RE: arsclist reel to reel player/recorder

["George Brock-Nannestad" <pattac@xxxxxxxx>]

From:           	"Copeland, Peter" <Peter.Copeland@xxxxx>
To:             	"'ARSCLIST@xxxxxxxxxxxxxxxx'" <ARSCLIST@xxxxxxxxxxxxxxxx>
Subject:        	RE: arsclist reel to reel player/recorder
Date sent:      	Thu, 14 Jun 2001 11:47:01 +0100
Send reply to:  	ARSCLIST@xxxxxxxxxxxxxxxx

Dear Peter, for me the third lesson has not ended yet. On 14 June 
2001 at 11:47:01 you wrote (among other wise things):

>     Classical information theory describes something called "the
>     impulse
> response test". In this, a very short transient spike is presented to
> the equipment under test, and the resulting output is recorded. The
> British magazine "Hi-Fi News" was recently been comparing top-end CD
> players using an impulse response. Perfect impulses are easy to
> generate in the digital domain. You have a large number of consecutive
> zeros, then change just one sample to +32767 (for sixteen-bit tests),
> and you can buy CDs with this test.

----- Now, what is the relevance of this digital signal? It is a signal 
which would never be generated by an A/D converter in the 44.1, 16 
bit realm, because, due to the criterion promulgated by Nyquist, 
there are at least *two* samples to represent the highest frequency 
that one desires to keep track of. For this reason, an anti-aliasing 
filter is fitted before the A/D conversion, so that we never have any 
signal component above 20 kHz. The kind of signal you mention 
can only be generated by extremely stupid digital editing, and the 
smoothing filter fitted after the output of the D/A converter ought to 
reduce it to - nothing. However, trigger times (jitter) and other 
electronic problems may feed this brief high-level signal on the 
output of the D/A converter into various undesired places, and 
certainly some capacitors will be DC "pumped". 

 The resulting analogue signals
> coming out of Hi-Fi News' CD players were all totally different from
> each other! The reviewer didn't go into the differences very deeply,
> but being familiar with the information-theory issues, I could see
> that some were made to give the widest uncorrupted frequency ranges,
> others made to give the best transient responses, etc.

----- Again, I am not entirely sure what you mean: the widest 
uncorrupted frequency range *will* give the best transient response -
 unless there are time delay problems.

>     And this is only high-end digital *reproduction*! We are currently
> attempting to tackle analogue-to-digital *converters* in a similar
> way, and develop a standard methodology so we may document the
> performance of such converters for all time. As far as I know, the AES
> has made no recommendations for such a test, but we are not in the
> main British Library building, so I cannot confirm this; can anyone
> correct me?

----- My comment here is that if you desire the precision that 24 bit 
really represents, you would want to look at some Agilent 
measuring equipment. They have digital voltmeters possessing that 
precision. What I would like the audio community to look at is price 
and the sample rate that is supported at that precision. Also the 
performance over time is interesting!

Kind regards to all,


George
Preservation Tactics