advertisement


Wow and flutter 'meter', etc

West sea, I've seen the effect on one Paul's polars it clearly reduces the height of the ripple that comes from the motor poles, ie 12 per rev. There might be a belt slip/grip improvement in there as well. Hard to say.

Too viscous and it goes bad, theres an optimum thickness for sure.
 
In case it helps / for the sake of comparison I ran the GL58 recording though my W&F prog here and got this result
http://jcgl.orpheusweb.co.uk/temp/WF-GF58.pdf
The varions seem alarmingly big! Have I got a scale factor wrong somewhere? My test files seem to give expected results, but it implies the GL58 is awful!

BTW If you increase the number of points plotted in the polar plots they can cover a number of rotations and this can then indicate if the rotation speed (or tone frequency) are off.
 
Jim

Very interesting

I've been thinking about the plots quite a bit to try to clarify in my own mind what would be useful. Realistically the worse W&F result is say 2%. But a good TT should be more like 0.2% max. So I think we need to be able to show that difference in some sort of scale maybe in two ways: The first as your recent Polar but perhaps with 0 rather that 0.5 in the centre. The second would be a polar where the 2% maybe only looks like 10% of the diameter so something more real but exaggerated to make it easy to spot.
 
Jim Audiomisc said:
In case it helps / for the sake of comparison I ran the GL58 recording though my W&F prog here and got this result
http://jcgl.orpheusweb.co.uk/temp/WF-GF58.pdf
The varions seem alarmingly big! Have I got a scale factor wrong somewhere? My test files seem to give expected results, but it implies the GL58 is awful!
Click to expand...

If you run the Cal file through the visualiser you'll see it's a perfect 1% peak. If you now visualise the GL58 it too hits about 1% peak, but the averaged RMS is not quite so shocking. Maybe 0.35%

If your polar plot is trying to map it to the platter rotation then results will be interesting because the idler makes several revolutions for each platter rotation.

Have you tried a polar plot of the Cal file? This will confirm your scaling.
 
Steveh100 said:
Jim

Very interesting

I've been thinking about the plots quite a bit to try to clarify in my own mind what would be useful. Realistically the worse W&F result is say 2%. But a good TT should be more like 0.2% max. So I think we need to be able to show that difference in some sort of scale maybe in two ways: The first as your recent Polar but perhaps with 0 rather that 0.5 in the centre. The second would be a polar where the 2% maybe only looks like 10% of the diameter so something more real but exaggerated to make it easy to spot.
Click to expand...

At present I've presenting the speed so that '1' means bang on the right speed. However I think veusz may allow me to set the polar range from -5 to +5 (percent) with -5 in the center. But to use that I need to alter the values given to the file.

Yes, the aim is to show up 'cogging' against platter rotation angle, which seems a useful thing to do. It also shows up if the rotation rate is 'off' because if we plot a few rotations they peaks shift around the plot from one rotation to another.
 
wd40addict said:
Have you tried a polar plot of the Cal file? This will confirm your scaling.
Click to expand...

I've been using my own calibration files that I generated when I first wrote the original program. However I need to recheck this and tweak the display as described above. May do this today, but should really be doing something else. :)
 
Had a quick poke about and noticed one error. I'd accidentally been using a set of veusz commands that generated a 'log' plot, so distorting the results when displayed on the polar graphs. The 'footer' file needs the line deleted that sets (incorrectly) the radial scale to be log!

I then altered my program so that is gives frequency (i.e. speed) error as a percentage *without* adding a 1 for convenience as Veusz will happily plot a polar with the radial scale from, say, -0.5 to +0.5. Footer duly altered I can now get results like the following

http://jcgl.orpheusweb.co.uk/temp/0p1pc_TwoPolars.pdf

That shows the results of a test calbriation file that alters the frequency by 0.1% +/- peak. Seems OK.

http://jcgl.orpheusweb.co.uk/temp/GL58-TwoPolars.pdf

This replots the GL58 (note the radial scale).
 
Here 'tis.

http://jcgl.orpheusweb.co.uk/temp/1pc-TwoPolars.pdf

Notes: The radial scale is -1% to +1%. The blue blobs are the raw input. The red ones are AES weighted, so tend to be reduced. My program assumes a stereo file that is longer than 10 secs. So I used sox to generate stereo, and concated this to double the length. Quick check with Audacity showed the join is clean.

I also used sox to generate a 48k rate version to check that my program is handling the timing correctly, and this gave the same results, so that seems fine.
 
Just in case it helps I've put up the updated/corrected files for the way my prog generates polar veusz plots here:

http://jcgl.orpheusweb.co.uk/temp/PolarFooter

Is the new footer used. Note the scaling which you can of course change. :)

http://jcgl.orpheusweb.co.uk/temp/veuszproc.c

This shows the modified proceedure that outputs plain percentage values without the bias '1' I used originally to 'fake' polar behaviour.

Overall, the program uses 10sec of audio to get reliable stats and a reliable mean frequency to base the percentages, etc, on. Note that if you get 'out of range' warnings this just means some values fall off the histogram range. They still contribute to the stats analysis, etc. Now back to finding files and photos from 1998-9!... :)
 
No problem!. Needless to say the GL58 isn't my main Hi-Fi deck, that's a 301, but it's my test bed for ceramic cartridges etc.

One of the problems of making these measurements is the quality of the test discs. All my tests have used LXT5346, but the punching of it's centre hole probably isn't perfect.

Her's two more files first the direct drive AT-LP120, then the mighty Collaro 4T200 with its 8lb platter. The traces displayed by the visualiser are virtually identical:

https://www.dropbox.com/s/qnu90ul3k27n8wu/ATLP 3kHz Mono.flac?dl=0

https://www.dropbox.com/s/7c25m3wgh58l4f1/4T200 3kHz mono.flac?dl=0
 
Thanks. :)

The method I use doesn't care too much about the tone frequency. So you may be able to cheat a bit and use tones on various discs at various diameters from the nominal center. I've not experimented but suspect pretty much anything around a few kHz will be fine. If you can use two bands at different diameters without moving the disc then the change in basic once-per-rev wow would indicate where the center/offset lays.
 
Here are results I get for the ATLP and 4T200 files.

http://jcgl.orpheusweb.co.uk/temp/4T200_TwoPolars.pdf

http://jcgl.orpheusweb.co.uk/temp/ATLP_TwoPolars.pdf

The both seem to include a slightly 'frizzy' section per rotation (approx) which makes me wonder if the test LP is noisy?

Before giving the data to the W&F prog I used sox to amplify the content by about 15dB as the level of the recordings was low. Poor SNR though will tend to risk 'added counts' and lift the measured frequency. For accurate results the signal needs a good SNR because the program only does minimal filtering to reduce rumble.
 
Not sure what's causing the fuzzyness. The visualiser starts with a limiting amplifier so not too fussy about level.

The plots are interesting and clearly show the hole is slightly off centre. This is a real issue when trying to measure W&F. I don't know if any of the modern test LPs come with guaranteed centreing.

At the end of the day the W&F an LP user hears is likely to be dominated by this effect rather than the turntable's inherent W&F.

Looks like Nakamichi had it right with their TX-1000!
 
Looking again, I suspect the 'fuzz' is much the same all around the platter rotation, but is 'hidden' by the AES weighted plot on one side due to the once-per-turn wow. So it is probably the noise level. The AES results are a better guide for audible wow and flutter anyway.

Yes, in reality once the turntable wow level is modest it will vary from one LP to another. But as per above, if you need to disentangle, measurements at more than one track diameter can allow this.

In principle we can overlay a set of rotations to improve the SNR of the results. But that then requires accurately knowing the actual underlaying rotation rate and ensuring it is stable. If doing this as a pro job it would mean using something to 'clock' the turns like a small mark on the side of the platter than triggers a sensor pulse on each rotation, and recording this on another track symultaneous with the LP tone.

This is an area where a DD turntable with a strobe tends to score as you can lock or adjust the rate accurately enough to combine rotations.
 
Apologies, Paul! I've been distracted by sorting and working on something else, so I hadn't made a full updated source available.

I have now, at

http://jcgl.orpheusweb.co.uk/temp/NewWandF.zip

This contains my current source code and a copy of the PolarFooter used for the veusz output.

Note two details. Firstly, it now doesn't list every time a value falls off the range of the histogram plot. Just summarises the count of these if they happen. (None of the stats are affected, just the histogram plot.) Secondly it now lets the user choose how many rotations to plot on the Polar Veusz file. Normally, the user should just say '1.0' and get one full rotation plotted as this avoids multiple turns overlaying and making the pattern hard to see. But plotting many turns lets you see if the pattern is consistent and/or if the rotation rate is fast or slow. (The plot assumes exactly 33.3rec RPM.)

It limits the number of rotations to 5.0 to avoid accidents. But the user should ensure that this value plus the start time don't run you off the end of the input file!

I've been meaning to add other details to the polar plot so it shows the input file name, the rms W&F and tells the user that the red blobs are raw and the blue ones AES weighted, so haven't updated the complied full ROX app yet. Aim to, but want to get on with something else.

Yes, I expect stereo input. I record / capture stereo by default. It also means I can do a frequency measurement on each channel and then average them together. This can slightly improve the accuracy if the noise on the two channels is uncorrelated or antiphase.

Hope this helps.
 
You must log in or register to reply here.


advertisement


Back
Top