We have recorded 9 different versions of Take Five onto aiff files which can be downloaded here. The password is eleanor.  Please listen and decide which you like the best. There are cartridges from both Soundsmith, retipped ones from Axel and a couple of originals.

I will of course reveal which cartridges are which at the end of the poll!

Momentary pause whilst I put the original recordings on! I had equalised them down so they all had the same volume and succeeded in making them all sound rubbish!!

We have decided the recordings are a little down in treble because of the record rather than any problems with recording method. Track 9 is actually a control - it is the CD!!

thanks peter,

i've downloaded all versions. the ending .aiff is important on your files. otherwise some computers don't recognise the format.

number 8 is absolutly my favorite!!! the others are so-so. number 2,4,7 & 10 sounds i.m.o. better than the rest. the difference between this four cartridges are minimal, but i prefer the sound from number 4 & 10.

i can send you a .aiff file from my mmc 20cl van den hul if you like.

cheers and thanks for your test.

dario

Hi,

After extensive listening, my favourite of all, including my 1964 LP, is definitely the CD version. The image is broader (the piano being heard from centre to outside my right wall), the sax is more defined (you can hear Stan Getz actually blowing it) and realistic. The fact is that the CD was mastered using the original tapes, and it shows. Let's hope the audiophile 200 gr pressing will be mastered the same way: it will then sound superb, possibly superior to the CD...

Another classic worth having is the Gilberto/Getz "Girl from Ipanema" LP. The sound can be bad to astounding according to the pressing. 

Right!! We have decided that my record is a lousy recording!! I have therefore recorded four more tracks - this time Camel - The Snow Goose. 15Mb snippets so not too big too download - have also discovered that my 4000 has the right and left channel transposed and can't be a**d to change it! Makes working out the CD very easy!There are 4 tracks listen and judge - the contenders are:

The CD

MMC20CL

MMC20S with a new catilver and shibata stylus care of Axel

SMMC20CL+

Equipment used - Beogram 4000, Beomaster 5500, Beogram CD5500. Apple Powermac G5 Dual 2.0. Recorded to aiff file using CD SpinDoctor.

Can't add another poll so just post below your likes and dislikes!

All three cartridges sound great this time!

Maybe the Shibata is a little bit more refined, but this would indeed require a 24/96 encoding to fully appreciate this (please?). However, Apple Lossless is very good. The Soundsmith has more treble energy and less bass. It is a contemporary cartridge that will appeal to those who have soft speakers, or who like a lot of treble definition—how can there be so many things in a slice of engraved plastic?

I don't have this record, so I can't compare with my system unfortunately.

Anyway, is your MMC20CL an original? I have one to re-tip and I'd like to know what it sounds like... I guess it'd be very close to the Shibata modified 20S (well not a 20S anymore then!).

Jacques

I listened to the new snow* samples, and had a bit of trouble telling them apart - except for #4, which lacks snaps and is therefore obviously the CD

As I'm currently preoccupied with statistical digital signal processing, I thought why not see if I can work something out of these samples from a mathematical point of view. Here's what I did (just skip the parts that go over your head, this will get interesting later...):

• limited the samples to 20-30 seconds after the part where the music gets going - the actual music content plays no part in this analysis, it would be best if it was just white noise
• converted to uncompressed .wav so I can easily read the data into MATLAB
• summed the channels to make the samples monaural (this may have to be reconsidered)
• made a statistical analysis of the spectral content in each of the samples, using 512-point Welch approximation

The spectral analyses weren't too interesting at this point, as they represent the spectral content of the music as filtered and modified through the entire transmission channel:

• in the vinyl case: the vinyl mastering, the Beogram with its cartridge, the preamp, Peter's G5 and whatever happened after that numerically
• in the case of the CD: the CD mastering, then some unknown stuff (did you rip it directly, Peter, or played through a real CD deck?)

Anyhow, the CD sample should contain the same music as the vinyl, perhaps affected by different mastering, but it sounded pretty darn close to the same. Therefore, we can do a trick and divide the vinyl sample spectrograms with the spectrogram of the CD sample. In principle, this should give us a spectrum estimate of the vinyl-cartridge-Beogram-preamp combo, in other words: something very much related to the frequency response of the cartridge!

So here is what it looks like. The frequency axis (horizontal) is linear and in Welch approximation data points, so 0 is really 0 and 512 corresponds to half the sample rate, i.e 44100 Hz / 2 = 22.05 kHz. The bottom of the bottom end is obviously bogus since there won't be any subsonic information in the recordings, likewise the last climb is bogus as it goes over anything that realistically could be reproduced by the test setup (including the CD!). The Y axis is logarithmic difference between the spectrograms (each grid line 1 dB).

Now the conclusions:

The red graph (snow3) is obviously very different from the other two, with a pronounced treble (remember the linear frequency scale, so '300' is already clearly above 10 kHz). I believe this would be the SMMC.

The other two are surprisingly identical, although the blue graph (snow1) has a little more bass / middle, up to 3-4 kHz. Then it has a quite funny discontinuity around 5 kHz -  I can't even begin to think what causes it - some kind of a resonance in the suspension / cantilever?

Now there you go. Now back to discussing musicality, timber and clearness

To my ears, the differences are so small as to be insignificant as far as my enjoyment of the music in these samples is concerned.

I have listened several times and I can't tell one from the other!

Listening in my workroom at the moment on Beomaster 5500 / CX50 system.

When I can be arsed to move I will try on BC2300 / Beolab 1's in the lounge and BC8500 / Beolab5000's in the sunroom.

I am a bit of a dinosaur and this is my first attempt at downloading and playing music via computer.

What digital quality am I listening to here???

joeyboygolf:

To my ears, the differences are so small as to be insignificant as far as my enjoyment of the music in these samples is concerned.

I have listened several times and I can't tell one from the other!

I fully agree there regarding my own ears. That's why I was suprised myself that a statistical analysis can dig this much out of them! (and even more suprised if/when I hear it was actually correct )

A couple of corrections I noticed already...  the Y axis is of course linear in dB terms, so don't look at the grid lines: the top is +10 dB and the middle is 0. And the dip is more like at 9-10 kHz, not five. And as the spectral content was not normalized, you can think of sliding any graph up and down to fit "better" as it'll just mean a volume change. In dB terms, the differences between the cartridges are quite small.

joeyboygolf:

What digital quality am I listening to here???

16-bit stereo with a sampling frequency of 44.1 kHz, non-compressed - so as far as the file format goes, it is identical to CD audio. Anything you'd normally call "digital music" will be somewhat worse due to compression, only the "lossless" formats maintain full quality.

The following might be helpful. My Beogram 4000 has its DIN lead mixd up so the tracks are reversed - hence the change compared to the CD - Snow 4. This was played on a CD5500 through the BM5500 and feed taken from the tape out to the Mac G5. The Beogram attaches to the phono input so that they use the same path.

Snow 1 is the MMC20SH - a 20S retipped by Axel with a shibata diamond and new cantilever.

Snow 2 is the MMC20CL

Snow 3 is as correctly guessed the SMMC20CL+

Maybe you could analyse the original tracks - the record was not very good but the cartridges were still set up OK. 4, 5 and 7 are different to what you have already had.

Nice to see that Axel does a good job as far as the listening comparison is concerned.

I have him repair all my MMC2's and MMC4's to as close as possible to their original spec and I've been very happy with them. Probably around 25 by now and quality control has been excellent.

Peter :

The following might be helpful. My Beogram 4000 has its DIN lead mixd up so the tracks are reversed - hence the change compared to the CD - Snow 4.

That doesn't actually change anything, as  I summed the channels before doing anything else anyway so the analyses were effectively "mono button down". The CD version was the reference signal - if I had graphed it as well, it would've been just a straight line in the middle.

Also, this isn't any real frequency response measurement, that would require playing a test record. As I took the louder passage of the samples, it is also possible that the wider excursions of the stylus do something to the response - the analysis might look slightly different when done from the quieter part of the music.

I'll see if I have energy to do the others as well; I typed the MATLAB code pretty much directly in so I don't have a script to automate this... Some of the other samples had the clipping problem, which would artificially weigh the higher frequencies.

Very interesting - the CL+ is easily distinguished from the other two - I suppose it rather depends on what you like. Certainly offers a choice.

Quite! And also, the S retipped with a better stylus has nothing to be ashamed of (except perhaps the colour )

I'm still perplexed by that dip. That'll perhaps be enough incentive to analyze the other samples as well, just to find out if it is some kind of an anomaly with that particular sample.

That was deliberate! I actually got two done - an MMC5000 and a MMC20S. The black one is for my Beogram 4001! I can do you pink noise if you want - in and out of phase as I have a test record.

I looked into doing this for the other samples as well, but unfortunately the CD version (cart9) seems corrupted (zero file size) - and this is a rather pointless analysis without a straight reference.

Pink noise would give us the correct frequency response of the system and is much easier - you probably have  the software you need on your Mac as well! At least Audacity has a spectrum analysis step.

Pink noise or 1/f noise has a spectral power density that is inversely proportional with the frequency. Plotted on a log-log scale with desibels on the y-axis and frequency on the x-axis, a flat system fed with pink noise should show up as a straight downward slope.