Why does it sound different on the Radio...

[BuffaloGhost]

Sounds like you have a tonne more experience than I


Sent from my iPhone using Tapatalk

[PJSprog]

All solid points. Unfortunately, there is also the opposite end of that spectrum, which I experienced recently while listening to some music via Sirius/XM Satellite radio in my car. Some of the classic rock stations have so overly compressed their signals that it ruins the music. Specifically, one station played Rush' Hemispheres (which is a rarity to begin with) about a week ago. The downbeat szforzandos after any quiet or buildup sections were squished so much that the music became unintelligible and indistinguishable mush. Hemispheres is my all-time favorite rock album, but this sounded so bad I had to change the channel.

There are extremes at both ends, and neither are good. Finding that balance is the aforementioned "dark art."

[Simon Barden]

For the uninitiated, traditional VU meters on analogue desks were never particularly accurate, plus the 0VU point could be set up differently on each desk, depending on the equipment it was connected to. Different makes of meter reacted differently to complex sounds and were only really accurate when using a sine wave. But 0VU was always set up to allow reasonable headroom for loud peaks, and when recording to tape, the gentle compression characteristics of the tape when levels beyond the nominal maximum for clean recording were sent meant that you could generally drive the VU meter into the red to get some nice tape compression, saturation and extra harmonics.

Never confuse 0VU with 0dBFS (dB full scale) when recording digitally. 0dBFS has no headroom above it, whereas 0VU did. Whilst the (normally floating point) maths inside your DAW can cope with really ridiculous virtual overloads, there is still no reason to use very hot signals on channels, and certainly not on the master bus. Most plug-ins modelled on analogue hardware use similar relative input levels for the start of distortion as the hardware had, so unless you positively want to drive a plug-in hard for effect, then you are generally best keeping the input levels into the plug-ins at around -12dBFS or less. There are some plug-ins I know of that do require hotter signals to work at their best, especially I find de-essers need a hot signal, but it's always best to find some way to reduce the signal level before it leaves the channel and hits the master bus or group bus.

Whilst your DAW will happily cope with signals above 0dBFS internally, the moment the signal leaves the DAW software from the master output and hits the D/A converter, then 0dBFS is the limit, and any signal above 0dBFS will get truncated to 0dBFS, and then you get digital clipping. Even if the loudest signal is at 0dB, you can still get inter-sample clipping in the D/A converter, if the 0dBFS value (or very near 0dBFS) has another after it and represents the peak of a rapidly rising and falling waveform.

The standard representation of digital audio waveforms as a series of stepped values is very wrong. What each sample value represents is a single point in time, and you can join them up with a line representing the analogue waveform to get the true picture of what goes on. And this is exactly what the D/A converter does - it doesn't output stepped values, it outputs a continuously varying voltage reconstructing the true waveform. But if there's a virtual peak that's higher than 0dBFS, then it simply can't output more than its maximum voltage level, and so the analogue signal is clipped at that point.

Consider a few consecutive sample values representing a signal peak that are at -32,-18, -2, -3, -19 and -24dBFS. The peak of the signal will be between the -2 and -3dBFS values, but whilst the maximum sample value is showing -2dBFS on the master meter, the inter-sample peak will be at say +3dBFS. So this is beyond what the D/A converter can produce, so the tip of the sharp curve will be flattened at 0dBFS for a short period. This clipping has no harmonic relationship to the original signal, so it sounds harsh, rather than the generally nice and warm and much slower analogue 'clipping' you'd get from slightly overdriven tape. It makes long-term listening fatiguing (along with very limited dynamic range introduced by the 'loudness wars' of past years).

So always use a meter on the master bus that has a true peak indication (which has software that mimics the D/A converter to work out what the real signal peak value is), and make sure that never reaches 0dBFS. I limit my maximum signal level to -1dBFS on the true peak meter just to be sure, when mastering, but when mixing down before mastering, I normally work to -18dB to allow room for EQ and compression at the mastering stage.

Digital radio normally has to operate in a very limited bandwidth, just because there are so many more calls for digital communications, mainly phone and TV broadcasting. The number of digital radio stations far outweighs the available frequencies available for good quality transmission, so digital radio normally has to be broadcast with a very reduced bit-depth and so a requires the content to have a very squashed dynamic range as a result. It's often far worse quality than AM used to be, and certainly nothing like a good quality FM broadcast can be like. In the UK, live classical broadcasts were often sent out on FM at CD-quality.

Most broadcast audio material is normally produced to fairly recent Loudness Unit Metering level requirements set by the broadcasters/film companies etc. which measure dynamic range and overall loudness over short, medium and averaged long term time periods (the long term one has a level threshold below which the level is excluded to stop silences or near silences dragging down the overall average loudness value significantly). Current versions of DAWs normally have these LUFS scales and meters built-in, so it's worth reading up on how to use them.

The majority of internet streaming services for audio and video are now processing the material so that all the sound is played back at a reasonably constant level, so that audio at a reasonable average level and with a good dynamic range doesn't sound significantly quieter than audio compressed and limited within an inch of its life (as certainly used to happen in the 2000s for most commercial material).

Our ears normally perceive 'louder' as 'better', and the loudness wars really started once vinyl stopped being sold in stores, as all-digital medium meant that the physical limits on the loudness that could be put on a record without the needle constantly jumping out of the groove no longer applied.

In fact, the methods now used by the content providers normally means that a very 'loud' track will be played back at a much lower level than a track with good dynamics and headroom.

There are plug-ins that can help you to optimise the level and dynamics of your track for playback on different content providers' systems. The Meterplugs Dynameter (with input from mastering engineer Ian Shepherd) is a good one for this (though I'd probably wait for a sale before buying it). https://www.meterplugs.com/dynameter

But they also offer a free 'check a track' service, which tells you (within a small margin of error as they don't have the exact algorithms used by those companies but can work backwards from known examples) what level your track will be played at on a number of platforms. Herre's the website: https://www.loudnesspenalty.com/

[BuffaloGhost]

This is all amazing and valuable advice. Thank you so much.


Sent from my iPhone using Tapatalk

[Simon Barden]

Editing and mixing should be done at levels you can hold a comfortable conversation over

I'd suggest listening a bit louder than that at times. The louder the signal, the more sensitive the ear is to pitch detection and level imbalance. If you have it too quiet, you can miss things that come to light when played back louder. You don't need to do it all the time, but at least one loud playback is worth it before signing off on a mix.

[Simon Barden]

Which sometimes is my downfall. I use my ears more than looking at the numbers.

Which is what you should really be doing most of the time. Some people spent too much time looking at the screen and not listening. As long as you aren't peaking too high on the meters, it's often better to shut your eyes and just listen and adjust as necessary. If it sounds right, it is right; the general public aren't going to listen to it whilst looking at a DAW EQ or compression page.

The Focusrite LiquidMix (a really nice piece of hardware that became obsolete because Focusrite couldn't write 64-bit Firewire drivers for it) that emulated classic hardware EQ and compressors, had knobs for adjustment of the parameters in addition to full software control. A lot of people liked using the knobs only, because it meant they adjusted until it sounded right. When they then looked at the software page, they then found they had really big EQ boots or cuts that they wouldn't have done using the software, because they would have looked too extreme when just looking at the figures.

[Simon Barden]

And a bit of investigation just now shows that digital radio is multiplexed on to a number of fixed frequency bands and those bands are normally leased by several companies. The standard DAB codec is MP2 - 128kbps, so already frequency limited and quite lossy. But if one company is trying to squeeze more stations on its allocation of data on that frequency than full data rate allows (as there is also error correction and station information broadcast in each packet), then they can reduce the bps rate to fit more stations in, but the audio quality suffers. There is a DAB+ format that some countries have been experimenting with, but it is not backwardly compatible with DAB, as it uses AAC+ compression codec which is more efficient than MP2. So it can get higher quality broadcast in the same sized packet, but you need a receiver than can decode it, which is why it has only generally been trialled and not introduced.

[dave.king1]

I'd suggest listening a bit louder than that at times. The louder the signal, the more sensitive the ear is to pitch detection and level imbalance. If you have it too quiet, you can miss things that come to light when played back louder. You don't need to do it all the time, but at least one loud playback is worth it before signing off on a mix.

Absolutely but it is always last step and I do it after the ears have had a break, I have a constant battle with band mates who insist that the only way to listen is loud with the bass pumped which is not only tiring but covers issues.

Should add that editing and mixing should be treated like an office job with a break every hour, give the ears, eyes and neck a rest, if you know what you are looking for and know your plugins the mastering step should be pretty quick

[Marcel]

Very well said Simon. As an old school analogue audio studio person I'm not all that familiar with all the various DAW situations which is why I avoided mentioning them, but thanks to you I'm learning fast.

As for ideal listening level when mixing I was taught during my Diploma in audio engineering training that 70dBA to 80dBA is the "sweet range" for most mixing endeavours, however where required and depending on the recorded material critical listening tests of completed mixes should also be done at 60dBA for 'background music' situations and for short durations at 90dBA to 100dBA for party/night club scenarios...

It is a widely help misconception that all radio stations keep their music libraries in MP3 format. The four local commercial radio stations that I do contract technical work for all hold their music libraries on local HDD's in predominately AAC format, and all their online stream(s) is done via a digital split at the studio so arguably the quality of the stream is vastly superior to what is broadcast over the airwaves. Except for the microphones and loudspeakers the studios themselves are all digital environments... And as the stations are networked if any one station looses its local HDD library then it will "borrow" music over the network via VPN from a sister station until the local HDD and library is restored, and neither the DJ nor the listener is ever aware of where or which HDD the music track itself is actually sourced from...

[Simon Barden]

I forgot to mention that the process of MP3 or AAC compression can also create digital clipping with very 'hot' mixes, especially when using the lower low bit rate compression. You can have a really hot and compressed mix that still sounds fine when uncompressed, but sounds harsh when converted to an MP3. So that's another reason not to really push the limiter at 0dBFS when creating a master mix and allow a reasonable dynamic range.