inharmonicity on a rhodes' tine ?

[Mark II]

Alright, I spent some time examine the rhodes sound with MATLAB ®. I used the FFT on a single key, copied a sample with the length of 1 second and transformed it into a spectrum... but I dont want to bore you with all the technical details.

the main question when stretched tuning is concerned is if the rhodes' tines suffer from inharmonicity of the partials. inharmonicity is when the higher harmonics are not integer multiples of the fundamental frequency.

I wrote a little script that finds all the peaks in the spectrum (=harmonics) of a rhodes' tine and its frequencies. and this is the result:

for example:
F1 (second key on my Mark II 73)    

Code Select Expand

   'n#'    'mag(dB)'     ' freq (Hz) '    'var equal temp'    'note'    'var harm'
   [ 1]    [-10.4876]    [    43.7393]    [        3.3961]    'F1 '     [  3.3961]
   [ 2]    [       0]    [    87.4786]    [        3.3981]    'F2 '     [  3.3981]
   [ 3]    [ -2.7309]    [   131.8909]    [       14.2065]    'C3 '     [ 12.2065]
   [ 4]    [ -3.6319]    [   175.6302]    [       10.0459]    'F3 '     [ 10.0459]
   [ 5]    [ -1.5443]    [   219.3695]    [       -4.9686]    'A3 '     [  9.0314]
   [ 6]    [-12.9016]    [   263.1088]    [        9.7845]    'C4 '     [  7.7845]
   [ 7]    [ -3.5731]    [   306.8481]    [      -23.9745]    'D#4'     [  7.0255]
   [ 8]    [-22.6191]    [   351.2604]    [       10.0459]    'F4 '     [ 10.0459]
   [ 9]    [ -6.7262]    [   394.9997]    [       13.2195]    'G4 '     [  9.2195]
   [10]    [-13.0943]    [   438.7390]    [       -4.9686]    'A4 '     [  9.0314]
   [11]    [-11.6204]    [   482.4783]    [      -40.4462]    'B4 '     [  8.5538]
   [12]    [-12.0037]    [   526.2177]    [        9.7878]    'C5 '     [  7.7878]
   [13]    [-20.6433]    [   569.9570]    [       48.0202]    'C#5'     [  7.0202]
   [14]    [-13.0879]    [   614.3692]    [      -22.0772]    'D#5'     [  8.9228]
   [15]    [-33.1867]    [   657.4356]    [       -4.7844]    'E5 '     [  7.2156]
   [16]    [-15.4013]    [   701.8478]    [        8.3869]    'F5 '     [  8.3869]

explanation
n# = number of harmonics, 1 = fundamental frequency
mag(dB) = magnitude in dB of the harmonics (measured)
freq (HZ) = frequency of the harmonic (measured)
var equal temp = variance of the measured harmonic to the next note of the equal temperament in cents (1 cent is a 1/100th of a semitone in equal temperament)
note = next note of the equal temperament
var ham = variance of the measured harmonic to the natural harmonic in cents

Result:
I tuned the F1 3 cents too high (which is pretty good as you all know the game of moving the tuning spring up and down the tine)
all other harmonics are not integer multiples of the fundamental frequency

for all other tines from E1 to H1 it can be said that the harmonics differ from real integer multiples of the fundamental frequency.
the range of variance is from 3 cents to 10, some single harmonics even have a variance of 100 cents (which is a semitone), but as only one single harmonic per tine is that way off it could be a bad tine !?!).
From C2 on the variance of measured to natural harmonics becomes smaller and is neglectable from C3 to C4. I didnt examine the higher octaves yet.

As I am not able to do the same thing on a acoustic piano someone else has to see if the harmonicity of a piano is bigger than the result above in order to judge whether stretch tuning in neccessary (/an option) or not. I hope this is in some way helpfull for answering the question about stretch tuning.

kind regards
Mark II

[Rob A]

So if I am correctly interpreting this data, the required stretch between F1 and F2 would be about 3 cents. That isn't much compared to the chart in the manual, but it's more than I expected. I find that the pitch on the low end varies by a lot more than that as the note rings out and the pickup magnet affects the vibration.

So, I'd be interested in seeing the results of that partial analysis on more samples, so see if there's more variation than three cents.

The thing that's confusing to me is the first row seems to describe the fundamental, so why is there a 3 cent var harm shown? That should be zero, right? I can understand it being off the euqal temperment value, but it can't be out of tune with itself *1 can it?

Now that I look more closely, if I compare the var harm between N=1 and N=2 there is almost no inharmonicity, i.e. the first partial appears to be incredibly well in tune with the fundamental. I think I need a better explanation of the columns, but I appreciate the analysis you've done here.

Actually: I think this just about sums it up:
87.4786 / 43.7393 = 2.0000. So if that's the actual observed frequency of the partials, then you just proved there's not inharmonicity in the first harmonic.

[keysandslots]

Putting the theory of this aside for a moment, has anyone ever played a stretch-tuned Rhodes and non-stretch-tuned Rhodes side-by-each to compare how they actually sound?  Mine is stretch-tuned and I'm happy with the sound, in fact I prefer the stretch-tuned sound to the non-stretch-tuned sound but I've never been able to A/B it because I only have one Rhodes.

Randy

[Rob A]

I'm actually now kind of interested in seeing the MATLAB code. I'm especially interested in whether six significant digits are justified, so I'd kind of like to see the FFT part of it at least. I'm guessing you used a relatively large number of FFT bins.

I think it would also be productive to run this routine on a piano F3, so see how well it detects/describes the inharmonicity we know to be there. I can provide a sample if need be (just not this second).

I think the main thing to look  at is the relationship f1/f0. How much that deviates from 2.0000 is what we care about, whether that gets scaled to cents or not is less important.

When I tune, I use a software that lets me average the pitch over a fairly long window (A couple hundred milliseconds). Even with that smoothing, the pitch will vary a few cents in the low end as the note rings out. (Thankfully it's much more stable in the midrange). In practical terms, adjustments of a cent or smaller are really difficult in either extreme of the range. My display shows hundredths of a cent.

[Mark II]

at first: sorry, I might be lacking the right vocabulary

I am using a 1 second sample and N=2^16 = 65356 FFT coefficients resulting in a frequency resolution of 0,67 Hertz

delta f = fs/N = 44100 samples per second/ 65356 samples = 0,67 Hertz

the var equal temp column gives the variance of the measured peak of the spectrum to the next note of the equal tempered (e.t.) scale. it bridges the gap from frequency to the name of the note of the measured peak.
for example, one peak is at 43.7393 Hertz, the next note of the e.t. scale is F1 and its frequency is 43.6536 Hertz. the variance of this measured frequency (43.7393) to the e.t. scale (43.6536, is not in this list) is 3.3961 cents.

another example:
the third harmonic is found at a frequency of 131.8909 Hertz, the next note of the e.t. scale is C3, the variance of this measured frequency (column 3) to the e.t. scale is 14.2065 cents.

the last column is the important one. It subtracts the measured variance of the hamonics (column 4) from the natural variance of the harmonics which is not shown in the list but can be found here:



from
http://en.wikipedia.org/wiki/Harmonic_series_(music)
the numbers below the notes show the number of harmonic
the numbers above the notes show the variance in cents from the note of the e.t. scale, for example third harmonic is 2 cents too high than an e.t. , meaning that the third harmonic is 2 cents higher than notated

so, for the second harmonic of F1, the natural variance would be 0 cents.
Meaning, for all real natural harmonics due to the scale in the image above the data in the last column would be all zero. If that is the case (all data zero in the last column the partials are absolutely harmonic, no difference to the scale of natural partials.

hope this makes it easier to follow the list.
feel free to ask.

Mark II

[Mark II]

Putting the theory of this aside for a moment, has anyone ever played a stretch-tuned Rhodes and non-stretch-tuned Rhodes side-by-each to compare how they actually sound?  Mine is stretch-tuned and I'm happy with the sound, in fact I prefer the stretch-tuned sound to the non-stretch-tuned sound but I've never been able to A/B it because I only have one Rhodes.

Randy

same here, one rhodes, one tuning method. sorry

Mark II

[Rob A]

I am using a 1 second sample and N=2^16 = 65356 FFT coefficients resulting in a frequency resolution of 0,67 Hertz

delta f = fs/N = 44100 samples per second/ 65356 samples = 0,67 Hertz

I kind of suspected that the resolution may not be high enough to really warrant the displayed precision. Is it possible to increase N in your code?

From 87.5Hz to 88.17Hz (.67 Hz interval at f1) is 13 cents. So I'm worried about how precise we are in measuring the harmonics.

I'm kind of interested in the peak detection code too.

the var equal temp column gives the variance of the measured peak of the spectrum to the next note of the equal tempered (e.t.) scale. it bridges the gap from frequency to the name of the note of the measured peak.
for example, one peak is at 43.7393 Hertz, the next note of the e.t. scale is F1 and its frequency is 43.6536 Hertz. the variance of this measured frequency (43.7393) to the e.t. scale (43.6536, is not in this list) is 3.3961 cents.

Variance to ET isn't too material here. The real thing to look at is the ratio of first harmonic (f1) to fundamental (f0). If you choose to express it in cents, fine.

Thanks for putting this analysis together, by the way.

Added: it's been over 12 years since I picked up MATLAB, so maybe doing a larger FFT isn't feasible.

[Mark II]

I kind of suspected that the resolution may not be high enough to really warrant the displayed precision. Is it possible to increase N in your code?

sure.

I'm kind of interested in the peak detection code too.

that is too easy to open to other persons  :wink:
This was my first MATLAB experience, and this spectrum analysis was kind of leftover from other things (more to come later).
I can send the m file in a PM as I guess you will be the only one interested in details. do still have a MATLAB version on your computer ?

Variance to ET isn't too material here. The real thing to look at is the ratio of first harmonic (f1) to fundamental (f0). If you choose to express it in cents, fine.

Actually it's a variance to the natural harmonics and there for the same as your suggestion, a variance to f0. in the natural harmonics the second harmonic is a true octave of the fundamental frequency. I thought it might be easier to go this way as I wasnt sure how many harmonics should be analyzed. your way was my first trial. I skipped that one.

kind regards
Mark II

[Rob A]

I don't have a MATLAB anymore, it went away with my Pentium 90 machine.

Actually, I was just bothered by why the last column didn't show a 0 variance versus harmonic for the fundamental. It seems like it would have to.

then later it occurred to me just to consider f1/f0.

then I wondered about your precision as described above.

then later it occurred to me that if you didn't have enough resolution you may not be detecting f1 reliably.

then later it occurred to me that if you weren't using a windowing function that maybe that interferes with detection of the harmonics. After doing this way back when I started thinking harder about the effect of the windowing function on the spectrum.

I think you are on the path to success here. I'd like to see this with N increased substantially (I'm willing to spend your CPU cycles for you!). I'll provide a piano sample soon so we can compare the results.

[Mark II]

mmh

a lot of things to tinker with. :wink:
I may add I used a blackman window on my sample before the FFT.

the last column didnt show a 0 variance as I wasnt able to tune the tine to this amount. move the tuning spring a bit and you will get a 10 cent step. so I thought anything in the range of +- 5 cents is pretty in tune for f0.

have to leave now.
will come back with details

yours

[edit:]btw. it was your "dimension of tone" thread that started all, so I have to thank you.

Mark II

[Rob A]

Two cents is a frequency ratio of 1.0011559128538237. My opinion is that's enough resolution.

So working backwards from that to N:

For two-cent precision at the first harmonic (let's call it 87.5Hz) we need 0.10 Hz resolution. (0.0011559128538237 * 87.5).

You said that delta f = fs/N = 44100 samples per second/ 65356 samples

So to get to our target 0.10Hz we want  N = fs/delta f =  44100/0.1 = 441 000 or better. That's not a huge increase. Of course powers of two are better, so say we take N as 2^19 = 524 288. That would get us in the realm that our tuning tools' precision will be the limiting factor at least.

[Rob A]

See what MATLAB makes of this:

http://stashbox.org/248389/F1-Yamaha-G3.wav

I know it's considerably sharp. I'm way overdue for a tuning.

[Mark II]

mmh, my code is wanting a major update.
f0 detecting lacks, and this is the basis for all other investigastions (Matrix of harmonics is based on f0)
as I said before this code is a kind of leftover from another routine which had to have a simple design, but now with your acoustic piano sample it is getting everything mixed up.

I will have to update my code for higher N and therefore higher frequency resolution in the spectrum plus a better f0 detection (maybe with an second input, give him the name of the note that was played in the sample)

see you later

Mark II

[Mark II]

so, here some more input, based on my old code (N=2^16) an analysis on Rob's sample of F1 played on his yamaha grand:

Code Select Expand


   'n#'    'mag(dB)'     ' freq (Hz) '    'var equal temp'    'note'    'var harm'
   [ 1]    [-47.8352]    [    43.7393]    [        3.3961]    'F1 '     [  3.3961]
   [ 2]    [-13.4078]    [    86.8057]    [       -9.9705]    'F2 '     [ -9.9705]
   [ 3]    [ -5.4037]    [   129.8721]    [      -12.4968]    'C3 '     [-14.4968]
   [ 4]    [ -7.2270]    [   173.6115]    [       -9.9686]    'F3 '     [ -9.9686]
   [ 5]    [-20.4223]    [   216.6779]    [      -26.3422]    'A3 '     [-12.3422]
   [ 6]    [-10.0226]    [   261.0901]    [       -3.5499]    'C4 '     [ -5.5499]
   [ 7]    [-11.8612]    [   304.8294]    [      -35.4018]    'D#4'     [ -4.4018]
   [ 8]    [-26.4444]    [   349.2416]    [        0.0676]    'F4 '     [  0.0676]
   [ 9]    [-20.2585]    [   393.6539]    [        7.3109]    'G4 '     [  3.3109]
   [10]    [-10.2690]    [   438.0661]    [       -7.6259]    'A4 '     [  6.3741]
   [11]    [       0]    [   483.1512]    [      -38.0333]    'B4 '     [ 10.9667]
   [12]    [-12.1171]    [   528.2364]    [       16.4166]    'C5 '     [ 14.4166]
   [13]    [-12.1235]    [   573.9944]    [      -39.7615]    'D5 '     [-80.7615]
   [14]    [-12.1235]    [   573.9944]    [      -39.7615]    'D5 '     [ -8.7615]
   [15]    [-12.4744]    [   619.0796]    [       -8.8544]    'D#5'     [  3.1456]
   [16]    [-13.3650]    [   665.5106]    [       16.3499]    'E5 '     [ 16.3499]


I will change the code in order to increase the resolution of the spectrum, but cant promise till when this will have happened.

I will leave the explanation to you, Rob :wink:
kind regards

Mark II

[Rob A]

I'm quite bewildered why the 11th partial is 47dB above the fundamental.

I ran my various spectral analysis tools on it (none of which have the required resolution, btw) and did not see that effect.

It also shows the second partial as flat with respect to the first (f1/f0 = 1.9846) so I obviously have some doubts about precision. I know my piano isn't squish tuned.

I'm also kind of skeptical that f0 seems to agree with the Rhodes table in the first post to four decimal places. Is that supposed to be the observed value of f0? Or the correct equal tempered/A440 value for that note name?

[Mark II]

I'm quite bewildered why the 11th partial is 47dB above the fundamental.

I ran my various spectral analysis tools on it (none of which have the required resolution, btw) and did not see that effect.

I did, I doublechecked this with Audacity, as I was astonished about this fact aswell. Audacity shows the same effect.

I'm also kind of skeptical that f0 seems to agree with the Rhodes table in the first post to four decimal places. Is that supposed to be the observed value of f0? Or the correct equal tempered/A440 value for that note name?

Might be due to reduced resolution. F1 frequency is 43.6536 for tempered@440Hz.

kind regards
Mark II

[Rob A]

http://www.afn.org/~afn49304/youngnew.htm

this paper makes me think about an alternative approach. Tell me if you think this would be difficult to implement.

I think it's mainly of use to consider f0 and f1 and their ratio (for our purposes here--determining whether stretch tuning is valid conceptually for a Rhodes).

Instead of a full-bandwidth FFT on your signal, can you apply a notch filter around f0 (based on some user input) then do a more high-resolution determination of frequency at that peak? Then do likewise for f1?

Or, as yet another idea, how about we add the test tone to a sine wave of fairly precisely calibrated frequency and count beats? Filtering steeply to isolate the relevant partials should help make that easy to read.

Added: if readers are scratching their head wondering what the hell we are talking about, try this reasonably friendly link:
http://www.kenfoster.com/PianoTuning/SuperTuning.html

[Rob A]

I'm going to play with this for a little while:

http://www.tunelab-world.com/

I read on a piano tech list that it can measure inharmonicity directly.

[Rob A]

http://www.tunelab-world.com/

I read on a piano tech list that it can measure inharmonicity directly.

Oh yeah, this is bloody brilliant. Dead simple, and it conclusively proves me correct.    :shock:

I'll get a bunch of data together and post it up here shortly.

[Rob A]

Okay, first the discussion of methodology:

I used TuneLab pro, which is a purpose-built tool for measuring the inharmonicity of a piano and computing a stretch tuning for it. You can repeat my results if you choose to, this is a program with a fully functioning demo mode. It is designed to facilitate getting a custom stretch curve for your piano by taking representative measurements of inharmonicity across the scale. It then computes a tuning curve, which it allows you to store and manipulate in several ways. Its purpose is to allow a piano tech to quickly tune pianos with a high degree of accuracy.

I calibrated the sound card in my computer using an inexpensive chromatic tuner I own as a pitch reference (I couldn't find my tuning fork). It's worth noting that my soundcard was about 13 cents out of true pitch, which I noticed as I did a quick check on my Rhodes (if every note is out by 13 cents it is an offset problem).

So when you build the tuning for a new piano with this program, you measure inharmonicity of several notes across the scale. By default it goes by octaves, C1-C5. I used at least three readings per note to make sure I was depicting the scale as accurately as I could (again, it has the built in feature to average readings).

When I got to C3 on the Rhodes, it started to complain about the inharmonicity number being too low to be a piano! I had to manually accept each result--the error message suggested I check for clipping or other distortion. A nice sanity check on the results though.

I wanted to make double extra sure I was accurately measuring the Rhodes, so I took additional readings from G2, G3, and G4.  This rather brilliant program will let you read every note if you so choose (the Accu-tuner only uses three notes to compute the whole stretch table)

Now, on to the data:

These are the observed inharmonicity coefficients of the Rhodes home piano and the Yamaha G3 grand:

Rhodes:


Yamaha:


And the stretch tuning curves computed by Tunelab for both instruments:
http://stashbox.org/257495/Yamaha.PNG
http://stashbox.org/257494/Rhodes.PNG

And finally, my analysis:

You will note the complete flatness of the Rhodes curve. It follows directly from the essentially zero inharmonicity constants measured. So, stretch tuning a Rhodes is not desirable for the same reasons as stretch tuning a piano. The only reason to do it would be if you needed to remain in tune with an acoustic piano in the upper octaves. Note the treble end of the tuning curve for my grand piano goes to almost 45 cents sharp at the top. Nearly a semitone. However, the bass is only about 5 cents flat.

Take a look again at the stretch tuning chart from ye olde manual:
http://www.fenderrhodes.com/org/manual/fig5-4.gif

Ironically enough, if you do follow the first procedure described in the manual chapter 5 (the tuning procedure without using a tuner), you'll wind up with a non-stretched tuning on a Rhodes, but you'd wind up with a stretch tuning on a piano.

Finally, check out the Wikipedia article:
http://en.wikipedia.org/wiki/Stretched_tuning

It specifically mentions Rhodes pianos and inharmonicity of tines. We now know that to be incorrect. As shown clearly above, Rhodes tines are essentially free of inharmonicity. The deviations from zero that show in the computed constants are likely to be limitation of the precision of the measurement.

[MuMajor]

This is such great stuff. I will have to try that Tune Lab myself as well at some point.

Mark II's experiment seems like very good evidence for there being no need to stretch tune the Rhodes, as the inharmonicity is virtually nonexistent. However, when I play my Rhodes without stretch tuning applied, to me it still feels like the top was slightly flat. I have been pondering over whether I should stretch tune, but the findings above do not motivate me to do it. It is kind of bewildering, though: there seems to be no inharmonicity involved, but the ears still seem to have some problem (especially) with how the the upper part of the keyboard sounds like.

So, I was wondering if this could be more like a matter of being so familiar with how my piano is stretched (the tuning, that is) that my ears kind of want the Rhodes to have some stretch in the upper notes as well, even if there was no inharmonicity-based need to stretch tune the Rhodes. I don't know, just a thought..

[BJT3]

Rhodes nerds...

[james page]

Guess I'm a little late to this party but I'm very interested in the topic. I tuned Rhodes for years with Dyno-My-Piano using a Conn Strobe which I still have. I recently bought Tunelab to tune my upright piano with and thought I'd try it with the Rhodes. When checking for inharmonicity I got the same results as Rob.
To achieve "stretch" when using the Conn I would allow low notes to roll flat on the display and allow the higher ones to roll sharp. I guess I'll try a similar approach with my first Rhodes tunelab experiment, even though the tine "swing" seems to be much smoother/reduced on the TL display.
Anybody have any updated rhodes/tunelab experience to share?

[pianotuner steveo]

Tune lab DOES NOT work well with a Rhodes.

It is meant for acoustic pianos only. ( Or EPs with strings)

[james page]

Why not?

[pianotuner steveo]

Because it does not have strings is what I was told. (Almost no inharmonicity)

The program was designed for stringed pianos, whether they are acoustic or electric.

It sounds like crap using with a Rhodes.

Tunelab is a $300 program for acoustic piano tuning. Do not waste your money on it for tuning reeds or tines.

[james page]

Because it does not have strings is what I was told. (Almost no inharmonicity)

The program was designed for stringed pianos, whether they are acoustic or electric.

It sounds like crap using with a Rhodes.

Tunelab is a $300 program for acoustic piano tuning. Do not waste your money on it for tuning reeds or tines.

You are right in that it was designed for ac pianos and its ability to measure inharmonicity is a great feature. But just because you don't use that particular feature doesn't mean it can't be helpful in tuning a Rhodes.
I bought it to tune my acoustic piano and I loved using it. But I also just tuned a Rhodes with it and it was definitely an improvement over my Conn strobotuner.
Would I recommend someone buy it to tune a Rhodes? Maybe.
What method do you use?

[bourniplus]

I've used Tunelab for a Rhodes and it worked well. In fact I've also used it to tune my melodica and my accordions. You can select different tuning curves from "default" (no stretch at all) to very stretched curves.

[pianotuner steveo]

I use a Korg OT 120 to tune Rhodes, and it is far easier and sounds better than using Tunelab.

Why are you folks spending $300 on Tunelab for hobby use?

[bourniplus]

No need to buy the full version for "hobby" use. The demo version is fine and it's completely legal, it just pauses for two minutes every 14 notes. However I'm not sure if the demo is available on Mac.

[pianotuner steveo]

I use IPad. There is no demo version for that, but I do use it 5 days a week so the expense was not that great for me.

[pianotuner steveo]

We should probably start a new thread for this question, but how does one tune a melodica or an accordion? ( short version)  is this similar to Wurlitzer tuning?

I thought melodica reeds were plastic.

[bourniplus]

I've never tuned a Wurly, but I believe there are similarities; the reeds of an accordion or melodica are tuned by filing either end of the reed, making it sound higher or lower.

[james page]

I use a Korg OT 120 to tune Rhodes, and it is far easier and sounds better than using Tunelab.

Why are you folks spending $300 on Tunelab for hobby use?

In fact I bought Tunelab to tune my acoustic piano and since I already had it, I tried it on a Rhodes and it worked great there too. It's just a tool and as such it is only as good as the operator using it. I trust that you can do a fine job with your Korg tuner but please don't tell us "it is far easier and sounds better than using Tunelab".

[pianotuner steveo]

Well, it is far easier for experienced tuners. Not sure why you don't believe that. Been tuning Rhodes with korg tuners for 35 years. I tried Tunelab once and will never use it on a Rhodes again. Sounds like crap. You can plug the harp directly into a Korg tuner. It takes me about 20-30 minutes to tune a Rhodes with a Korg tuner.

Tunelab was not designed for Rhodes or Wurlitzer tuning. It does not work well with a Rhodes. You can't plug the harp directly in and tune with Tunelab. You have to amplify the piano which can bring in all kinds of outside factors based on EQ, etc.

Only as good as the operator? What does that mean?

[james page]

Well, it is far easier for experienced tuners. Not sure why you don't believe that. Been tuning Rhodes with korg tuners for 35 years. I tried Tunelab once and will never use it on a Rhodes again. Sounds like crap. You can plug the harp directly into a Korg tuner. It takes me about 20-30 minutes to tune a Rhodes with a Korg tuner.

Tunelab was not designed for Rhodes or Wurlitzer tuning. It does not work well with a Rhodes. You can't plug the harp directly in and tune with Tunelab. You have to amplify the piano which can bring in all kinds of outside factors based on EQ, etc.

Only as good as the operator? What does that mean?

I don't see how anything could be easier than tunelab. The display is large and very easy to read and extremely accurate and precise. And tuenlab can't sound like crap because it doesn't make a sound at all. It's just a tool.
No it wasn't designed for Rhodes or Wurlitzer tuning. Nor was a Fender twin reverb designed for Rhodes or Wurlis but they sound pretty good through one. And no you can't plug directly into tunelab. But aren't you already using a good clean, reasonably flat system to set up and tune your Rhodes? If so, Tunelab is good to go with that. Don't even need to plug it in. Please don't tell me you tune directly into the tuner without even listening to the piano.

By "only as good as the operator" I meant that these are only tools. You could put a novice in front of a piano with a $1000 tuning system and it would not compare to what a pro could do with a $5 tuning fork.

[pianotuner steveo]

I said nothing about Tunelab making any sounds- not sure where you got that from. I also said nothing about the display being hard to read.

I do the first pass of the tuning with the harp up, and plugged into the tuner. After that, I put the harp down and listen. Fine tuning adjustments are then made by ear.

Tunelab takes far longer. You need to sample notes first ( the first time) plus there is far more flipping the harp up and down.

I've been tuning instruments since 1968, professionally since 1979, and for me, my method is FASTER and easier.