Why is reading crosstalk so inconsistent across the different tools?

[J.R. Boisclair]

I know I must seem like I'm completely hung up on crosstalk, but I just had more experiences from which I can only conclude that there is no reliable way to ever objectively set ideal crosstalk - it's all relative to the particular test record being used.

I decided to experiment. I found four copies of the old Command Test Record (CS100) is excellent or NM condition and bought them. This record has tracks for L and R crosstalk that contain a 1KHz signal in one channel and nothing on the other (like some other records), but it had been lauded by some in the analog world as being among the most accurately cut test records out there. I also bought two other copies of the AP Ultimate Analog test record. Including my other copy of the AP test record, I ran crosstalk measurements using the same softwares (A+ and the Visual Analyser oscilliscope) using all 7 records. The results were less than encouraging. The variation across results was up to +/- .9dB, consistently. That means depending on which two records one might use, if one shows the R to L difference of +.9 dB, the other would shows the L to R difference of + .9dB - a difference in crosstalk measurement of 1.8dB! More amazingly, the difference across just the AP Ultimate Analog test records was -.7/+.8 dB, a 1.5dB variation that, if choosing one measure as right and the other as “wrong”, would have the cartridge tilted in the completely opposite direction!

Frankly this suggests to me that the whole mania of trying to get anything right using any test record is nothing but random.

Not totally random at all. Results are simply explained by variation across a normal distribution curve. This means that if you took the average results over a large number of test records, you would be pretty damn close to perfect. However, you aren’t collecting enough information if you’re only looking at decibels. You have to watch the phase. If you did this, I think you would see a little better consistency across records.

The problem starts with the symmetry (or lack thereof) of the cutting stylus. As I’ve discussed in my videos, this is a big problem and very common. I have spoken with many mastering engineers and none of them inspect for this.

Not a single test record to date offers any assurance that their four cutting angles, cut path and stylus symmetry were defined and margin of error noted before cutting the lacquer. THEREFORE, not a single test record can be trusted as a reference for any given playback alignment test.

We are doing something about this.

 

[J.R. Boisclair]

It’s not particularly about cutting process. It’s about flatness of record.

A record might seem perfectly flat but it isn’t. There are minor warps over the record from outer groove to inner groove. The problem is you cannot flatten the record with clamping method you use. Actually only successful method I have encountered that can flatten those minor warps is vacuum hold down.

There is a distinct difference between AM and AP test records in terms of azimuth but I get same azimuth point when there is vacuum hold down. When there is no vacuum hold down perfect azimuth point differs slightly. This happened with every vacuum hold down turntable and none with others.

So, the problem is vinyl material and flattening record with a machine helps but vacuum hold down solves it.

I haven’t done a controlled variable test of this hypothesis. I think it is worth inspecting

 

[tony22]

Yes @mtemur, granted, I have only a periphery ring and a record weight, but have you tried to measure crosstalk across multiple copies of the same record?

 

[tony22]

You have to watch the phase

A+ does display phase for each direction of crosstalk. Within this range of variation seen for the separation measures, the phase didn’t vary by more than a few degrees. Here is just one example (of several).

 

[mtemur]

Yes @mtemur, granted, I have only a periphery ring and a record weight, but have you tried to measure crosstalk across multiple copies of the same record?

No, I haven't. If you have an ORB record flattener I recommend you to flatten those 3 AP records even if they look flat. After that check if azimuth readings are still differ 1.5dB.

 

[tony22]

If you have an ORB record flattener

I’ll have to borrow one.

 

[tony22]

Hello @Balle Clorin. For today's exercise I took all four of my Toshiba records, and my Denon test record, and measured the amplitudes of the crosstalk tracks using Visual Analyzer. Here is the chart:

Test Record	L->R	L->R	L->R Crosstalk	R->L	R->L	R->L Crosstalk
Toshiba 1	-18.18​	-44.9​	-26.72​	-42.81​	-17.51​	-25.3​
Toshiba 2	-18.27​	-49.4​	-31.13​	-46.59​	-17.67​	-28.92​
Toshiba 3	-18.23​	-46.87​	-28.64​	-50.01​	-17.71​	-32.3​
Toshiba 4	-18.02​	-42.44​	-24.42​	-49.56​	-17.7​	-31.86​
Denon	-11.57​	-33.83​	-22.26​	-47.17​	-10.88​	-36.29​

I know you like the Toshiba record, but unless I'm doing something wrong in VA it doesn't seem like there is any more consistency with these as any other record where I have multiple copies. What is odd, however, with the Toshiba record is when I look at the "no signal" channel on each test track, that channel shows essentially nothing on the scope - not even a hint of a sine wave. It's like the low amplitude channel has literally no signal on it. So it makes the extreme negative values almost (to me) meaningless as no signal is really being measured. I'd love to get your thoughts on this. And it should go without saying, look at the very different readings from the Denon record. All done with periphery ring on each record and Stillpoints LPI v2 record weight. All records otherwise appear flat when rotating on the platter before the ring goes on, although yes I wish I had vacuum hold down. But I honestly don't think that would suddenly makes all these experiments we've been doing come to any happy conclusion.

 

[Balle Clorin]

Thanks,I did som comparisons too.. I choose my Toshiba since it gives A median value compared to my 20 test record Tracks, and it gives high values.
I posted A table, will try fo Find that post..

Here is check against, AM, Ortofon and Ultimate test record. How I clamp the record or not affects the resultt….

EDIT : strange spell check is confused by different language

Anyway crosstalk seem to vary a lot across record and vary minute differences when clamping the record or record condition matters. Also the how you measure matters, I like to use a filter below 400 or 100hz to remove rubbish from the measurements


More here

Crosstalk measurement - Which record is best?

I want to start a new thread to discuss my findings and share some thoughts about crosstalk with you. There are some threads here dealing with crosstalk from time to time, but then the topic drifts away. For my measurements I used a linear tracker from Uher, the PS936. So I had not to deal with...

www.audiosciencereview.com

 

[tony22]

I like to use a filter below 400 or 100hz to remove rubbish from the measurements

:head smack: I forgot to use a 400Hz high pass filter in VA! Thanks for reminding me. I had done this in previous tests but forgot to do it for these records. I’m still concerned about what, in VA, appears to be just random noise in the “no signal” channel for the test tracks. To me it suggests that for these records, trying to get matching crosstalk figures may not be the right way to go about it.

 

[microstrip]

I set azimuth looking at crosstalk using an audio spectrameter and focus on the peak value of the 1kHz component - it is the important value, as it rejects noise contributions. When using a scope our eyes carry this filtering. .

 

[tony22]

I set azimuth looking at crosstalk using an audio spectrameter and focus on the peak value of the 1kHz component - it is the important value, as it rejects noise contributions. When using a scope our eyes carry this filtering. .

Visual Analyzer displays the spectral content as well, in a separate window below the scope window. So you go for peak spectral amplitude at 1KHz, or matching them, L-R and R-L?

 

[Balle Clorin]

Using A bandpass 400-3000 removes both low end runbish and clicks

 

[Balle Clorin]

The result with Merolemez test record with varying Azimuth cutting. Band pass filtering 1khz.
. Cartridge is not adjusted , the tracks have different azimuth giving diffferent crosstalk
a perfect test record and perfect stylus and generator should give blue and orange dip at 45 degrees
. Left is V15 setup with Ortofon record, right is ATOC9 setup with Toshiba record,

 

[Balle Clorin]

Hello @Balle Clorin. For today's exercise I took all four of my Toshiba records, and my Denon test record, and measured the amplitudes of the crosstalk tracks using Visual Analyzer. Here is the chart:

Test Record	L->R	L->R	L->R Crosstalk	R->L	R->L	R->L Crosstalk
Toshiba 1	-18.18​	-44.9​	-26.72​	-42.81​	-17.51​	-25.3​
Toshiba 2	-18.27​	-49.4​	-31.13​	-46.59​	-17.67​	-28.92​
Toshiba 3	-18.23​	-46.87​	-28.64​	-50.01​	-17.71​	-32.3​
Toshiba 4	-18.02​	-42.44​	-24.42​	-49.56​	-17.7​	-31.86​
Denon	-11.57​	-33.83​	-22.26​	-47.17​	-10.88​	-36.29​

I know you like the Toshiba record, but unless I'm doing something wrong in VA it doesn't seem like there is any more consistency with these as any other record where I have multiple copies. What is odd, however, with the Toshiba record is when I look at the "no signal" channel on each test track, that channel shows essentially nothing on the scope - not even a hint of a sine wave. It's like the low amplitude channel has literally no signal on it. So it makes the extreme negative values almost (to me) meaningless as no signal is really being measured. I'd love to get your thoughts on this. And it should go without saying, look at the very different readings from the Denon record. All done with periphery ring on each record and Stillpoints LPI v2 record weight. All records otherwise appear flat when rotating on the platter before the ring goes on, although yes I wish I had vacuum hold down. But I honestly don't think that would suddenly makes all these experiments we've been doing come to any happy conclusion.

Cartridge?

 

[tony22]

Using A bandpass 400-3000 removes both low end runbish and clicks

Balle, have we ever discussed the phase issue? It seems in your exhaustive efforts on this matter that I have not seen whether you feel it is important to try and get very close or matching phase values (e.g. to ensure the generator is optimally aligned).

 

[microstrip]

Visual Analyzer displays the spectral content as well, in a separate window below the scope window. So you go for peak spectral amplitude at 1KHz, or matching them, L-R and R-L?

I match for equal residual divided by peak for both tracks - in practice equal residual with most cartridges,

 

[tony22]

I match for equal residual divided by peak for both tracks - in practice equal residual with most cartridges,

I'm not sure I understand this. Apologies.

 

[microstrip]

I'm not sure I understand this. Apologies.

I divide the residual value by the full signal value to get a normalized value. As in fact the full signals are almost the same I am mostly comparing the residuals.

I would apologize for not being clear if this was not a characteristic of AI generated posts!

 

[Balle Clorin]

Balle, have we ever discussed the phase issue? It seems in your exhaustive efforts on this matter that I have not seen whether you feel it is important to try and get very close or matching phase values (e.g. to ensure the generator is optimally aligned).

I is A lot of work to get the phase, but I gave that too somewhere, in ASR “Fun with vinyl measurements”. Much easier with AMagiknthat gives phase and db in one go….

 

[Balle Clorin]