This Old Flute - nach Meyer

[Carey]

I took a little time tonight to install Tartini, and R and it's working fine. I wanted to see how best to record what I was doing, so I hooked up the web cam and played around with the audio settings on the capture software. I turned the record volume lower than normal for voice so the flute's strong notes wouldn't be clipped. Here's the first video I shot this evening. I post it not for it's technical content, but because it shows my surprise at seeing myself. Ha!

Now I've only had a flute since mid-May, so don't expect much in the way of skill at playing yet. That and our session went for 7 hours yesterday, and I think my lips are shot.

http://www.youtube.com/watch?v=K6NEeGTsR9Y

It did work to play the video with Windows Media Player and record it into Tartini or Audacity. From Audacity I could export a WAV file and analyze that file with Flutini. I thought I could do the same with Tartini, but I couldn't sus it out, so I just played it and let Tartini listen to that and do it's thing.

The plots below were made from playing directly into Flutini and Tartini and following the normal display procedure.

I didn't use the same file, but I did play pretty long single notes from D thru D and on up to A and back a number of times. With Flutini I could watch the bars move and counts increase so I knew it was hearing the right notes. With Tartini I could watch the tuner and see that it was hearing me. I exported the data from Flutini and displayed it with the same R script.

Both gave pretty similar results, which is good, because I'm really testing the tool at this point more so than testing the flute. But, a puzzle - why are so few notes displayed when I watched all of them register on the tuner dial? Do I need to configure something so it will display notes it's not expecting? On Tartini I set it for the key of C# (A=440) as that most closely matches what I see.

This is Flutini:
<a href="http://picasaweb.google.com/parkscarey/ ... 5426"><img src="http://lh3.ggpht.com/parkscarey/SJfMuEt ... lutini.jpg" /></a>

And this Tartini:
<a href="http://picasaweb.google.com/parkscarey/ ... 1890"><img src="http://lh6.ggpht.com/parkscarey/SJfMtzJ ... artini.jpg" /></a>

[Carey]

Perhaps it would work better to tell Tartini the flute is a D and lower the A=440 value. I'll try that tomorrow.

[groxburgh]

Perhaps it would work better to tell Tartini the flute is a D and lower the A=440 value. I'll try that tomorrow.

Hi Carey
Don't worry about telling Tartini what key/pitch you're in because that only affects the graphed output that Tartini is showing you. But do tell Polygraph what pitch you're at eg 430 if you've got the slide all the way in, and you'll get a reasonable plot. You'll still have some problems because of F# being so low in pitch that some of it is getting assigned to Fnat etc...
And you can email me Tartini's output file (rtest.txt or whatever you call it) and I'll have a play with the different settings to show how to get a meaningful plot when the flute is basically not near A440 and not very in tune at any pitch.

Cheers
Graeme

[Terry McGee]

I was hoping you'd step in here, Graeme. Seems to me that the flute tuning or something is giving potentially misleading results, not a good thing when you have the chainsaw running.

If, as Carey seems to think, the flute is distinctly flat all over (compared to modern pitch), a lot of the notes are going to be interpreted wrongly - low F#s as sharp Fnats, for example. You can minimise that if you are aware of it by adjusting the ref frequency, but if you had a flute that had really bad tuning as well as really flat (or sharp) tuning, the whole picture could get rather confusing.

I wonder if we could devise some way to set the fences between notes more meaningfully than the nominal +/-50 cents? It would be interesting to plot the incidence of all the recorded notes from say Carey's flute against a semitone scale and see where the minima actually fall, and place the fences there. Could that be done, either manually or automatically?

In some cases the fences could be further from the intended notes, eg if you are not intending to play Eb's or Fnats (i.e. you've opted for diatonic in D), the fences need not be on the quarter tones, but could be at Eb and Fnat. The fence between F# and G would have to remain around the quarter tone.

Terry

[Carey]

Terry, I wonder if RTTA is the right tool to be using at this point. Or rather I wonder if using it in this manner is appropriate. The flute is not in tune well enough to need the sophisticated analysis of live playing that RTTA provides. The needle on the tuner is good enough. But the needle on the tuner does not persist.

As you say the issue is confusing the different notes. I will make 14 seperate files and look at the notes one-by-one, then there will be no confusion about which is which. I will play notes into Audacity where I can select out the better (more representative) ones and miss out the ones where I don't get the tone started well etc. and then feed them to Tartini, like with like, and see what we see. I suspect this will lead to a better procedure when approaching instruments that are way off.

Some things are better left to the human than to try to make the computer do it, especially when it's not mundane (yet.) (Refer to the fixing the tool conversation here.)

Aside: I spent 10 years creating software for professional race teams and I knew I'd spotted something that I could sell when I said to myself - "Now THAT's a stupid job for a human!"

[Carey]

OK, going back to a known quantity to test the testing tools, here's the second Tipple flute I made. the process seems to work OK on a decent flute. (This one thicker walled than Doug's specs so I had to figure out the holes on my own, so don't blame Doug for the tuning here.)

<a href="http://picasaweb.google.com/parkscarey/ ... 0658"><img src="http://lh5.ggpht.com/parkscarey/SJhRq4h ... 0flute.jpg" /></a>

Wonder if I forgot to play 2nd E? I know I didn't bother with Cnat.

[david_h]

Why not just scan along the Tartini 'Pitch Contour' and move the reference frequency around until the almost horizontal parts in the middle of the notes (of some, or maybe only one, pitch) line up with the grid.

I may be getting the 'wrong end of the stick' with all this RTTA stuff but in all the clips of the experts I have looked at (that we are not allowed to talk about), and in my own efforts, most of the tails of the distributions in the summary statistics are in the transitions between notes and around cuts and things. About the only thing my playing and the experts have in common is that somewhere in the contour of all but the shortest notes is a horizontal bit (or wavy if its fiddlers with vibrato) that is returned to repeatedly and varies by only a very small amount during the tune (although in my case it sometimes slowly drifts sharp or flat before I get to the end).

Sorry to sound like one of those cynics over on the thesession.org but it seems to me that the devil is in the detail. I find Tartini a superb tool but I get questions rather than answers from summary output. Thinking about a post-processor for the Tartini output to select the 'horizontal bits' but time probably better spent practising.

Ha ha - crossed with Carey's last two, must remember to check before submitting

[Carey]

David, I think you hit the nail on the head. I had just recorded a scale of full breath notes in Audacity that I am going to feed to Tartini by clipping out the stready tones from the middle of the notes. Same thing, just done on the front end where I know what it is I'm up to so the software doesnt' have to guess about it.

Thanks for your feedback! These sorts of comments are what will make this effort pay off for all of us.

[david_h]

Another thing that is revealing if you record directly into Tartini and watch the trace is that sometimes it gets the octave wrong (I think there are settings to reduce this).

But amazingly when playing a long note and altering the character of the sound it will sometimes change its mind about the octave and go back and re-write the contour of the part of the note that it has already displayed - spooky when that happens.

[ImNotIrish]

I have a Nach-Meyer LP (low pitch) which had the usual tuning issues, etc. It has an ebonite hj as well. A few years ago, on the advice of someone here on the board (and not knowing any better), I started to file the holes and the embrochure hole, trying to raise the pitch of the instrument and better the tuning. The flute plays, with the slide all the way in, almost at A=440. I have only based this on my plaing along with a concertina player. I can blow the flute in to tune, but it requires rolling out, and really blowing! I don't really have a clue how to go about using Tartini, but I can make a sound sample and post it if someone else would like to figure the tuning out. The flute has been sitting in my bathroom, put together for about two years. I pick it up and play it occasionally, and can get a decent tone out of it after a while. Ill try and get a sound clip posted in the next couple of days.
Arbo

[Carey]

Great! I look forward to hearing it and inflicting it on RTTA. I expect Graeme even more so. I'd love to know the measurements along your flute now to compare it to mine and see if it is in range. I wonder if the holes were in the same place to begin with? In other words, are we solving the same problem?

[ImNotIrish]

Where should I take the measurements from? You have to realize that I initially just started experimenting but enlarging the hole and embrochure with no clue about anything except my ear and an old generation 'D' whistle as a guide. Get back to me about the measurements.
Arbo

[monkey587]

Now I've only had a flute since mid-May, so don't expect much in the way of skill at playing yet. That and our session went for 7 hours yesterday, and I think my lips are shot.

http://www.youtube.com/watch?v=K6NEeGTsR9Y

If you've only been playing that long, and that video represents what you sound like, then I recommend not retuning the flute at this point. Otherwise, once your embouchure develops you'll likely have a flute that's way sharp and has strange intonation. Take it from someone who has modified the tuning of flutes and then later regretted it...

[Carey]

I wondered about that. I'm not very accomplished, but that video is exceptionally poor playing. I posted it because I was amused by my bearded countenance when I saw it on the monitor, and thought I'd share it. I'd have to say I'm getting there, embouchure wise. I can play into the third octave without overblowing.

I'm going to send my nach Tipple off to a better player and we'll see how that does in their hands. Should give a point of reference at least.

Thanks!

[Carey]

Now I have the Tartini problem again. It starts up, but as soon as I click the record button it locks up, using 98% cpu. I have to kill it. Rebooting does not improve the situation. Anyone have any ideas how to get it going again?

Carey