed in a way that the sounds on DTTV have not. Now it may be significant here that the FM recording used for the above results was taken from a mid-day repeat broadcast. The BBC may well only be employing the level compression at some times in order to avoid the sound levels becoming awkwardly low. However the DTTV R3 broadcast used for the above was the same mid-day repeat, and did not have the level compression. Hence the implication is that if you wish to avoid the sound being tampered with in this way, then you might prefer DTTV R3 to FM.

From the above results we can draw an interesting conclusion. If we accept that – when data reduction is used – that the best results can be expected with the least amount of reduction, then the implication is that the sound for DTTV TV should be better than DTTV R3! This is because the DTTV TV broadcasts allocate bitrates of 256k/sec for audio, whereas DTTV R3 only gets 192k/sec. Of course, the same argument also implies that DTTV R3 should be better than DAB R3

One feature of the results which is particularly intriguing is that it seems to clash with a common assumption. Like many audio enthusiasts I tend to expect TV sound to be mixed to ‘spotlight’ what is in front of the camera, and to be level compressed to accomodate TV viewers who might not want the full dynamic range of an orchestra. However there is no sign of this in Fig 5! This said, I do have the impression that BBC1/2 TV do tend to level compress or spotlight the sound in a way that BBC4/R3 do not – although at present I have no reliable data to test this impression.

Given these results, perhaps there are good technical reasons for the impression I have gained over the last year or two that the quality of the sound on Proms via DTTV BBC4 can be as good as or better than on R3 – either on FM or DTTV. The result may not be like this for other stations. However, if you haven’t tried it already, then DTTV may offer better sound quality than you might expect. So perhaps you should give it a try.

Jim Lesurf
18th Aug 2006

[1]   Philharmonia Orchestra. Conductor - Christoph von Dohnanyi Violin - Christian Tetzlaff

prev.gif - 2352 bytes  ambut.gif - 3891 bytes  next.gif - 2248 bytes