Crappy Sound At 8 Bit

[Guest Caroline]

I have copied and pasted a previous stream below. We are having the exact same problem of the sound being really crappy (full of static) at 8 bit. Just to clarify - is there any way for us to record and then KEEP the sound format at 16 bit? Will the sound eventually have to be converted to 8 bit no matter what, and thus the sound will be crappy no matter what?

 

 

D/4 PCI will accepts 8 bit sound files but only computers prior to 1998 have

8 bit sound cards. Meaning we need to compress but that leaves the recording

full of static.

 

Your software requires:

 

8 bit, 11025 Hz, Mono

 

We can not record an original sound file in 8 bit without static. It is

impossible. We use a professional sound studio for our sound files and they

can record in 16 bit but NOT 8 bit without static. 8 bit is too old for

their professional sound studio. 8 bit has not been used for recording

quality sound files for almost 10 years. Recording studios currently use 24

bit or 32 bit technology for excellent files.

 

We have many recordings in 16 bit , 11mhz, Mono to try to use your software.

They sound perfectly clear BEFORE we compress them. Of course we have to

compress to match your requirements of 8 bit, 11025 Hz, Mono. The result is

static and a sound quality to the phone user that is completely

unacceptable. Very unprofessional.

 

Your software is archaic if it only accepts 8 bit recordings. 8 bit

recording cannot be made by todays computers and compression of 16 bit

creates too much static no matter what software we use to compress it.

 

The norm of a computer sound card is 16 bit since 1998. We do not have

access to a computer sound card prior to 1998 nor should we be expected to.

 

Questions:

 

1. Why does your software insist on 8 bit recordings when your clients can't

record to meet this spec since 1998?

 

2. We obviously need a work around for 8 bit, using the software we bought.

What do you recommend?

 

3. Do you have software that uses 16 bit?

 

4. This D/4 PCI is not usable in its present form to us. Tell us all the

options we have to be able to use software employing a 16 bit recording (or

higher) from a professional sound studio?

 

 

 

SupportTeam:

Cut and Paste of topic 2271 replaced with a link:

http://voiceguide.com/forums/index.php?showtopic=2271

Edited 03/31/2006 09:33 PM by SupportTeam

[SupportTeam]

Looks like your current studio cannot make the recordings of the quality you need.

 

Have you listened to the sound files which came with VoiceGuide? (see VG's \system\voice\ subdirectory). I think you'll find they play quite clearly...

 

You may want to forward some of those sound files to whatever studio you're using now and just tell them to make their sound files sound as good as those ones...

 

Have a look at the list of people that are accustomed to making recordings for the telephone equipment here: http://www.voiceguide.com/servCallflowDesign.htm

You should probably approach one of them instead of using your current studio...

 

The Dialogic cards and most other telephony cards out there use pretty much the same sound file format. The sound format chosen by them is more then adequate to deliver clear sound over the phone lines. If your sound file sounds hissy etc it's not because of the sound file format (otherwise every automated recording/playback in the world would sound like that...) but because of the recording itself.

 

If there was an inherent problem with Dialogic cards' sound quality then you'd find that many more people would be complaining about it...

[ktruk]

Caroline: A bit of background may help you and others appreciate the sound quality issue:

 

1. The world uses digital trunk lines to connect all cities/towns and countries. This is ISDN, and it works at 8KHz and 8 bit. Fact. You can't change it or do anything about it. This has been so for nearly 2 decades. So every call you make or receive is at this 'quality'.

 

2. You will note that almost very call you make on digital lines (and most analog) is clearly intelligble, at the expected quality and quite noiseless/static free. In fact ISDN lines are so quiet, the ISDN specification had to be extended to put noise on the line, so people didn't think they had been cut off, (Called CNG or Comfort Noise Generation). So, if the lines are clearly okay with 8KHz at 8bit, recordings at this spec should be just as good for you in theory.

 

2A. Even at 8KHz 8Bit, this is far better than older long-distance POTS (old fashioned analog) lines that were only good for 3KHz and an equivalent 4-6bit resolution due to noise and interferance. In fact some early voice modems only worked at 4KHz and 4 bit spec and were 'okay' at the time.

 

3. The encoding of your voice is done at the exchange and decoded at the receivers exchange (analog line) or digital handset (digital line). The encoding compresses 12 bit sound into 8 bits using 2 different methods. Depending on your country, it either a-law or u-law methods. When you call another country using a different scheme, it is converted for you. So, no matter what you record, or how you do it, it ends up at at 8Khz and 8bit as far as the telephone system is concerned.

 

4. If you record at 16 bit, only 12 bits get compressed to 8 bits, so depending on the device doing the compression (telephone hardware card or software driver), you may get slightly better quality at 16 bit, even if it ends up at 8bit. The quality difference is slight, so many systems don't bother and just take the 8bits and pass it on to the telephone system. If you send the telephone interface any other format (say 11.25KHz 16bit) it will be converted to 8Khz 8Bit anyway. Its just that some cards (eg: Dialogic) are designed for this format by default.

 

4A. Why didn't they bother? because a few years ago, when disc-drives were expensive and only a few 10's of megabytes, it made a lot of sense to store huge sound files at 8bit, half the size of 16 bit files. Also, some schemes did the compression first, like dialogics VOX format that is a 'ADPCM' format that makes better use of the bit-range using 'Adaptive Differential Pusle Code Modulation'.

 

5. So, how to get some decent recordings: Record at a high multiple of the frequency rate, so if you are aiming at 11.025KHz, go for 44.100KHz or if 8KHz, go for 32KHz, then the downsampling of the signal will not suffer from "artifacts"/static. Record at 16bit or higher resolution and downsample to 16 or 8 bits using something like Cool-Edit. Record your sound files as 'raw' wav/pcm format, not some intermediate format. Then you you will get some good recordings. Your studio should know all about this, and even if they don't have the downsampling software (which they should) you can do this yourself. Just get the high-quality recordings and downsample them yourself.

 

6. Audition your recorded and downsampled prompts on the phone, listening to them over your speakers/headset will always sound lousy/bad. Over the phone, they will sound better and as expected.

 

I hope this helps.

 

PS: If your sound files are played back unitelligibly with loads of static, then you are using the wrong file format or have the encoding scheme wrong (a-law/u-law).