Interesting article on the Korf site

[Mike_S]

Think it was just conjecture on the 3 different principals. It would be nice to see the same setup done with all 3 alignments overlay the graphs.
To go deeper, use different stylus types on each alignment types, then overlay.

My take away is that none of them are perfect, use the one that works for you.

 

[tothmes]

Of course the harmonics are not at 1khz but a multiplier of them and I think they are very much attenuated by the RIAA filter. That could be one of the explanations of him producing low thd figures, especially with that replicant stylus. I think if he had focussed on or extended his article to focus on 1 specific harmonic, it would've been more interesting (and given a bodeplot of his riaa filter).

That certainly is a problem in his presentation. But shouldn't the result be a skewed balance between the (still assumed only) three harmonics, by a reduction of each increased step, and hence an lower measured THD? For this incorrect distortion profile to affect the overall shape of the graph, the balance between the different harmonics would have to vary over the radius of the disk. I can't recall any such effect described in the original papers, but I haven't read them in a while.

At least we're setting up an interesting list of problems in future comparisons between different tests of his hypothesis.

 

[tothmes]

Would have been interesting to superimpose the distortion figures from the same stylus on a linear-tracking arm.

That was my first thought too. Why not do the same experiment on a linear tracker of decent quality. That would give a very interesting insight, regardless of the outcome (either the same or different)

That, and also a comparison between 9 and 12 inch arms. As a way to finally get some actual data instead of opinions.

But a problem would be the risk of the distortion spectrum getting coloured by different resonance profiles between the different arms. For his particular task this kind of preliminary first publication must be done with the same arm.

 

[Thomas_AA]

That, and also a comparison between 9 and 12 inch arms. As a way to finally get some actual data instead of opinions.

But a problem would be the risk of the distortion spectrum getting coloured by different resonance profiles between the different arms. For his particular task this kind of preliminary first publication must be done with the same arm.

Before any of that I think he needs to specify the level on the record and compare with a theoretical graph using calculation of the same level. From the site of the record, "pegel" is given as -10 dBu. I have no idea what that means in cm/s.

 

[tothmes]

Before any of that I think he needs to specify the level on the record and compare with a theoretical graph using calculation of the same level. From the site of the record, "pegel" is given as -10 dBu. I have no idea what that means in cm/s.

There's a lot of peer review and standardisation of methods to be done before we're there. Including this.

 

[ripblade]

Virtually every graph on the web shows the RIAA curve with a zero crossing point at 1khz.

There's a lot of nit-picking going on, but what I think Korf is trying to illustrate is that Lofgren et al calculations are not deterministic of actual distortion levels. As he revealed with his mass/resonance experiments, there's little correlation between theory and practice. At least he's attempting to measure what's going on rather than regurgitate the same old 'facts' without question. Some people will be offended by his conclusions.

We'll have to wait for his next installment to see where he's going with this...

 

[Thomas_AA]

Virtually every graph on the web shows the RIAA curve with a zero crossing point at 1khz.

There's a lot of nit-picking going on, but what I think Korf is trying to illustrate is that Lofgren et al calculations are not deterministic of actual distortion levels. As he revealed with his mass/resonance experiments, there's little correlation between theory and practice. At least he's attempting to measure what's going on rather than regurgitate the same old 'facts' without question. Some people will be offended by his conclusions.

We'll have to wait for his next installment to see where he's going with this...

It is just that others have said this long time ago, eg Yosh at the biglobe pages said it is not possible to ger those graphs due to other distortions present.

 

[ripblade]

It is just that others have said this long time ago, eg Yosh at the biglobe pages said it is not possible to ger those graphs due to other distortions present.

Which brings us back to his opening question, why did he choose Stevenson over Baerwald for his arms? The graphs certainly explains to me why I can't tell the difference between them in practice; I take it on faith to use one over another but the difference between them is exceedingly minor in comparison to the total palette of distortions.

 

[mr_petit]

Virtually every graph on the web shows the RIAA curve with a zero crossing point at 1khz.

There's a lot of nit-picking going on, but what I think Korf is trying to illustrate is that Lofgren et al calculations are not deterministic of actual distortion levels. As he revealed with his mass/resonance experiments, there's little correlation between theory and practice. At least he's attempting to measure what's going on rather than regurgitate the same old 'facts' without question. Some people will be offended by his conclusions.

We'll have to wait for his next installment to see where he's going with this...

His mass/resonance experiments were very flawed imho. The HFNRR record has imho a fubar vertical resonance test track that he uses, where the 20hz and 18hz track are recorded louder than the rest of the tracks. Because those resonances don't move if I increase the effective mass by 10gr or more... and I have exactly the same 'resonance' as he has (as in very high amplitude) regardless of the cartridge I use. I also have other vertical resonance tracks on other records that don't show this phenomenon. All this faulty data he said he used for his calculator, for which the math he used is very flawed (in his published formulas (he doesn't publish his exact calculator) he assumes every cartridge has the same damping factor....). So his mass/resonance experiments have no real value I think.

Also in the conclusion of this experiment we don't know if the low distortion figures published from that replicant stylus are actually lower than the mathematical models from Loefgren/Baerwald. We only know that they are considerably lower than that from an elliptical stylus and a conical stylus, and that the distortion from that elliptical stylus and conical stylus increase rather heavily to the inner grooves, so where wavelengths get shorter.
Is that a surprise? Not really to me. For me the surprise is that that already happens at 1kHz, provided we see THD here and not say groovenoise that gets amplified by stylii that can't track that noise.
I mean, I see a 'consistant distortion' of somewhere between 0,5% and 1%.
Translated to signal that is already -40dBu. Adding up that the test record signal is -10dB, thats a total of -50dB. To me that's a value that could easily be groove noise and not THD, and that the upslope we see towards the inner grooves with the elliptical and conical stylus are tracking errors. And that the alignment/tracing errors are so small that they hardly rise above the groove noise, especially as the THD value is extra attenuated by the RIAA. I mean, would the graphs be radically different if where the graphs display 0,5% would be 0,75%. That is (out of my head) the difference (3dBu) if the THD is just as loud as the groove noise.

 

[tabarddn51]

''we don't know if the low distortion figures published''

Hi, Uhm, according to Mr Google when the paper by Lofgen was written. ''In 1938, the format known as the "LP" (Long-Playing record) was not yet available; instead, records were primarily 78 RPM (revolutions per minute) shellac discs, which offered shorter playing times per side compared to the LP format, that was introduced much later in 1948 by Columbia Record''

Cheers

 

[Thomas_AA]

Which brings us back to his opening question, why did he choose Stevenson over Baerwald for his arms? The graphs certainly explains to me why I can't tell the difference between them in practice; I take it on faith to use one over another but the difference between them is exceedingly minor in comparison to the total palette of distortions.

The distortion level that Korf shows with the replicant stylus is between 0.4-0.6%. Without a clear picture of a spectrum plot would look like for this particular record with unknown level (-10 dBu is not understandable in my world. -10 dB referenced to 0.775V?), it is not really possible to say what is going on. I get around 0.1% 2nd harmonic, close to one of the null points at 3.54 cm/s RMS or 5 cm/s peak. Given other distortion that either is baked into a record or occurs during playback, it is probably difficult to trace it to the typical tracking distortion plot. Perhaps playing a quiet lacquer would reveal this.

 

[JP]

The distortion level that Korf shows with the replicant stylus is between 0.4-0.6%. Without a clear picture of a spectrum plot would look like for this particular record with unknown level (-10 dBu is not understandable in my world. -10 dB referenced to 0.775V?), it is not really possible to say what is going on. I get around 0.1% 2nd harmonic, close to one of the null points at 3.54 cm/s RMS or 5 cm/s peak. Given other distortion that either is baked into a record or occurs during playback, it is probably difficult to trace it to the typical tracking distortion plot. Perhaps playing a quiet lacquer would reveal this.

And don't forget that distortion sums by vector; it can cancel out. One can't just wave away the confounders introduced when taking a pure mathematical construct and testing it in real life. Those confounders have to be identified and addressed, one way or another. You also have the logical gymnastics of proving a negative.

Virtually every graph on the web shows the RIAA curve with a zero crossing point at 1khz.

The "zero" is arbitrary. The point is that the filter doesn't magically stop having an effect at 1kHz. Indeed, if you remove the treble time constant and normalize at 1kHz, you can clearly see that there is a frequency effect (~0.25dB @ 2kHz, but goes all the way up to ~10kHz), and where there's a frequency effect there's also a phase effect.