Bit and Sampling rates - Real or Imagined?

[pietercape]

Hi,
I am new to Audacity but so far I love it, especially the simplicity (no knobs and dials) and it works for me. So far!

I am thoroughly confused by the discussions on bit-rate in all the forums and, as I am not a technical person, I have a basic problem understanding the benefits of 32 bit recording/editing if the basic problem seems to be the "throat" through which it must all pass, the on-board soundcard of your computer which in most cases appears to be 16 bit.

My music collection is mostly classical but includes all genres. I have completed converting my CD collection (EAC-Wav-imported to MediaMonkey-Flac) and the 2000 composition 9000 tracks took hundreds of hours to classify, tag and catalogue. now it is the turn of my Vinyl collection which includes many 60's & 70's recordings. I accept that many of these were not the "cleanest" even when brand new and do not expect to end up with "pristine" digital conversions. I come from an era when we spoke of "garbage in, garbage out" I use a Graham Slee "Bridge" Gram Amp1 for the turntable signal and then the line input to my Vista based PC. The soundcard is a Realtek High Definition Audio that can be set to 2 channel, 16 bit, 96000Hz. Output is a Cantatis Overture 192 soundcard with dual mono mode Burr Brown PCM1794A DACs and audiophile op-amps (Specs at http://www.cantatis.co.uk/upload/file_15_4.pdf) It was expensive but provides superb sound to my Zennheiser Classic II headphones.

I have checked the soundcard review forum for input soundcards and the two most mentioned are the Behringer UCA202 and the Edirol UA-1EX. I tried the Edirol 18 months ago and was not impressed, I returned it. So I am still stuck with a 16 bit soundcard for converting my LP's.

Here are two quotes from the forums:

"Steve Aug 2010

It may be worth mentioning here that although Audacity does not currently support 24 bit recording on Windows, it does support 24 bit / 32 bit float for importing, editing, processing and exporting. For anyone that wants to take advantage of 24 bit recording, and use Audacity for editing, they could record with an application that supports ASIO (such as Sonar or Wavosaur), then export the recording as either 24 or 32 bit WAV, then import that file into Audacity."

"Waxcylinder 3 Aug 2010

I record from my old Technics SL-150 deck with SME 3009 Improved tonearm, through an ART phono pre-amp and then on to an Edirol UA-1EX (the pre-cursor the the UA-1G) set at 16-bit using the latest drivers from the Edirol website."

Steve says that Audacity does not support 24 bit recording and elsewhere the team recommends using 32 bit float for all imports. If the recording came through a 16 bit soundcard (the throat) what does Audacity do to upgrade it to 32 bit? What can it do to enhance the quality of the sound?

Waxcylinder (regular contributor to the forum) in the quote above also uses a 16 bit setting on his Edirol.

I am obviously showing just how ignorant I am in technical matters but I do not understand how Audacity can enhance a 16 bit recording by upping the bit bit rate. What do the added bits consist of? What is the benefit of doing this?

If there is no improvement then surely we can just stick to 16 bits particularly if we are going to lose bits when we downsize to 24 bits which is what I would like to use as MediaMonkey does not play 32 bit float.

[kozikowski]

what does Audacity do to upgrade it to 32 bit? What can it do to enhance the quality of the sound?

Those words are more slippery than you think they are.

Super High Sample Rates and Bit Depths don't necessarily increase the quality of the sound. You won't suddenly notice that the sparkling quality of the singer is a lot more sparkling or "transparent" or "velvety" or any of those other audio phrases.

The advanced settings are used for post production to avoid damage -- and as a side issue, may be what the client is expecting, even if he/she has no idea why.

If you need to do advanced production/filtering/corrections, 32-floating is far less likely to damage the sound than lesser formats. For just one example, if you apply a filter that results in an increase in the sound signal (surprisingly easy to do), the show may go "beyond maximum" and become permanently damaged -- unless you're doing your production in 32-floating in which case sound values beyond maximum are supported and may be managed.

People who do capture in 24-bit do it because of the wild nature of live capture and to get the digital system noise floor away from the restriction of the microphone electronics. The show sound channel noise goes from dead quiet (if you do it right) to deader quiet.

I do live voice capture and use either 48000 or 44100 at 16-bit because I have a very controlled environment and expect little or no post production filtering. It works for me. I would expect the higher formats to add almost nothing to the work but take up much more disk space to store the performance. Every time you boost the specifications, the file sizes go up. Sometimes strikingly so. Newbies accustomed to downloaded MP3 files are stunned at their first high-quality recording that maxes out their hard drive.

Koz

[pietercape]

Thanks for that response. I am not looking to do a lot of editing as I am happy to live with the quality of my Vinyl recordings and so I will take your implied advice to stick to 16 bps recordings.

After my post I searched Hydrogenaudio for more information on 16/24 bit recoding and found a link to an older Audacity thread http://audacityteam.org/forum/viewtopic ... 53&start=0 that seems to confirm that recording at 24 bps using a 16 bps card as the source just adds "padding", no real improvement in quality.

Still think that Audacity is perfect for my needs and who knows, I may be tempted to buying a 24/96 capable soundcard with a good quality ADC.

Are there any recommendations?

[steve]

I have a basic problem understanding the benefits of 32 bit recording/editing if the basic problem seems to be the "throat" through which it must all pass, the on-board soundcard of your computer which in most cases appears to be 16 bit.

If by "throat" you mean the limiting factor for sound quality, for your set-up that is going to be on the analogue side.
The quality of your records will be one limiting factor and the quality of the analogue electronics between stylus and the Analogue to Digital (A/D) converter will be the other.

You've got a very nice phono pre-amp, but the quality of on-board sound cards are renowned for being poor quality. The Realtek High Definition Audio A/D converter is pretty good, but the electronics between the mini-jack input and the A/D converter are usually the cheapest that the manufacturer can source. On top of that the component layout of those analogue components is usually designed to fit anywhere that there is enough space rather than a layout aimed at sonic excellence. It is not uncommon for the sound card circuit tracks to wander half way across the mother board, which is a very bad environment for low level analogue signals due to all of the high frequency interference that is present.

I'm rather surprised that you were unimpressed with the Edirol UA-1EX as it should be hugely better than the on-board sound card. Perhaps you got a faulty one, or maybe there was some other factor that caused the results to be poor? I use a UCA 202, and while it is not really "audiophile" quality, it is a lot better than the on-board sound card of my laptop. The Edirol UA-1EX should be significantly better than the UCA 202, though I've not had one to test.

what does Audacity do to upgrade it to 32 bit? What can it do to enhance the quality of the sound?

Converting from 16-bit to 32-bit is just a matter of how the numbers are represented.
A simple analogy would be that rather than counting 1, 2, 3, 4, 5, ... counting 1000, 2000, 3000, 4000, 5000, ...
This does not improve (or reduce) the sound quality in any way.
Where it does improve things is when there is any processing of the audio.

Extending the analogy, let's say that you have sample values of 0, 4, 7, 9, 10 and you amplify the audio by a factor of 1.1. This will produce sample values of 0, 4, 8, 10, 11 (the results are rounded to the nearest value). Note that there is now a jump between the second and third value. This is distortion.
Now let's say we are using the "thousands" scale and do the same thing. We then have the numbers 0, 4000, 7000, 9000, 10000 and when we amplify by a factor of 1.1, it produces the values 0, 4400, 7700, 9900, 11000. In this case there is no distortion, each result is exactly correct.

Unfortunately it is not quite as simple as this and there is a trade-off for very slightly increased noise when converting back from 32-bit to 16-bit. So you take your choice, distortion or noise.

For a single processing operation such as amplifying the recording or applying a fade-in or fade-out, there is probably little to choose between 16-bit and 32-bit (assuming that you are starting with a recording using 16-bit hardware), but the more processes you apply, the more calculations are done on the data, and the inaccuracies of 16-bit calculations mount up, which very rapidly makes 32-bit processing the more attractive option.

One other relevant factor is that Audacity always processes at 32-bit precision. This means that with the default Audacity settings, if you have a 16-bit audio track and apply any processes to it (this does not include simple cut/copy/delete/paste editing) then the 16-bit data is processed at 32-bit, then converted back to 16-bit when it is returned to the audio track. This gives you the worst of both worlds. The audio gains distortion due to rounding errors, AND it gains a little bit of noise when it is converted from 32-bit back down to 16-bit.

The short answer is that if you intend to do any processing of the data, the highest quality will be achieved if the audio track is in 32-bit (float) format.
If you are only doing simple cut/copy/delete/paste editing, then there is no harm in using 16-bit audio tracks.

Note that the errors and noise that we are talking about with these digital calculations will usually be extremely small, and are likely to be totally swamped by noise and inaccuracies produced by the sound card's analogue circuitry.

[pietercape]

Hi Steve,
Many Thanks for taking the time to respond to my question. Your explanation (and I understand that you have greatly simplified it) makes a lot of sense and has made it a lot clearer.

As I said in my post, I accept that my vinyl recordings have limitations but I like them that way, warts and all. I have taken care to set my front-end up as well as I can with the best components I can afford and while good certainly not the best and the out sound card, the Cantatis Overture is really very, very good but of course I have no control over the motherboard layout.

I am using Audacity for two things. It is great for click removal if you are prepared to spend the time and I do it simply by zoom and cut. It is also great for splitting tracks and, best of all, it is great to join tracks. Many classical recordings on CD split compositions into separate tracks where the composition is meant to be played "attacca", that is as one long piece with no breaks. Audacity does a great job here, again simply by cutting and pasting. I am not really editing at all and I guess that 16 bit is good enough for me.

Any recommendations on a good PCI soundcard especially with a good ADC will still be appreciated though!

Thanks to the Guru's at Audacity for a great program and for your contributions to the forum.

[waxcylinder]

I too am surprised that you did not like the UA-1EX and suspect, like Steve suggested, that you may have had a duff one.

Originally I ran mine in 16-bit mode as I had it in its standard mode using the default windows drivers. Latterly (at the request of Gale Andrews who wanted me to do some testing) I downloaded the latest Edirol drivers and set the device into its Advanced mode. In this mode my understanding is that it runs at 24-bit rather than 16-bit. I record into Audacity with Audacity set at 32-bit floating to get the best headroom for editing as Koz suggests - and set at 44.1kHz. On Export I get Audacity to downsample to 44.1kHz 16-bit PCM stereo WAV files, using triangular dithering.

The results I have got with this have been very good indeed. I listen on a Rega Planet CD deck, feeding a QUAD 33/303 pre/power amp and on to QUAD ELS-57 electrostatic speakes which are famed for having a very detailed sound.

=======================================

BTW I would recommend looking at Brian Davies excellent ClickRepair software. It costs a little ($40) but its results are just a little shy of magical IMHO - see this sticky thread: http://forum.audacityteam.org/viewtopic.php?f=28&t=1994

WC

[steve]

Click removal is a process that will benefit from leaving Audacity's default Quality setting of 32-bit as it is. If you change it to 16-bit then the click repair effect will needlessly add a bit of distortion and a bit of noise. The only real downside to using 32-bit float as the default is that the Audacity Projects will be twice as big as they would be with 16-bit tracks, but as you do not need to save the projects (once the track has been safely exported as a 16-bit WAV file you can close the project without saving it) the size of the project is not important.

As waxcylinder mentioned, Brian Davies' ClickRepair has had a lot of excellent reviews on this forum (and elsewhere).

Normally I'd recommend something like the Edirol UA-1EX as a high quality but inexpensive sound card upgrade.

[waxcylinder]

Click removal is a process that will benefit from leaving Audacity's default Quality setting of 32-bit as it is...

I believe true if using Brian's ClickRepair too.

The first part of my workflow is:
1) capture one side of LP - Audacity @32-bit 44.1kHz
2) Export a 32-bit WAV file
3) Process this through ClickRepair (produces a 32-bit WAV as its output)
4) Input the cleaned WAV into Audacity
5) delete the original capture track
6) alll other editing/processing done in Audacity.

WC

[kozikowski]

I'm rather surprised that you were unimpressed with the Edirol UA-1EX as it should be hugely better than the on-board sound card.

If you can hear it. Nowhere is the monitor and speaker system addressed and if the music can't make it out of the computer into the room, then nothing you do will make much difference.

The other more awkward possibility -- particularly if you're transferring your collection of Edison Diamod-Cuts that you bought new -- is your hearing might not be up to it. Some of these changes require a young woman in a quiet room to hear.

To give you another point of reference, 44100, 16-bit Stereo is the sound standard for a Music CD. When it was introduced, the audiophile community screamed the standards weren't high enough and some of them (but not many) proved they could hear the difference. Those are the people still listening to vinyl, but that's the technical dividing line. Anything higher than that contributes almost nothing to strict musical quality -- all other factors being equal.

The music on a Music CD is of fixed quality. There's no such thing as making a CD a lot better by messing with the data. All you can do is make it worse. That's why we recommend doing the whole pathway at 44100, 16-bit Stereo if your destination is Music CD. That way there's little or no conversion damage.

Then there was the study of college kids who far preferred distorted MP3 to other technologies.

Koz

[pietercape]

Thanks Steve and Waxcylinder for your feedback. It is perhaps time to try the Edirol again. My main problems at the time were with ASIO and I have learned much about set-up since then.

I have just completed three recordings of one side of an LP. I set my line in and stereo settings on the soundcard to 2 channel, 16kbps, 96000hz. (max capacity)

Recording 1: I set Audacity to record at 24kbps/96000hz and exported in FLAC 6 24 bit. This resulted in a 249.3mb file and a bitrate of 2592 as shown in MediaMonkey

Recording 2: Audacity set to 32kbps/96000hz exported at Flac 8 24 bit. File size was 249.5mb bitrate 2591

Recording 3: Audacity set to 16kbps/44100hz exported as Flac 6 16 bit. File size 64.9mb bitrate 673kbps

I listened to the three tracks with my Zennheiser Classic II headphones which are very good and purely subjectively (and I understand the limitations of listening comparisons) I like recording 1 best. It sound bright and airy. Compared to recording 1 the 16kbps/44100hz recording ( #3) was good but subdued almost muted.

So for me the million $ question is whether the better colour in the 24/96 recording compared to the 16/44.1 recording is worth the extra cost of a soundcard and the significant increase in storage space required. Mmmmmm.

On a question of logic and given that the input recordings were all 16/96 why is it that the 24/96 recording sounds brighter or more colourfull and airy. (I am running out of words here trying to describe the sound ) Is it due to the lower sample rate of 44100 in the 16 bit recording?