Discussion has raged in this and other forums over the years about whether tuners are helpful in assessing the tuning accuracy of your flute and your playing, matters of critical importance to the flute player. Tuners have not faired well in the discussions, sometimes unfairly. To me the biggest problem with using a tuner is “staying honest”. If you watch the tuner while you play notes at it, you are drawn to correct them, possibly not a bad thing in terms of training you, but not a good thing in terms of achieving a meaningful and impartial assessment. But I suspect all that is about to change, due to a new development by Graeme Roxburgh.
Graeme, a flute player and computer scientist in New Zealand, mentioned on the Chiff & Fipple list a computer application called Tartini, developed by researchers at his university. Tartini permits you to record sounds and then view their characteristics, including pitch. I thanked Graeme for bringing it to our attention, and wondered if it might be possible to extend it further. In particular, I was looking for a system that would automatically scan through the results, rejecting the bum notes inevitably created accidentally, and collating the good ones, producing a graph of the intonation of the flute (and of course its player). It would be truly lovely if you could just play a tune into it, but would be good enough if it could just handle scales. What would be important is that you can play without being aware of the results, and that the machine, and not the user, would be responsible for analysis. Honest, impartial, easy.
Well, by George, I think he’s done it. Rather than delving into Tartini itself, Graeme employed a stats package called R to do the post processing. You record into Tartini, export the data into R and, bingo, up comes a filtered, collated intonation graph that Graeme has programmed. The combined system is tentatively titled the Tartini-R Polygraph.
But, hang on a tick, how do we know it’s accurate? I tried to shake it up…
I suggested we try it on an instrument that cannot alter its intonation. Graeme produced credible looking plots from Jackie Daly’s accordion playing. I still wasn’t convinced, so used Excel’s random feature RAND() to make up a “tune” using 5 each of all the notes in 2 octaves of the D and G scale. I converted that into music using ABC Navigator to play the “tune” and create a MIDI file using the flute voice. (Sounds a bit like Schoenberg but not as bad.) Interestingly this showed up some tuning problems on the Polygraph. Graeme investigated and found that the MIDI system was introducing tuning blips at the starts of notes. He used Audacity to notch out the blips, and Tartini-R now reports the tune as being in tune. So Tartini-R was right - it was the MIDI tune that was not!
I’d have to say I’m impressed and pretty excited about this (after all, this is something I’ve been hanging out for for over 30 years!). It’s going to show up problems that we flute makers might prefer not to have shown up. I can imagine we’re going to be receiving a lot of critical emails enclosing Tartini-R plots. Some of us will deal better with that than others! Perhaps we should accept bribes to keep the genie in the bottle? Whoops, dropped the darned thing …
It’s also going to show up problems that we flute players might prefer not to be reminded of. But better to know you play the second octave flat in the privacy of your own home, rather than having it come up for open debate at the session. And of course, it isn’t limited to flute. Fiddle players would find it equally valuable, although there it’s all about the playing and not the instrument.
So, it’s clear that the system has great promise, and is worth developing further. But have we done enough to prove it? We don’t want to release it, get a lot of people worried about their flutes and flute playing and then find out it has a problem we hadn’t thought to check. What further tests should we subject it to before releasing it?
Speak now! I can’t keep my thumb over this bottle forever …
Terry
