CD stoplight - some analysis

Well, much has been said of CD stoplight, and one thing that stuck in my mind was the claim that those who believe in it working have not presented any evidence to support their claim other than a claimed perception of improvement.

So, how to test this using ( something that has become a catchphrase where I work) "a more scientific way of doing it"?

Much of the debate has raged over the issue of bit errors. Stoplight doesn't affect bit errors as far as I know. A digital system such as making a digital copy of something (ie using a PC) should be unaffected by it. It's where real time conversion to analogue takes place that it has an effect, especially where the clocking is derived from the data of the CD.

Thus, in order to measure the effect, a recording must be made from analogue, and compared with and without stoplight.

So I did. I made two copies of a track from the original CD (ripped to .wav on hard disk, then written at 12x onto two Philips CD-R80 52X CD-Rs). One of these copies was then treated with stoplight on the outside edge and on the ridge near the spindle, the other was left without.

These were then played without changing anything. A recording was made from the tape-out of the amplifier onto a PC. The recording format was .wav, and it was done at 24bits, 192kHz in order to provide a much higher theoretical resolution than any effects we might be looking for. Also, using a sample rate that is not a direct multiple of the CD 44.1kHz sampling rate means that timing errors (since this is via analogue, they can't be matched for start and stop time precisely) relative to the CD sample rate are similar for both recordings, rather than one having a greater offset than the other.

So, onto some results...



The very beginning of the very first note of the track used (Starsailor - Tie up my hands). There are some subtle differences in the waveforms. Which is which is entirely inconsequential because we're looking for differences.

Do little differences like those seen in the above image amount to much necessarily? Some analysis is needed...





Now, that's fairly significant difference between the two....

Again, I have not said which is which.

Discuss.

 

i must admit there is a hell of a differance regardless of weather it works or not.

know that the CD stotlight pens cleans the sound up the 1st one of both pictures should be with the pen as they contain less varation.

 

oh eck
im off for the weekend

:duck:

 

Interesting - at last someone's throwing some science at the issue

My vote says the 2nd one is with the stoplight as the effect I've noticed is usually a cleaning up of harsh or edgy treble.

btw, why have you smudged out the title bar in the first two images?

Michael.

 

The title bars show the filenames, which reveal which trace was which. Perhaps it is unimportant, but I'm interested in getting some comments on the traces and the frequency analysis before revealing which is which, so as to try to avoid potential bias.

 

Nice graphs
but is the difference actually audible:guitar:
or are we yet again in the terrain of subjective hearing :crazy:

 

Repeat your experiment but using the same CDR a number of times. If they all look the same then you may have the start of something.

Paul

 

Indeed. Since the test occurs in real time, it will take some time to repeat (plus the output file of 5min30s at 24/192 is around 400MB...). However, I do intend to repeat (as science is repeatable)...

 

Very interesting,

I would say that the green penned CDR is the second one as it contains less high frequency content.

Scientifically, the next thing to do is to repeat the experiment a until you can get a statistically significant sample and demonstrate that it is not sample to sample variation/experimental error.

 

make sure you keep your eyes closed though .
i'm going to buck the trend and say that the 1st one is the stoplit version as it looks to my ropey eyes and brain that there is more detail coming through but what do i know?
cheers


julian

 

There are several points to make here I suppose:

Firstly, as Paul Ranson points out, you need to do this several times with each disk.

It seems reasonable that you could get this kind of difference using the same disk multiple times. In other words: it's experimental error caused by other nuisance variables.

For instance, 24 bit's is an awful lot of bits, when you think about the general level of signal noise floating about Again, the ADC in your PC could be generating a significant deviation in your results. I've also found that the sampling frequency of most PC sound card clocks isn't particularly stable..

I'm very suspicious of short sampling intervals when used in conjunction with the FFT. Using Log scale for frequency is both a good an bad idea What is the vertical scale in your plot. [You'll recall your theory: good accuracy in the time domain necessarily requires poor accuracy in the frequency domain, and vice-versa..]

Try re-running the FFT plot but just slide the begining of the sampling interval a few 10's /100s/1000s of samples either way in time.. Look at the plots again.. Can the same data generate the changes in FFT? (I suspect you might find this is the case - it might also depend on how your software is windowing the samples for each FFT..)

 

Fascinating post! Thanks very much Isaac for taking the time to actually do something, rather than just waffling like the rest of us.

The first one appears to have a number of odd blips (spikes) in the frequency response that the second one lacks, so maybe it's the second one, with the blips cleared up... Or maybe the blips are indeed part of the signal and an un-stoplighted disc smudges them away, so the stoplighted one would be the first one, with the blips intact... Or maybe that apparently "hashy" treble in the second one is what the stoplight clears up, so maybe it's the first one...

Or maybe they're statistically insignificant experimental errors.

Can't wait to see how this turns out.

To be honest, my vote is probably that the differences are nought to do with the stoplight, and repeated experiments will show this to be the case. Why? Because if it were that obvious, Densen - or somebody - would've demonstrated it already, and gotten Sony/Philips to change the CD standard to include the green barrier or something. Surely. Right?

Dunc

 

This is probably anal, but when we are talking such small difference, dosn't the CD burner have to be taken into account, also the back ground processes of the computer at the time.

For instance is the stop light CD was burnt and then straight after the normal CD, could the drive have any bearing, as it may have been working harder for the first CD?

Anal I know but these things must be considered

 

Er why?

Given that we are searching for audible differences, why not use the dedicated faculties given to us at birth to determine the products worth?

Geez this scientific crap is just barking up the wrong tree IMO. After all, if scientific research were the be all and end all of good sound quality, Sony product would rule the world, and valve amps would sound dire.

 

But Mike, if you just use your ears then you are obviously imagining the improvement. Some will only accept there is an improvement if you can measure it. Surely thats far more important than acually listening to some music

 

Why is anyone arguing? There's no argument! Isaac feels it's worth his while to do some objective tests, and a fair few of us feel the same. If you don't, that's great, go listen to your music and leave us nerds alone - we don't want to hear it!

Dunc

P.S. - What is everyone doing up at this ungodly hour on a Saturday???

 

If someone is taking your cash by selling you a green pen, then I can only say that I wish I were taking your money

Luddite.

It is...

They do. [OK, Sony, Panasonic, Phillips etc.]

The profit of Wadia/Levinson/Krell is loose change to Sony..

They do.

 

Well I know it works, so for me there is no need for measurements. However if Isaac can come up with some objective measurements to demonstrate a difference thats great. It would then stop all the naysayers jumping on our backs every time we mention the green pen.

 

Whats even more worrying is that Dat is posting at this hour from the states. Maybe he is nocturnal.

 

I am up late planning further dastardly deeds of blind test terror