Why is reading crosstalk so inconsistent across the different tools?

[tony22]

There are hardware based tools, like the Foz, Foz 2, oscopes, and the new MoFi phono preamp, and there are software tools like Adjust Plus+, AM v1, and AM v2. Some measure only crosstalk, some crosstalk and phase. But at least so far IME (using a number of these HW and SW tools), I find that the results are hardly in consensus agreement with each other. The question is - why? Shouldn’t any of these techniques yield results that, if not in agreement in magnitude, at least be in some agreement with regard to the difference between R->L and L->R? Isn’t the math ultimately (or shouldn’t it be) the same? To say it’s frustrating is (for me at least) putting it mildly.

 

[mtemur]

I get the same numbers from AM V1 and V2 also with Fozgometer but I don’t have it anymore. I digitized azimuth tracks and inspected on Izotope RX a couple of years ago. The results were the same with other methods, maybe minor differences that’s all. But I can’t claim they always match maybe there are other factors I can’t think of now. On the other hand test discs result differently depending on cutting and pressing variations.

 

[tony22]

Yes, I get similar numbers with AM v1 and v2, but vastly different numbers with Adjust Plus+. What’s troubling is the matching. I can be within .5dB across channels with Adjust Plus, but then with either AM the delta will be on the order of several dB. That just doesn’t seem right. I can’t recall what the results were with a friend’s Foz.

 

[mtemur]

I don’t know why adjust+shows different numbers but confirmation of AM results with Izotope RX satisfied me.

 

[tony22]

And you recall that AM was consistent with the Foz?

I sometimes wonder if checking the stylus for perpendicularity to the cantilever, or its angle offset there of, and adjusting azimuth so the stylus drops into the groove precisely perpendicular is a more correct way to deal with this adjustment.

 

[drrsutliff]

My AM v1 was consistent with my Foz as long as the Foz had a fresh battery and recently calibrated which I did most times before I used it. I never had Adjust Plus to compare but with AM v1 I can consistently obtain less than 0.5db across channels. My Foz is a permanent loan to my audio group so have not rechecked recently.

 

[tony22]

Here’s what’s bothering me. JR has a discussion on his site about azimuth and crosstalk, where he mentions if you want to do it right, follow Mikey’s method (to which there is a link). Mikey’s method is to use the Ultimate Analog tracks with a DVM (and ideally a band pass filter) to measure at the speaker terminals. Alternatively, he mentions that a Foz, or AP+ when used with the Ultimate Analog disk, both yield consistently similar results.

I use AP+ with both Chris Frickert’s disk or the Ultimate Analog disk and get crosstalk measurements across channels within .4 dB of each other. When I use AM V1 or V2 with their respective disks, they both yield results where there’s a crosstalk channel imbalance of several dB. Now, if JR describes a technique and implies that Mikey’s technique is equally good, and Mikey implies that, as well, either a Foz or AP+ will yield consistent results by comparison, doesn’t that kind of mean AM is maybe wrong?

 

[J.R. Boisclair]

If you read my latest blog post on my visit to Agnew Analog in Greece you’ll begin to see what the problem is. In short: lack of control at time of cutting with respect to cutting stylus symmetry, stylus insertion and cutterhead angles.

BEFORE you ask why we bother with targeting optimal playback alignment targets at all given the obvious slop in test records, read the blog to the end.

 

[tony22]

If you read my latest blog post on my visit to Agnew Analog in Greece you’ll begin to see what the problem is. In short: lack of control at time of cutting with respect to cutting stylus symmetry, stylus insertion and cutterhead angles.

BEFORE you ask why we bother with targeting optimal playback alignment targets at all given the obvious slop in test records, read the blog to the end.

JR, I appreciate your blog, but I think in the case of my observation it may not be precisely addressing my worry. If I use one record, the Ultimate Analog Test Record (and I have for purposes of this examination), and I measure crosstalk with every tool I have that is able to use those tracks (AP+, AM v1, and even the Visual Analyzer scope on my laptop - this last equates essentially to Mikey's method), the channel crosstalk difference is not the same. Sure, I can see fractions of a dB difference being acceptable (which to me is close enough correlation), but when one method yields difference results that are few to more dB larger than the others - using the same record, then something is wrong somewhere.

BTW, some of us were worried about zenith even before Mikey wrote about it.

 

[J.R. Boisclair]

JR, I appreciate your blog, but I think in the case of my observation it may not be precisely addressing my worry. If I use one record, the Ultimate Analog Test Record (and I have for purposes of this examination), and I measure crosstalk with every tool I have that is able to use those tracks (AP+, AM v1, and even the Visual Analyzer scope on my laptop - this last equates essentially to Mikey's method), the channel crosstalk difference is not the same. Sure, I can see fractions of a dB difference being acceptable (which to me is close enough correlation), but when one method yields difference results that are few to more dB larger than the others - using the same record, then something is wrong somewhere.

BTW, some of us were worried about zenith even before Mikey wrote about it.

Oh, sorry! I guess I did misunderstand your question. You found using the same record gave varied results over your measurement methods.

I can only guess: variations in averaging methods across the three measurements is most likely answer, filtering slopes will be different as well. That will clearly change the results too.

Both software methods are opaque as they do not reveal HOW they sample and process the data. For this reason alone, AM and Feickert could never be used in any scientific study. It is therefore impossible for us to draw a conclusion as to which is most accurate.

 

[tony22]

It is therefore impossible for us to draw a conclusion as to which is most accurate.

Agreed. I think using a scope gets a better view of what’s really going on.

 

[mtemur]

I cannot speak for Adjust+ but AM is pretty reliable cause I checked it with Izotope RX azimuth function. The results were the same. It also matches with Fozgometer as @drrsutliff indicated but I don’t have FM at hand to check again.

 

[tony22]

Then something is really weird. I used Visual Analyser and the measured results of the difference between R-L and L-R were within .4 dB of using AP+. AM V1 using the same record gives a R-L and L-R difference of almost 4 dB. My azimuth adjustment for this cart, at least per the rotation of the cartridge as the means to assess it, is less that 1/2 degree rotated from horizontal. With the methods that are a close match, I get crosstalk numbers of -35.7 and -35.4 (with AP+) and phase difference within 10 degrees, not quite as close but within .4 dB when checked with Visual Analyser.

So how can Izotope RX be in line with AM, but a software oscilloscope package be in line with AP+?

 

[mtemur]

Then something is really weird. I used Visual Analyser and the measured results of the difference between R-L and L-R were within .4 dB of using AP+. AM V1 using the same record gives a R-L and L-R difference of almost 4 dB. My azimuth adjustment for this cart, at least per the rotation of the cartridge as the means to assess it, is less that 1/2 degree rotated from horizontal. With the methods that are a close match, I get crosstalk numbers of -35.7 and -35.4 (with AP+) and phase difference within 10 degrees, not quite as close but within .4 dB when checked with Visual Analyser.

So how can Izotope RX be in line with AM, but a software oscilloscope package be in line with AP+?

I don't know and I wonder how can FM also be in line with AM and Izotope RX.

 

[tony22]

It’s very frustrating. I’m thinking the least ambiguous way to do it is with reading the data directly and just doing the math and subtraction personally.

 

[mtemur]

This is what I was getting with Izotope RX with AM test disc and ZYX cartridge when I recorded azimuth tracks. Results were in 0.5dB in line with AM V1. I wouldn't bother with Adjust+ and oscilloscope. Just my two cents.

 

[tony22]

Sigh. As I said , frustrating.