Your opinion about wow&flutter on walkmans

@CDV Your measurements have confirmed what has been stated before by @Simon Green: the original belts that came with these walkmans were better than what it is available today.
And there's also a significant variation between samples, if you were to choose from a pack of 10 identical belts, you will likely find a couple that are close to the original spec. However, if you buy only one, the luck may not be on your side.
That also explains the insanely low figure for the CFM-130, it simply comes down to an almost perfectly cut belt, because that unit does not have a big flywheel, nor capstan servo or other fancy stuff.
I think it's probably impossible to make belts as good as they were back in the day, if they're not made at an industrial scale, especially those thin 0.5-0.8mm thickness ones that are used in a lot of walkmans.

About the WRMS specs and DIN, in my measurements DIN was always more than double the WRMS value. If we let the weighting aside, the relationship between RMS and peak-to-peak value of a sine wave is: peak-to-peak ~= 2.8 x RMS.
Given that DIN is a +/- value, that represents a peak-to-peak measurement: if we take the service manual specs of the RS-TR333: +/-0.2% DIN is almost exactly 2.8 x 0.07% (which happens to be 0.196%).

I would be curious to compare WFGUI measurements with a dedicated hardware wow&flutter meter. So if someone has a harware meter, please do some measurements on the same unit with same test tape and post the results.

 

@Valentin, I suggest you read my article on Medium. I spent quite a lot of time researching it with more than two dozen sources, cross-referenced. I am not an electronics engineer, and my math skills have long faded, so if you find an error please do not hesitate to alert me - article on Medium are editable

Scroll to the bottom and you will see:

• Effective value or root mean squared (RMS) value: for a pure sine wave it comes down to 0.707 × peak value
• Average value or mean value: equals to 0.637 × peak value for a pure sine wave.
• Peak to peak value: 2 × peak value for a pure sine wave.

On this picture you can see that peak to peak = 2 × Δf, and only either a positive or a negative portion of a cycle (Δf) is used to calculate wow & flutter number. So, in ±0.17% DIN, the number 0.17 is Δf/f, not 2Δf/f. You can see from the picture that this is a 4 Hz wave, it is the frequency where W&F number is the highest and is not adjusted for weight. 4 Hz - highest weight.


WFGUI works fine. It is hosted and being used by a guy who is well-known in tape community as one of the bearers of the analog audio banner. The app is developed by another guy, but is cross-tested against other apps and hardware meters, just read the user's manual to the app, it is very short.

The disparity between WRMS and DIN values - sometimes it is 0.7x, other times it is 0.5x, other times it is 0.3x - can point to a particular problem. I guess, higher difference between WRMS and DIN signals about short but powerful spikes, which are smoothed out by JIS standard, but picked up by DIN standard. These, probably irregular, short spikes may indicate... what? Problems with motor? A sticky testing cassette? You should know better, as you work on many walkmans, I am just an amateur

As for new belts being not as good as the original - this may be true. On the other hand, on two of the walkmans I have replacing the belt before the original one has disintegrated improved W&F. Also, Marian's belt seem to ensure about 0.03-0.05% better WRMS than thicker belt from another supplier, and this may be crucial if the number is around 0.2% WRMS - it can make or break the listening experience.

 

Just watched your video on YouTube @CDV .:
Thanks for creating it, very insightful and great companion to the Medium article.

Edit:
never mind, I noticed you posted this link already in another thread yesterday.

 

@CDV I have read again you article on Medium: it's great, probably the best wow&flutter reference to read out there.
There is nothing wrong with it with one minor exception: the statement that the DIN value should be about double the WRMS value. Don't know where that statement comes from, if it's taken from somewhere else or it's a deduction made from empirical observation.
After doing multiple measurements on different units and also taking the WFGUI picture from your article as a sample as well, it became clear to me that the ratio betwee WRMS and DIN is NOT a constant number and will vary depending on the shape of the speed variation waveform.

Let me explain more clearly: as I stated in the previous post, for a pure sinewave, the ratio between RMS and peak-to-peak is 0.35 (it will be 2.8 if we put it the other way around, peak-to-peak over RMS).
Of course, it should be pretty obvious that, in practice, the speed variation waveform (that is displayed in WFGUI) will, in most cases, be different from a pure sinewave and will be more or less random like.
Below, I attach the generic formula by which RMS is calculated (independent of waveform shape). It's obvious that the result of the integral is dependent of the waveform shape and will only be a fixed value for some particular waveforms, like a sinewave.



The ratio between peak-to-peak and RMS is called crest factor. Given the crest factor varies, the ratio between WRMS and DIN measurements will also vary. https://en.wikipedia.org/wiki/Crest_factor

Below I attach multiple measurements which are all done with the same standards, so the weighting can be taken out of the equation as a factor, because it's the same on all measurements.
I stated both the ratio between RMS and peak-to-peak values and between peak-to-peak vs RMS, in order not to confuse you, because you're probably used to see the RMS as a function of peak-to-peak.
It can be seen that in the picture where the waveform closely resembles a sinewave (the one from your article), the ratio is close to the 0.35 mentioned above. Of course, these are only a handful of measurements and this ratio can have more variation than what's seen here.

My conclusion is this ratio varying between units does not indicate any problem, but just individual variations between mechanisms, mechanism designs and servo circuits.
For example, DD units tend to have more flutter (in the first 3 pictures that can be seen as small but frequent variations) than wow. The flutter will be even more dominant in DD units with quartz lock.
The particular microscopic imperfections of each motor pulley, flywheel pulley and belt cut will create a different wow&flutter waveforms.

EDIT: Something that is also worth stating is: the wow&flutter numbers alone will not tell the whole picture and the waveform shape is also important. The most noticeable will be the slow variations, especially on music with long constant tones.




What you say about very short spikes is a valid perspective (because those short spikes will change the crest factor), however we should see those on the waveform. Note that the DIN and WRMS values are mathematically calculated based on the displayed waveform,
so it would be impossible that the peaks are displayed in the numeric value but not on the waveform.

As in regard to WFGUI, indeed it is working as it should, as it's based on mathematical calculations (those are always precise). The thing I was questioning was the system as a whole, including sound card ADC, interface, driver, CPU, OS.
These all add delays to the system, which may not be able to capture fast transients. However, this is likely not an issue.

About the belts, thinner belts always work the best where the mechanism was designed for that. However I did seen variation between different samples and even instances when replacing the belt increased the wow&flutter figure significantly.
I'm not stating that Marian's belts are bad (or any other brand), in fact his belts are my number 1 choice if they're available for a particular unit.

 

Well, I said that it is about double of WRMS. Yes, it is empirical, I thought it was clear from the article. If not, I will update it to make it more clear. I think the one thing is always guaranteed that RMS is numerically no higher than Q.Peak. (Is it true?) On the other hand, these values should not be directly compared, it is like comparing length and area. Peak or Q.Peak is like length, it is a spike of the amplitude with disregard for time, while RMS is calculated over at least half a cycle, and in case of JIS it is calculated over a 5 s period, so it is more like area (well, integral is area under graph, duh). But like with length and area, let's say a disc's radius and its area, we can say that the larger the radius the larger the area - this relationship holds. Only in case of disc (or in case of pure sine wave) we can use a precise formula, but in case of real life W&F measurements we cannot.

I mention this in the article as well. There is a formula at the end of the article, and I do mention that it works for pure sine wave only.

Right, and I was stupid enough to not do an inverse. Thanks for mentioning this!

There are test files around that you can download, they are generated with frequency generator. For a 4 Hz modulation frequency WFGUI shows WRMS = 0.7 * Peak. In this case Q.Peak = Peak. Of course, for real life data this won't be the case.

I do not have much experience with repairing players and decks, but it seems to me that a bigger difference between WRMS and Q.Peak indicates short and more violent bursts, or it may indicate bursts that happen with the regular frequency, but the duration of the burst is longer, like a burst can take 0.05s or 0.1s, but occur with the same frequency of 4 Hz. So, it seems to me that the closer WRMS and Q.Peak, the better. And the closest they can be is WRMS = 0.7 * Q.Peak. The numbers I see in most specs are about 1:2, sometimes 1:3.

Q.Peak (DIN) is considered more useful to uncover issues with the mechanism that cause micro-bursts of speed, while RMS (JIS) is more forgiving. RMS changes less in time, so it is easier to pick a value and put it into the spec, while Q.Peak jumps around, so are you going to quote the highest value, even if it happened just once in half an hour? Maybe it was an outlier. I guess, this uncertainty about which value to choose AFTER you already performed a test resulted in 2-sigma approach, which I don't mention in the video, but mention in the article. It is a statistical approach, which throws away bursts that happen less than 5% of the time. But who is going to use this method for comparison now, when only half a dozen (at best) manufacturers still make decks? OTOH, vinyl is on the rise, so maybe there is some use for this method there.

It depends on the sampling frequency. Same Nyquist theorem that says you need sampling frequency being at least twice higher to capture the true shape of a wave. Although for W&F, which is limited to 200 Hz or so, capturing fluctuations with 400 Hz is peanuts, so yes, we should see spikes if any on the waveform.

TL;DR: I think we are generally in agreement

EDIT: the machines with 0.04% and 0.08% WRMS from your testing - what are these? Are these walkmans or a component deck? Can you tell make and model?

 

Thanks guys ... even here reading there is to learn .. very well

 

@CDV A very comprehensive response !

In regard to WRMS/peak value, it's up to you to decide if the article should be updated or not. My position is that the ratio between the two should not necessarily be a particular value in order to consider a unit good in terms of wow&flutter.

I do agree that short pulses with high amplitude will influence the DIN value more. Such peaks will still influence the RMS, but not as much, as the area under the graph for such short peaks will obviously be small.
At the end of the day, it is up for debate if such peaks are audible or not, probably a debate that existed back in the day, otherwise the Germans would have adopted one of the other standards and not create their own.
I have not seen such short but high amplitude peaks in my measurements, so maybe if we do more measurements with print screens and post we can see more info and draw a conclusion about this potential problem.

To be honest with you, I never looked at the DIN value, to me if a unit has a 0.08% WRMS spec and I obtain that after the repair, I consider it perfect. Of course, the debate is in regard to units that do not have a spec (that is why I started this thread in the first place).
I will be posting some mp3s with the same recording played on different units (a blind test, not saying which is which), as I'm curious what your and other members' perception will be about wow&flutter. This way, a more objective assesment can be made.

About the digital capturing, I agree that the sample rate is unlikely a problem. Signals captured, in my experience, don't have sharp rising/falling edges, so the harmonics count can't be that high to exceed the 20kHz bandwidth of a sound card.

The units with 0.04% WRMS and 0.08% WRMS are both walkmans: a WM-D6C and a WM-D3, both completely restored.
Also to be noted the unit with 0.42% WRMS does have a problem. I will post a separate thread about that and the solution when I'll fix it.
The unit with 0.28% WRMS has that value because of the belt, it has increased from 0.2% (original belt). I will search for a better belt and post a separate thread as well.

 

As I said, made a recording on a cassette (the recording was done a unit with very low wow&flutter, to exclude it from the equation) and played the cassette on 2 walkmans with different measured wow&flutter figures.
I chose this kind of music specifically because it's very fast and does not contain long constant tones. Both playbacks were recorded on the same PCM recorder.

The question is: what is your perception about wow&flutter in these 2 recorded files. I would like as many responses as possible and then I will reveal which is higher and the measured values on unit A and unit B.
1. Which has the higher figure, A or B ?
2. What do you think is the difference between A and B ?

NOTE 1: A slight skip somewhere in the middle of the track is caused by skipping to a different part of the track, so ignore that. Also the 2 playbacks are not volume matched.
NOTE 2: As mentioned at the beginning, the track was not chosen based on preferences, but rather because it's very fast, so slow variations in pitch are harder to detect.

Of course, ideally a walkman should have a w&f value so low, that even in the worst case scenario with very slow music it still perfoms acceptable. However, it is my opinion that an average between the 2 extremes should be taken as a reference and not the worst case scenario.
This can be achieved by a combination of my results and CDV's.

EDIT: As @CDV suggested, please give me your responses via private messages.

 

I suggest collecting replies via private messages to avoid dependance on others' answers.

 

@CDV : your expectations of W&F of the belt driven walkmans might be a bit on the high side.
As example, besides @Valentin ’s factual results of restored walkmans, I attach a screenshot of the WM-701C (super) walkman results l from @walkman archive . These are from a carefully serviced unit.

Here is the link to the entire article:
http://walkman-archive.com/gadgets/walkman_sony_05_701c_v3.htm

 

@Emiel, without explicit mention what kind of measurement this is, it is hard to judge. If it is unweighted, then it can easily be several times higher than weighted ("WRMS").

My expectations are simple: a more expensive and sophisticated walkman should perform better than a cheap plastic one.

 

You are right, but with the 701C and it’s relatives you have hit the jackpot, not cheap at all Let’s ask Hugo and doctor Walkman what these values actually represent.
We’ll also be getting values of the DX100 in a few weeks, as Valentin is servicing mine. It does require a bit of work, since I did not use at all in since the purchase in 2010.
That player was close to 650 Dutch guilders (without inflation 290 euros) at the time of release in ‘91-92. Sony did not have many other more expensive portable playback only devices then.

 

I for one will be interested in learning how S2G-ers rate this.

As for my uneducated self, I do not care about techno and listening thru my 'puter' speakers I kinda liked A better:
probably because it has more W&F and made it sound more 'natural'... As my favorite audio-Guru (Art Dudley, of Listener fame, then with Stereophile) once said: "I Love Distortions!"

Going to listen through my Sennheisers and Pono Player tomorrow, but, ... maybe... record some PF "Echoes" or maybe Schubert so that old-farts like me would actually had a chance to relate to the tune...


NOTE: post edited, I must have spent too much time at Schiit Audio site and slipped into that same 'shit-crap' language

 

I did that, accidentally, for El Diablo:



I did get service manual for the Leader, but as far as I remember calibrating W&F meter was out of my league. So, I used Nakamichi LX-5 freshly restored for me by Willy Hermann to check my own measurements. The results, in pictures:

 

I provided my response 1:1, and am curious to find out what the outcome is.

 

@Jorge, Willy Hermann reported 0.067% weighted peak. Peak value according to DIN standard is reported as "plus-minus". On the other hand, his number corresponds to your JIS WRMS number. Usually - in practice, not according to formula - W.Peak is two-three times higher than WRMS. Therefore, what he measured was WRMS, not W.Peak.

So, willfully or not, he misled you into believing that your deck has spectacular performance, while it barely meets the manufacturer's specs: less than 0.06% weighted RMS, less than ±0.11% weighted peak. Anyway, restoring a forty-year old deck close to manufacturer's specs is a job well done.

You may want to remove the first picture or at least blur your name and address.

 

Here is a plot with W&F specs of cassette decks from 1993 Stereo Review buyer's guide. Sometimes W&F was reported as RMS, other times as WRMS, yet other times just a number, so take with a grain of salt. But I presume that WRMS means JIS weighted measurement.

Two decks for $900 and $1250 are Bang & Olufsen Beocord 4500 and Beocord 7000. I can only hope that these are DIN numbers, so WRMS should be around 0.04%, much more respectable. The $300 deck with 0.12% WRMS is Fisher CR-W981, the $250 deck with the same W&F is Fisher CR-W781.

Anyone wants to plot a number for the TEAC W-1200?

 

I created my own small test too to complement Valentin's. See the link to Google Docs.

Four different machines play four pieces from Grieg's Peer Gynt, plus one piece by Sarah McLahlan. I created five folders, each folder has the same fragment played on each of the four devices. Some of the machines do not have Dolby and some of the machines are mono, so I converted everything into mono and did not use Dolby decoder.

The goal of the game:

1. Rank files in each folder from best W&F to worst W&F. For each folder, specify the file with the least wobble first, worst wobble last.
2. Mark the files that you consider unacceptable for this musical fragment. There are different fragments, so some fragments may play better or worse on lesser machines.
3. Feel free to write any additional comments.

For example, an entry for one folder in the report card may look like this (your ordering may be different):

• 01
  • A1345E - (this is, like, as good as digital! No W&F whatsoever!)
  • 5BCFD1 - ok
  • 14ADD9 - barely tolerable
  • E32235 - intolerable (absolute garbage, my ears bleed!)
• 02
  • ...

Repeat the above for all five folders. Again, the most important is ranking for each of the folders.

Please, pay attention to wobbly sound only, do not account for noise or frequency response or playback speed (one machine clearly has a different playback speed, try to assess wow & flutter only).

SEND THE RESULTS TO ME USING PRIVATE MESSAGE.

I want to gather at least ten responses. I will be updating this message with the number of people responded.

To speed up the voting, I am going to summon a bunch of people who posted in the last several days: come on, guys, please listen to the moozik and vote in Valentin's test and in mine!

@Ayupchap, @Mister X, @Ridvan, @Valentin, @RTB, @Emiel, @Billybrass1, @Cameron, @capStan, @autoreverser, @Philip Taylor, @Lordusher, @Jorge, @Silver965, @TooCooL4, @Longman, @Ken80s, @DrZoidberg, @Easthelp, @Radio Raheem, @Zwaantje ... I got tired

P.S. The experiment has flaws. In particular, clipping seems to have been caused by one of the machines, not by hot levels during recording. Anyway, this is all the equipment that I have.

Here is Out Of the Shadows on YouTube, used in Test #5:

 

Thanks @CDV , I'm not used to getting homework here but this task I'll for sure accept.
I'll try to complete it before or in the weekend.

 

I am not a big fan of blind/double-blind tests: I hear what I want to hear and BeeGees played through a cheap bbox sounds better to my ears than if played on my main stereo. Simply because that is how I listened to BeeGees when I were in my teens, when I Actually Cared for this kind of music...
I must have mentioned this elsewhere, but for some tunes I prefer cassettes recorded from CDs to the original CD/PerfectSoundForever sound, my only explanation is that a bit of extra W&F added to the surgically-Perfect sound makes it sound more Natural to me... Silly Me!!!!

on the technical side: I think that ABX test might be more appropriate, ABX selects out some folks with compromised hearing. But I believe I must have made my point of Not looking at the specs one time too many! For those who missed my ramblings, go here

for the folks who do not bother with such nonsense, PLEASE keep in mind that in each and every such test there Must be a Negative Control !


For the record: I might have been duped by Willy Hermann,........ Good Point... I have no comments to that!!!...