Piano World Home Page Forums Our Most Popular Forums Piano Tuner-Technicians Forum Quirky Microphone Inharmonicity Measurements!

I use 2 microphones for software piano tuning, both inexpensive but with very good reputations. 1. A USB microphone (plugs directly into my netbook)... the Samson C01U; 2. a "measurement" microphone (plugs into an XLR to USB converter call "Icicle") .... the APEX 220.

Both work really well with Verituner, although they differ very slightly on "A440" (about .10-.15 of a cent difference).

I recently did an IH measurement of the Hailun 218, from top to bottom, first with the Samson, then with the Apex using Verituner piano tuning software. The IH measurements went very quickly, which is a good sign in Verituner.

Then something weird happened with the Apex: it completely and totally FAILED to to provide a inharmonicity measurement of any kind for A#1!!! I completed all the remaining keys in the bass, easily getting IH measurements for all of them. I then when back to the intransigent A#1, and tried again and again; finally rebooting Verituner, and then rebooting the computer itself.

Nothing. A#1 could not be measured, at all, with the Apex 220. So I switched mics, from the Apex measurement mic to the Samson USB, for that ONE key. The recalcitrant A#1 inharmonicity measurement was quick and easy with the Samson, and of course, that's NOT a measurement mic.

I am stumped. Why should the Apex measurement microphone pick up inharmonicity info for EVERY key save one single key.. the A#1?????

There is a postscript to this odd story.... I tried moving the Apex to various different positions near and far from A#1.

Finally, the oddest thing occurred. I pointed the microphone AWAY from the string and, in fact, held it completely outside the piano itself, facing AWAY from the piano.

Surprise. I got a complete IH measurement, but only under these completely bizarre and counter-intuitive circumstances.

I can only reason that either there is something incredibly weird and quirky about Apex 220 measurement mics, or this particular Apex measurement mic. Or just maybe ALL measurement mics do weird things when measuring inharmonicity.

Very, very odd.

Probably has more to do with the directional sensitivity differences between the two microphones and some mixture of room acoustics.

Did you check the signal level? If the Apex mic is overloading and saturating, that would induce harmonic distortion that would mask the true inharmonicity. I think the Verituner is designed to detect this condition and avoid making a bad measurement. But when you pointed the mic away from the piano, the level would have been lower, and perhaps no longer saturating with distortion.

I keep the Windows "recording level" window active at all times; what's more, I keep the level indicator as close as possible to the Verituner level indicator. They are basically right beside each other, so that I can easily see what each of them is doing.

I'm wondering if the electret capsule itself was vibrating inside the microphone housing at the same or very nearly the same frequency as one of the more prominent partials of the A#1 string (I observed the same behaviour on both strings of A#1) on this particular piano??

The anomalous behaviour has proved to be eminently repeatable, by the way. Last night, I cleared the IH information from A#1 and tried measuring again, with the Apex microphone pointed at the piano, in various different orientations. Again, Verituner wouldn't "accept" the information that particular microphone was sending it.

I'm going to check (yet again) tomorrow. The piano room itself is kept at a very constant 44% RH and 22 degrees Celsius. So I don't anticipate any radical change in the characteristics of the string or the hammer that might affect the IH variable!!

I may make a wav. file recording of each string of the offending A#1, with the Apex and with the Samson microphones. I would post those files here via Dropbox.

Measured once again, this morning.

I'm at a complete loss. Now Verituner's IH information for the intransigent "A#1" is working perfectly with the Apex measuring microphone.

Go figure.

I've had it.

Case closed.

I'm forced to conclude that IH measurement is, in its nature, a very fuzzy and inexact enterprise.

I keep trying to tell you that you are trying to find the exact centre of the Atlantic Ocean. You can't.

For one thing, we have to find an acceptable definition of centre, and then we have to accept that the centre is always moving, due to shifting beach sands and tectonic plate movement.

A vibrating piano string is non-periodic. As well, the pitch centre is always moving. As well, iH is an arbitrarily defined quantity that is useful for predicting an approximate upper partial pitch. All of these factors make it impossible to to 'dot the i's and cross the t's' as you would like to do.

Accept the fact that the bullseye of a tuning pitch is not a point, but a large circle, several cents in diameter. Doing so will lower your stress level and allow you to get to playing, as you so clearly do very well.

Circle vs point: a useful analogy, which I will try my level best to keep front and centre!

Just an FYI to the OP and others toying with these things:

Since the A to D is happening inside the mic with these USB mics, you have to be very careful of the levels. The Apex mic I think only has 105db of internal dynamic range (some other brands go to 120db) last time I looked and a well struck piano note mic'd at close range might exceed that. So you could be clipping the signal before the conversion to digital. I'm not sure if Windows and the level slider/meter is getting level info from the mic before the conversion to digital (not sure how that would even work) or after. If it's just a post conversion dbfs meter and the gating window integration is large the transient attack could well be clipping in the mic....

Good point R.P.

John, you could test this by recording your USB mics directly in audacity at several volume levels and looking closely (zoom to the single sample level) at the waveforms.

johnlewisgrant,

Maybe you've already tried this, but if you have, you haven't mentioned it...

Have you yourself tried listening to the mics' output over a set of good headphones?

You indicate that you're carefully monitoring the level indicators of the mic signal. However, my experience tells me that level indicators are not always a good indication of the quality of signal that is being sent down the microphone's plumbing.

Regards,

One other thought, in speaker testing you typically aim the mic at the source but for room measurement the typical positioning is vertical (perpendicular to source). The best measurement mics have calibration files for 0, 45 and 90 degrees but the concern there is amplitude change at different orientations. With measurement mics I would be inclined to shoot with the mic body parallel to strings on a grand and horizontal on uprights (same orientation basically as sound is coming up and out of the lids on uprights spinets - so perpendicular to sound field). I've not experimented with this thought - I get decent runs in Tunelab and EPT with the horizontal positioning.

While I'm just a hobbyist tuner, I build speakers and have a multi-channel home theater and also a recording studio so I've played with lots of measurements along the way....

Audio people choose microphones for the way they make things sound, which means that microphones change the waveform of a musical note. Once you change the waveform, you change the Fourier transform, which is the approximation of the waveform by the summation of harmonic frequencies. So any measurement of those harmonic frequencies is going to be changed by the microphone. Whether it matters or not depends on the accuracy that you wish to measure it.

Nonsense. The microphone changes the amplitude of the partials, not their frequencies. Non-linearity and distortions can create extra partials, but the frequencies of the existing ones will not be changed.

Paul.

Can you prove that? Show the math!

I am not so sure. I tend to agree with BDB on this one.

The issue here is the concept of frequency measurement. You are assuming that frequencies exist as some finite value outside time. I disagree. Frequency is the perceived sensation of a serial phenomenon. Without time, there is no frequency, and yet, that is exactly what we try to do when we measure frequency - to remove time from the sensation. The technique, as mentioned by BDB, is to use a Discrete Fourier transform, which is a finite, summation approximation, and is subject to influence by noise, distortion, and, above all, the time length of the sample. The particular algorithm, chosen by the person doing the transform, uses techniques to choose a particular bin or some mathematical averaging, if you will, of many bins from the transform, to come up with a particular number, which we call 'the frequency'. If we change the algorithm for example, from a Hann window to a Blackman window, we will get a different 'frequency'.

IMHO, as a result of several years of study and experimentation on frequency measurement of non-periodic waveforms, I feel that the noise and distortions added by the microphone and the electronic chain of devices required to deliver the data to the DFT software, along with the noise of the recording environment, room reflections, and the like, do affect the value of the number we like to call 'the frequency'.

BDB and pyropaul are both right - trust me, you would not want to record vocals via a measurement mic (sounds positively bad - dry, sterile and really bad), nor would want to measure precision (at least in the amplitude sense) with a SM57. An SM57 should not change the frequencies, but the partial levels will change and that does alter the color of the sound and gives an SM57 the SM57 sound. Just like an EQ - it doesn't change the frequencies, but does change the amplitude and hence the sound.

In our tuning world, the amplitude should not matter that much as long as the partials are detected to feed the tuning software. Maybe Prout has a better understanding of the code and where the noise floor is to ignore partials (if signal below (x)db - ignore type of thing)....

My point is that, while using different mics or mic placement will not change the frequencies perceived by our ears (actually, it might, as our perception of pitch is affected by partial amplitudes), it will change the measured frequencies due to the effects I describe above.

I couldn't find a single reference to any kind of distortion that would produce a frequency shift. Phase and group delay, yes, but phase differences are inaudible. The only thing that could cause a frequency change would be the Doppler shift (so if your source or microphone is moving).

As for amplitude accuracy, it obviously matters for something like the Entropy Piano Tuner, but probably not for the others where just the partials' frequencies are needed, not their amplitudes.

Of course, for recording purposes, the euphonic distortions produced by microphones and their frequency responses (which lead to terms like "warm sounding") do make a difference - but the frequency response is with respect to amplitude, not the frequency of the partials themselves.

See also Types of Audio Distortion

Paul.

Neither BDB nor I posit that the microphone can change the 'frequency' of the sensed signal. What it does do is change the structure of the waveform over time, and that changes the measured 'frequency'.

I challenge you to record a piano note simultaneously, using two different mics, then measure the frequency of the waveform from each mic. This is easy to do. Use a stereo mic recorded into audacity, then measure each wave form. You will find two different frequencies. They may be close, within 0.1Hz or so, but they will not be the same.

I guess since piano tones are aperiodic, then there is no "frequency". Of course different microphones will change the amplitude of the partials, but the partials' frequencies will remain the same, though I suppose the "frequency" of the aperiodic waveform could be different as you always have to select a finite time window to come up with a figure for this. So, on second reading, I probably agree with BDB in that sense, but not in any sense that the microphones can change the frequencies of the partials (without doppler shift at least - maybe if there's a massive low frequency component this could shift the apparent frequencies of the higher partials as the surface of the microphone's transducer moves).

For things like the EPT, it's the amplitude of the partials that are important, not the composite "frequency" of the overall sampled aperiodic tone. I don't think the bins used by the EPT are down to the 0.1Hz resolution anyway (not that anyone can tune to that accuracy consistently).

Paul.