GoodSoundClub - Romy the Cat's Audio Site - Only God knows what it all means.

Analog Playback
Topic: Only God knows what it all means.

Page 1 of 1 (18 items)

Posted by Romy the Cat on 05-24-2012

I got today an interesting emails from a site visitor pointing me to UK guy who user FFT analyses to evaluate stability of TT palters. This is very interesting as it gives very objective data about performance platter, suspicion, bearing, and drive topology. It does not give information how the proximity to the FFT perfection impact sound. I mean it is not the fact that the TT with best 3150Hz stability will sound “better” all together but it for sure give SOME idea of what is going on.

http://www.pinkfishmedia.net/forum/showthread.php?t=70027

The Cat

Posted by Paul S on 05-24-2012
fiogf49gjkf0d Very interesting. I see there is FM demodulation prior to the FFT, however. Recent experience with PWM (pulse width modulation) seemed to show that this process, in itself, smooths things over. When I read the header I hoped someone had focused some sort of superfast 4D measuring device on the TT. Now that I think of it, how would this work? Paul S

Posted by N-set on 05-26-2012

Actualy I was going to point into that thread where I was partially involved providing (i'm phixphi there) my EMT930 test signals for analysis but was a bit shy to do it. One of my "big" hopes is that perhaps Paul's softwould allow to capture some objective characteristics of TT sound, like i.e. spectral analysis helpsto get some sound characteristics of e.g. tubes, and other crap. The most interesting are those circular plots,sor tof a "speed portrait" of a TT. The ambitious plan would be to start the hard work of correlatingplots to sound, just like people correlate harmonic spectrum to sound (all that monotonic decay laws etc).AFAIR, what was missing for comparison are "portraits" of serious belt TT's. Anyone want to recorda test signal with his big belt TT and put it up in the thread for analysis (beware: it can do more harm to your ego than a timeline)?

Posted by Paul S on 05-26-2012

Oh, good, it's you, N-set. This is a COOL idea! Maybe you can set me straight.

The way I see it, to be significant, the full-spectrum sound recording alone would pose its own challenges, for sure. And what are the parameters for linear measuring, and storing measurements, and for extrapolating the data? I wonder what it would take in terms of procedure and equipment in order to wind up with "accurate" measurements in the first place, and then again after a carrier had been added and "removed". Looking at a square wave "post carrier", the top band is typically wide and fuzzy with residual carrier frequency, which I see (and hear) as "smeared". Yeah, NASA gets around this; but not the average audio "lab". Am I just "seeing" this wrong (as I am wont to do)?

Best regards,
Paul

Posted by N-set on 05-30-2012

Hi Paul! I'm sorry but I simply can't make any sense of what you are asking....Anyway, all the software has been designed by Paul R, I had no contribution here(Paul has contacted me asking for sampel of my EMT when the soft was already done)and is based on this idea:
http://audiamorous.blogspot.com/2009/02/some-thoughts-on-digital-wowflutter.html
so all the tech Q's are best sent to him. I did not have time nor free resources to go in depthof the method. I'm more interested in potential results it can provide.
Again, anyone with a good high mass table interested in recording a test tone and postingin the pinkfishmedia forum for analysis?
Cheers,N-set

Posted by N-set on 06-11-2012

N-set wrote:
AFAIR, what was missing for comparison are "portraits" of serious belt TT's. Anyone want to recorda test signal with his big belt TT and put it up in the thread for analysis (beware: it can do more harm to your ego than a timeline)?

I haven't been to the PinkFish thread for quite a while until yesterday. I see that Romy has kindly provided test signal from his SX8000.A big surprise! Why does the speed stability look so poor? I thought it would be rock-stable and we'd all take our small TT toys and go home.Well, it seems not. In the demodulated spectrum there is a whole comb of components. The group around 8Hz is probably cart/arm resonance.The polar plot has some clearly visible low freq. speed modulation (waviness) of unknown source (belt??) as well as thewhole figure is a bit elliptic, suggesting eccentricity somewhere, maybe both things are due to the test LP.As Paul R suggested in the thread, it'd be nice to repeat the test with some better known (i.e. already analyzed several times on different machines to see what is the LP's input) test LP.It seems that the best studied one in the thread is the Ultimate Analog (I also have Clearaudio and some of my tracks were analyzed with it,being expensive, super thick etc etc, it's still eccentric and even wrapped), although it's far from perfection.
Cheers, N-set

Posted by Romy the Cat on 06-11-2012

I have no idea why you are surprised that I submitted my test. Anyhow, I do not know what to look at and how to interpret what I see when I analyze the result of this test. The feedback that Paul R forwarded to me did interpret the result differently then you N-set do but I do not know whose interpretation is more reasonable I have no opinion on the subject.

This test is very interesting but it still has from my view 4 main questions:

1)    How what we measure by this test impact sound. The irregularity lower then 0.1% are not auditable and with all TT that they measure we are in .002% realm.
2)    If one TT measure 0.5% with horrible result and another measure 0.0005% with perfect result and the first TT (being absolutely the same topology presumably) sounds better then what does it mean?
3)    How to plug the subjective or sonic result into the analyses and correlate it with measurable data?
4)    Doing the analyses we compare apples and oranges as 2 TT of different topology have very little in common. I think it would makes senses to do the analyses of this type on the same TT and to see how different action on this TT would impact the analyses.
Anyhow, even I do like that we have such a great tool in disposal as Paul R offers but it did not affect in any way my feeling about turntables.

The Cat

Posted by N-set on 06-11-2012

Romy the Cat wrote:
I have no idea why you are surprised that I submitted my test. Anyhow, I do not know what to look at and how to interpret what I see when I analyze the result of this test. The feedback that Paul R forwarded to me did interpret the result differently then you N-set do but I do not know whose interpretation is more reasonable I have no opinion on the subject.

First of all I'm surprised by the results of the test not by you supplying the sample, of course! Please read carefully.
The spectrum diagram represents the low frequency components of the the rotational speed variation.
There is a certain analogy between THD number and Wow&Flutter number: bare THD tells you very
little, if anything at all, about the amp, one has to dig into that number deeper and see a detailed spectrum behind it.
Those tools of Paul do the same with W&F--this is a detailed analysis of different components
of the speed instability. The circular plot is one smart way of representing the speed spectrogram:
the distance of each point on the plot represents the instantenous speed as a function of the
angle in the course of the rotation of the platter (I think Paul puts circa 5 full rotations on one plot).
In the ideal world the figure should be a circle: a constant speed during the whole turn of the platter.

I'm not sure if Paul R gave you more output than what he wrote on the PF, but I think neither of us really gave
a solid interpretation as it is difficult at this moment: too little experience with the tool. One clearly sees a regular waviness,
there is something happening approx 16 times per revolution. This may be induced by the motor and/or amplified by the
motor-belt interaction (is your belt stiff or soft?).
Do you know at what speed does the motor go? This waviness is also superimposed on an oval rather than
a circle, so this suggests some eccentricity (you can see it at the very low end of the spectrogram, there are some
peaks there; 0.55Hz peak corresponds to the LP eccentricity).

Romy the Cat wrote:
This test is very interesting but it still has from my view 4 main questions:

1) How what we measure by this test impact sound. The irregularity lover then 0.1% are not auditable and with all TT that they measure we are in .002% realm.

Oh, this is a huge question and a big work to be done: correlate measurements to the sonic results. As I see it at this moment the tool is still in its teething fase, identifying merely some "hardware" mechanics in the plots. This is needed to learn the tool, but I hope with time the hard work of correlating with the sound will start (again an analogy to the advent of spectrum analyzers in amp analysis is in order IMHO). As you imagine a time (and a good will of the community!) is needed to establish enough statistics. The statements "irregularities below 0.X% are not auditable" I'd treat rather carefully, just like parallel statements made thoughout the years that THD below X% is not auditable...

Romy the Cat wrote:
2) If one TT measure 0.5% with horrible result and another measure 0.0005% with perfect result and the first TT (being absolutely the same topology presumably) sounds better then what does it mean?

There was an example in the thread of Voyd 3 motor machine showing a rather far from ideal plot (BTW it seems to support your view on multimotor machines: it seems impossible in the real world to synch the motors and keep them synced; look how at different revolutions of the platter
the Voyd speed instability gets shifted in the phase), but still the owner preferred it to the other deck with seemingly better plot. I think
no sane body believes that those plots will tell you everything about the TT and more importantly about it's sonic. Until one starts to correlate
plots with sonics it's difficult to judge, the only thing we have are hypotheses (i.e. the closer to the circle the better the sound), common sense and other intellectual crap.

Romy the Cat wrote:
3) How to plug the subjective or sonic result into the analyses and correlate it with measurable data?
4) Doing the analyses we compare apples and oranges as 2 TT of different topology have very little in common. I think it would makes senses to do the analyses of this type on the same TT and to see how different action on this TT would impact the analyses.

As per 3) this is the question of methodology. One has to develop it. By trial and error.
I do not 100% agree that comparing different topologies is meaningles. Quite to the opposite: instead of intellectual fantasy arguments
(like invoking "laws of physics" and other crap), one can measure different topologies and see how they differ between each other
in reaching what seems at the moment to be the objective: a constant rotation of the platter. Same with e.g. different amp topologies:
why wouldn't it make sense to spectrum analyze a PP amp and a SET and compare?
Obviously, what you propose in the second part has already been going on: perform a different action on the TT and then see the results,
both measured and sonical. People try different PSU, oils, dampings etc etc.

Cheers,
N-set

Posted by Romy the Cat on 06-11-2012

N-set wrote:
There is a certain analogy between THD number and Wow&Flutter number: bare THD tells you very little, if anything at all, about the amp, one has to dig into that number deeper and see a detailed spectrum behind it.

Actually there is not. When I was taking “irregularity lower than 0.1%” I did not mean THD but the rotational irregularity, or Wow&Flutter. I am not familiar with TT Wow&Flutter but in my past I dealt a LOT with reel tape and I know that if we are beyond 0.1% then we are out of dander.

N-set wrote:
I'm not sure if Paul R gave you more output than what he wrote on the PF, but I think neither of us really gave a solid interpretation as it is difficult at this moment: too little experience with the tool. One clearly sees a regular waviness, there is something happening approx 16 times per revolution. This may be induced by the motor and/or amplified by themotor-belt interaction (is your belt stiff or soft?).

This is the whole point – even if we do see the problem we still have no knowledge where this problem came from and is responsible for it. Did the record was cut properly. Did they cut the whole on exact middle? Did the platter during the record cutting had any speed regularity? Oh, this is a huge question and a big work to be done: correlate measurements to the sonic results. As I see it at this moment the tool is still in its teething fase, identifying merely some "hardware" mechanics in the plots. This is needed to learn the tool, but I hope with time the hard work of correlating with the sound will start (again an analogy to the advent of spectrum analyzers in amp analysis is in order IMHO). As you imagine a time (and a good will of the community!) is needed to establish enough statistics. The statements "irregularities below 0.X% are not auditable" I'd treat rather carefully, just like parallel statements made thoughout the years that THD below X% is not auditable... I disagree. The THD are deterrent from W&F as W&F are pretty much liner distortion existing in nature. In THD the order of the harmonics is important and the reason that creates the distortion is important. In case of Wow&Flutter the reason is NOT important but only the fact.

N-set wrote:
I do not 100% agree that comparing different topologies is meaningles. Quite to the opposite: instead of intellectual fantasy arguments (like invoking "laws of physics" and other crap), one can measure different topologies and see how they differ between each other in reaching what seems at the moment to be the objective: a constant rotation of the platter. Same with e.g. different amp topologies: why wouldn't it make sense to spectrum analyze a PP amp and a SET and compare? Obviously, what you propose in the second part has already been going on: perform a different action on the TT and then see the results,both measured and sonical. People try different PSU, oils, dampings etc etc.

What I was trying to say was that comparing the same topology give an explanation WHY the analyses looks like it looks. If you have a TT with analyses done on a given record then you can run the same analyses again but juts for instance loose the belt tension or supplying more air pressure in platter suspension. This single change would give an idea of how the analyses be different in case of one single change was done. Doing it for loooooooong time we might understand the reasons and predict what was responsible for this or that bad analyses.

Posted by N-set on 06-11-2012

Romy the Cat wrote:
N-set wrote:
There is a certain analogy between THD number and Wow&Flutter number: bare THD tells you very little, if anything at all, about the amp, one has to dig into that number deeper and see a detailed spectrum behind it.
Actually there is not. When I was taking “irregularity lower than 0.1%” I did not mean THD but the rotational irregularity, or Wow&Flutter. I am not familiar with TT Wow&Flutter but in my past I dealt a LOT with reel tape and I know that if we are beyond 0.1% then we are out of dander.

Of course we are talking about 0.1% or whatever of W&F! Sorry if I have confused you, but I use THD only as an example: a bare numbers tells you little, you have to see the spectrum behind that number. Getting 0.1% is easy, virtually every reasonable deck gets it,just like getting 0.1% THD is easy. Yet all those <0.1% decks sound different...I'd be careful...

Romy the Cat wrote:

N-set wrote:
I'm not sure if Paul R gave you more output than what he wrote on the PF, but I think neither of us really gave a solid interpretation as it is difficult at this moment: too little experience with the tool. One clearly sees a regular waviness, there is something happening approx 16 times per revolution. This may be induced by the motor and/or amplified by themotor-belt interaction (is your belt stiff or soft?).
This is the whole point – even if we do see the problem we still have no knowledge where this problem came from and is responsible for it. Did the record was cut properly. Did they cut the whole on exact middle? Did the platter during the record cutting had any speed regularity?

Some of those problems can be +/- identified: eccentricity, wraps, etc show up as 0.555Hz fundamental and it's harmonics.
As for the cutting quality of the test LP that's why it would be very educative to record your deck with a known LP, like the Ultimate Analog one.
It has been used in several test so far and seems +/- ok cut, although the purity of 3.15kHz tone is disputable (see some eraly Paul's analysis of this record).

Romy the Cat wrote:

Oh, this is a huge question and a big work to be done: correlate measurements to the sonic results. As I see it at this moment the tool is still in its teething fase, identifying merely some "hardware" mechanics in the plots. This is needed to learn the tool, but I hope with time the hard work of correlating with the sound will start (again an analogy to the advent of spectrum analyzers in amp analysis is in order IMHO). As you imagine a time (and a good will of the community!) is needed to establish enough statistics. The statements "irregularities below 0.X% are not auditable" I'd treat rather carefully, just like parallel statements made thoughout the years that THD below X% is not auditable... I disagree. The THD are deterrent from W&F as W&F are pretty much liner distortion existing in nature. In THD the order of the harmonics is important and the reason that creates the distortion is important. In case of Wow&Flutter the reason is NOT important but only the fact.

I don't know what kind of linearity of W&F you mean? Again, I was building THD <-> W&F anaolgy to illustrate a big step being made in measurements: from a single number (THD or WF) to much more detailed information (harmonic spectrogram and WF spectrogram respectively). No more than that.
I would very strongly disagree that with WF the reason is not important!!! What do you mean? I definitely want to know such reasons ! For example in my case the plot is slightly pentagonal with a bigger dent and a bit elliptic. This means that something relatively small is happening 5x per revolution and something bigger 1x. The only thing that in my case rotates 5x the speed of the platter is the idler. Looking at the spectrogram I see a strong peak at 2.7Hz. Bingo, one of the reasons for my speed variation is identified: the idler (either eccentric or uneven or both). Now I can think of measures to improve
the situation. You get my point?

Romy the Cat wrote:
What I was trying to say was that comparing the same topology give an explanation WHY the analyses looks like it looks. If you have a TT with analyses done on a given record then you can run the same analyses again but juts for instance loose the belt tension or supplying more air pressure in platter suspension. This single change would give an idea of how the analyses be different in case of one single change was done. Doing it for loooooooong time we might understand the reasons and predict what was responsible for this or that bad analyses.

I can't do any better than to agree with the above. But apart from such optimization of a given topology/realization, IMHO comparing different topologies, with a good selection of samples withing each class (belt, idler, DD, ...) has a lot of sense as to give a sort of "topology fingerprint" (again using my THD-WF analogy, each PP and SET topologies have their specific harmonic patterns--their topology fingerprints; same may be here: e.g. belt and idler may have their own topology-specific spectra of WF).

Cheers,
N-set

Posted by N-set on 06-11-2012

So as an example let me present here Paul's analysis of my EMT930 (all the credits go to Paul R).

The objective of this particalr run was to test an apparent audiopedophile(tm) legend that a 78rpmwarmup or preforming of an idler deck helps to smooth out the idler errors (eccentricity I guess).The deck was heroically spinning 78rpm for 4 days, the test LP is Ultimate Analog and the samples were taken rightbefore the warum and right afterwards.

Here is the polar plot (ignore the spikes--dust or static) for a quick glance:

First of all the warmed up and cold plots do not differ much and I even like the cold one morefor being more consecutive from revolution to revolution, so the audiophile rule did not work for me...what a surprise!
Then, one immediately sees that the shape of both cold and wrmedup samplesis pentagonal and the same for all the runs (there are approx. 5consecutive revolutions plotted)--there is some regular speed error 5x per revolution.This corresponds to (approx) 5x0.555Hz=2.775Hz.

Let's look at the demodulated speed spectrum

The strongest component is exactly those 2.7somethingHz (Paul gave the exact number from the datafiles).

Then there is a dent in the polar plot once per rev. I have little idea as to where it comes from,it may be disc eccentricity together with idler errors. This once per rev. error is visilble as (much smaller due tome having a lot of luck positioning well the test LP on the spindle)0.555Hz peak. Actually the deck was serviced by one of the reputable EMT serviceman in Lahr and he did tell me that when measuring WF he saw an increase of it once per rev, but since it was within the 0.075%spec, he has ok'ed the deck (he is using custom made test LP with a brass insert for an excellent concentricityand one of the EMT WF meters, 420 IIRC).

Then there is a 12.35Hz peak in the "cold" spectrum and it's 2nd harmonic in the "warmed up" one.This is most probably the motor (Paul tried to explain it by 3130Hz(the actual freq.)/3150Hz(the master of the LP) *25Hz=24.8Hzbut I did not get this explanation)

So what are the conclusions at this moment? The idler is the source of the biggest error. Right now there is a new original EMT idlerset up by the mentioned serviceman and said to be within the specs, all tip-top. My plan is to restore the old idler (at a fractionof the EMT price), e.g. with this guy

http://www.terrysrubberrollers.com/

and see how it performs compared to the EMT one. Beyond that I have no intent to mess up with the deck alignment again.Another thing is a dedicated power synthesiser to supply the 3phase power to the motor. Then new experimental possibilities open.
Cheers,N-set

Posted by Romy the Cat on 06-11-2012

N-set,

I very well understand what you are saying but I’m VERY hesitant to make any sense from it, not to mention to navigate my opinion from it. I think to use the anises like this in discreet fashion is a bit wrong for the reason I stated above. There are so many souses where the error might come from that it is impossible to say anything definitively analyses you do many analyses from the same setup and introduce each time very minor change. If you look at the site then you will see that the first test I submitted to Pool was huge amount of errors that took place 5 times during a rotation. Looks strange, doesn’t it? It turns out that it was the Lynx cared that I did not slave to Pacific’s A/D. The next text I use Pacific as master clock and the 5-poll error was not there. Also pay attenuation how lower the error if you discard the 5-poll. It was just a different record….

So, only God knows what it all means.

So not forget another important subject. The best sound I even hear from TT was from 1924 Gramophone... and I am sure that it has Wow and Flutter somewhere in realm of 15%-20%. Ironic, isn’t it?

Rgs, Romy the Cat

Posted by Paul S on 06-11-2012

Some time ago I got uptight about the idea of long TT belts, and this is because I put a strobe on one, a long time ago, after I thought I "saw" it fluttering, out of the corner of my eye. Anyway, it did flutter, and a longer one we messed with had a fairly complex "flutter within flutter" pattern, in other words, flutter harmonics.

In the end, of course, it's all in the hearing (or not hearing). But as long as someone else does the work, it's fun to be "vindicated" about the 3-motor gambit, and I'm waiting to hear more about the "big motor" and "DC" stuff, too.

I've used a Sota drive belt for many years now, but there was a time when I only used silk thread or pro recording tape on my Rec-O-Kut.

Paul S

Posted by N-set on 06-12-2012

This is a new tool, so some training is inevitable. Yes, I saw your first sample with the peaks and my first thought was that it was not the deck, as I could not imagine a mechanism which would produce such spikes so perfectly repetetive (ok, maybe a scratch).
Eliminating the test record input---I agree this is not easy, that's why measuring with several test LP's seems a reasonable work around.

Here is another example of my deck but this time with ClearAudio test LP, which I bought for 50Eu
in a hope to get something more concentric than Ultimate Analog...
Thick slab of vinyl, legendary German company, German pressing, and....fucking eccentric and wrapped (the cart goes like
on a rollercoster)!

The elliptic shape comes from the eccentricity (you can also see clearly visible the dent, but it's origin is unknown to me)
To confirm let's look at the demod freq. spectrum

Bam! A strong peak at 0.555Hz and some of it's harmonics--eccentricity (and probably the wrap)
Observe also the omnipresent 2.72Hz peak from the idler. This is well identified and not record dependent
(this characteristic idler freq. was also confirmed by EMT technician).
One also sees again the 12.4Hz and it's 2nd hrmonic, most probably motor related.

So I think with more and more examples and clever thinking (Paul R is the master) one would eventually learn the tool
and start correlating certain processes in the TT drive with the sonic results.

Yes, I listen to mechanical gramophone sometimes
too. Let me put forward a bold hypothesis: the ear/brain is not so much sensitive to the value of WF but to it's spectrum and stability in time.
The analogy in my hypothesis with what we know about THD is obvious.
Anyway, one should keep in mind that all those spectrograms, plots ect are just some representations
of the reality and not the ultimate reality itself. It helps to build the picture but it is not the whole story for sure.
Think e.g. of platter ringing. It also shapes the sound of the TT, etc etc

Belt flattering: very interesting experiment Paul! What would be nice is to try to correlate it with those circular plots: take
a TT which shows belt fluttering and see it's speed variation plots, trying to recognize that fluttering.
How did you strobe the belt?

Would be interesting to know if that's what's happening in Romy's MS. One mechanism is this: the motor coggs and excites belt+platter system
due to too little damping in the belt itself.

Posted by Paul S on 06-12-2012

N-set, you may be aware that back in the day strobes were popular for *other* reasons, so they were "around". We just darkened the room and varied the strobe until we could see the range(s) of belt motion. The motions were visible at different strobe frequencies, hence my "conclusions".

Remember that Romy's platter is HEAVY, and its rotational mass is very high. If I understand correctly, this rotational mass is also (intentionally) high compared to the motor torque. I don't know how to relate to audibility or even relevance the incremental differences shown on the plots, but those spikes appear to be larger than I would suppose from anything the belt could do so "suddenly" to that moving platter, given the relative "influence" of the parts in question. Dang me for even bringing it up, but this sort of repetitive "spike" could be from a "fault" in a bearing. Should be easy enough to check with a stethoscope.

Best regards,
Paul S

Posted by N-set on 06-13-2012
fiogf49gjkf0d Paul, the spikes are digital data processing artefacts, no problem with them. What would be interesting to explain is the waivyness. Is it the test record or the deck (or both)?

Posted by Paul S on 06-13-2012

N-set, for one thing, a certain amount of "flutter" is a normal part of analog recording processes, if indeed the test tracks are "analog" from end to end. I have no clear notion of what any (subsequent?) digital processing itself "looks like", but as I said earlier, the FM part definitely has a sort of "smoothed pulse" effect, generally speaking, and this might appear as "waves" on this particular chart. Also, like I said before, the residual carrier frequency/ies (that/those not entirely scrubbed from the wanted signal) also have a smoothing/masking effect, not to mention they are "present" in their own right, even though they are supposed to be "out of the band" under consideration.

Another question for the expert is, does everything "foreign" appear as a "wave" or as "eccentricity"? If so, then only frequency and magnitude add to these to make the chart, whatever the "cause" of it.

A last question: do the digital spikes generate their own "harmonics", and might any or all of these subsequently "dissipate" into "waves"?

Best regards,
Paul S

Posted by N-set on 02-27-2013

Here is a DIY software assembly for the advanced wow&flutter analysis, designed by edd9000 from pinkfishmedia:

http://teribil-audio.com/2012/06/software-based-wow-and-flutter-analysis/

I have never used the Timeline, but I suspect this soft is much more brutal revealing speed errors
than the latter (the intimacy of my dick's rotation is currently vivisected in the pinkfishmedia thread).
Have fun & many thanks Paul & Ed!

Page 1 of 1 (18 items)