Arguments on properly recording/sampling acoustic pianos (+more)

dikrek

You used AI agent to write that long response. Deny it and we will go on a giant off‑topic (it can be verified with antiAI although you might have touched the formatting here and there → reminds me of our private conversation). Cheeky.

I do not mind. AI is still learning.

dikrek The bit I disagree with most is “phase align” the six microphones.

Effing AI at it right there. So…

On piano, being a big musical instrument, you can place many, many mics. Ideally, I would have placed Nu‑multi pickups under each note's strign(s) and piezos on bridges. But we are restricting ourselves to microphones (typical acoustic recordings).
Ok, use 5 mics (close‑up) if you will. Or another one under the piano → that confined acoustic space also contributes to the overall sound.

Use 3 close ups (at least). Low, mid, highs, omni directional. One mic → no problems, two mics (stereo or binaural) → you can\will get phase issues unless you know what you are doing (pointing parallel, 90° or worst is 180° to one another). With omni mics it is expected and inevitable. More direct pattern mics, more individual ranges but heavier setup. Hence → better is with individual pickups.

It is cringe that I have to write this self evident rant but we all learn at some\any point.

The lid. Ok, keep it closed. The only way the sound comes out of that "box" will be from underneath (and rumble resonance of cabinet and lid) → the opposite sound of the soundboard where the strings are blocked by it. This setup has nothing to do with any other string instrument. None. So, good luck.
Remove the lid. Expose the string and the soundboard. Dampen as much as possible reflections from metal frame. It is not and should not be a part of the acoustics → it is a structural reinforcement. But being metal and hard it reflects and shifts the real sound of string, soundboard and cabinet.

Phase aligning is only valid between 0 to 200Hz above which it becomes an impossible task to align denser freq. waves. That is why in audio mixes all instruments but bass (incl. synth bass) and kick drums are cut off either with deep shelf (as I prefer) or more generally with a 12 to 24dB even 36dB or 48dB (extreme) per renova ("octave") slope low cut. Mics capture it but it is not needed.

Hence on those piano mics I place, only the low and low mid will have info there (0 to 200Hz) → the other (2 or n) mics will get a shelf cut there. Phase alignment will be much facilitated that way.

The stereo (room\studio) pair does not need time alignment → it is distant for the reason to capture late reflection off studio room acoustics. But it will have low information below 100Hz at least which is not important. The bass strings from ≈30Hz to 120Hz have their fundamentals lower in dB than actual harmonics of new (well cleaned) strings (usually wound coiled too). So you do not need that rumble especially on distant mics which will only mess up the phase. Phase between each mic in the stereo pair should be checked and mics adjusted if need be.

dikrek For a natural solo grand recording I’d start almost exactly the other way around:

Find the position where a very good stereo pair already captures the piano as a coherent instrument in the room.

That is because you do not want to record the piano but the acoustic environment in which it is placed\used\emitting sound.
I am talking about recording the piano itself. And in fact would prefer to replace the room stereo pair with IR (impulse response) sets.
For the latter though the actual room should be close to anechoic (if too expansive — it is — to do all this in a proper anechoic one).

dikrek The fundamental difference is that I’d try to record the piano, whereas your (100% unproven) approach seems to record a collection of piano components and then try to reconstruct the piano afterwards.

F‑off AI.

PASHKULI I think this clarifies our actual disagreement.

If your objective is to record a very dry piano in something approaching an anechoic environment and add the room afterwards, that’s completely legitimate.

But I don’t think that makes the result the uniquely “pure piano sound”. Anechoic capture removes the room. It does not remove the spatial complexity of the piano.

There are three reasons.

1. Even an anechoic grand does not produce one canonical signal

A grand piano is a large, directional acoustic radiator. What a microphone captures depends on where it is.

This has actually been measured in an anechoic chamber. Nakashima, Giron and Tohyama measured piano radiation on a 2 m-radius spherical surface, in 10° increments from above to below the instrument and around the full 360° horizontal plane, specifically to derive direction-dependent piano responses. (AIP Publishing)

So anechoic piano sound at position A is not identical to anechoic piano sound at position B.

Both are “the piano”.

That is why I think dry is a much better word than pure.

2. If you want to add the room later accurately, piano directivity matters

This is the really important consequence.

A conventional IR assumes some source position and radiation behaviour. But the actual piano sends different spectral energy in different directions, and that behaviour can change from note to note.

Researchers have tackled almost exactly the system you’re proposing: record instruments anechoically, then reconstruct their interaction with a virtual room.

Their solution was not simply “close-mic the instrument and convolve it with an IR”.

They recorded instruments — including grand piano — with 5 or 13 microphones distributed around them, then represented the instrument in the room simulation as a compound directional source, one component for each anechoic channel. (Welcome to DTU Research Database)

That’s the crucial distinction.

If you collapse the grand into a few close bass/mid/treble signals and then apply ordinary stereo IRs, you can certainly create a convincing and perhaps beautiful result.

But you haven’t recreated all of the piano–room interaction that you deliberately discarded.

3. Removing/damping parts of the piano is then a creative choice, not purification

This is where I think the “pure piano” argument becomes internally inconsistent.

If the objective is genuinely:

piano minus room

then put the actual piano into the anechoic environment and measure what it radiates.

But if before doing that you:

• remove the lid;
• damp reflections from the frame;
• carpet underneath it; and
• decide which structural contributions constitute “real piano”,

you’re no longer merely subtracting the room.

You’re changing the source.

That may produce a sound you prefer. Nothing wrong with that.

But it is a designed dry source, not some objectively purer version of the instrument.

And that’s ultimately my point.

Your approach could be very interesting for a sample library because you’re optimising separation, control and reconstructability.

But if you genuinely want the most rigorous version of:

anechoic piano → add arbitrary room later

the acoustic research actually points towards more spatial capture of the dry instrument, not fewer microphones divided into keyboard zones.

Ironically, your underlying idea is more defensible than the recording method you’ve proposed for implementing it.

And that makes it an interesting experiment — just not “irrefutable logic”.

PASHKULI The phase argument can be reduced to one point:

Multi-mic phase interaction around a piano is real, but the rules you gave for it are not.

There are three separate issues.

1. There is no 200 Hz boundary

You wrote that phase alignment is only valid from 0–200 Hz and becomes impossible above that.

For a pure delay:

phase shift = 360° × frequency × delay

So yes, higher frequencies are increasingly sensitive to small timing errors because their wavelengths are shorter.

But there is no special transition at 200 Hz.

If one signal is simply another delayed by 1 ms, compensating that 1 ms corrects the time offset broadband.

The real difficulty with several piano microphones is that they are not identical delayed copies. Each hears a different mixture of strings, soundboard radiation and reflections.

Therefore no single delay can perfectly align every note and frequency.

That’s true both below and above 200 Hz.

2. Turning an omni 180° does not create a 180° phase inversion

Real omnis become somewhat directional at high frequencies, so orientation can affect their HF response.

But physically rotating an omni microphone through 180° does not invert its polarity.

Those are different phenomena.

With spaced omnis, the important timing relationship comes principally from their positions relative to the source.

In fact DPA deliberately uses arrival-time differences between spaced omnis to create stereo localisation, and recommends spaced omni techniques for grand piano.

So “parallel good, 90° worse, 180° worst” is not a general phase rule for omni microphones.

3. Phase differences in stereo are not automatically errors

A/B stereo deliberately uses inter-channel arrival-time differences.

Those differences help create the stereo image.

That does not mean phase interaction is irrelevant. Combine sufficiently correlated signals with different delays and you can certainly get comb filtering, particularly in mono.

So checking mono compatibility and destructive interaction is good engineering.

But:

detecting a phase/time difference does not imply that the correct solution is to remove it.

Sometimes that difference is part of the recording technique.

Likewise, high-passing some close microphones below 200 Hz may reduce overlapping LF energy and therefore reduce interference.

That’s perfectly reasonable.

But that’s filtering, not phase alignment.

So I think five things need to remain distinct:

Time alignment — compensating an arrival delay.
Phase relationship — frequency-dependent phase difference.
Polarity inversion — flipping waveform polarity.
Filtering — reducing spectral overlap.
Comb filtering — interference when correlated delayed signals are summed.

They interact, but they are not interchangeable.

So yes: six microphones around a grand require careful phase management.

But the physics does not reduce to:

“phase alignment works below 200 Hz, becomes impossible above it, and two omnis pointing 180° apart are the worst case.”

Those specific claims don’t hold up.

PASHKULI Separate point on covering/damping the iron frame, because I think the premise here is backwards.

The fact that the frame’s primary function is structural does not make it acoustically irrelevant. And if the goal is “piano minus room”, altering the frame is altering the piano, not removing the room.

Three points.

1. The frame demonstrably participates in the sound-producing system

You wrote:

“It is not and should not be a part of the acoustics → it is a structural reinforcement.”

But structural and acoustic are not mutually exclusive.

Ingolf Bork of Germany’s PTB actually measured this in a grand piano.

He identified a slowly decaying component around 90 Hz as a resonance of the iron frame. Importantly, he found that although the frame was not radiating that component directly, its vibration was transferred to the soundboard and then radiated acoustically.

So the measured chain was:

iron-frame resonance → soundboard vibration → radiated sound.

Even more interestingly, Bork reports that removing this low-frequency component from the sound at certain listening positions produced an unnaturally synthetic character.

That is experimental evidence, not a theory about what the frame “should” do.

Source: AES record — Bork, Sound Radiation from a Grand Piano
Open copy of the paper

2. Calling frame reflections “not the real sound” doesn’t solve the problem

You then argue that because the frame is hard metal, it reflects sound from the strings/soundboard and therefore “shifts the real sound”.

But why are those reflections automatically less real?

They occur inside the actual piano.

The strings, bridge, soundboard, frame, rim and lid form a coupled physical system. A microphone above the strings records the acoustic field that this complete instrument produces at that position.

If your objective is an anechoic recording, the anechoic chamber removes room reflections.

It does not somehow remove the internal acoustic and mechanical interactions of the piano itself.

So if you then cover the frame specifically to suppress those interactions, you are doing something different:

you are modifying the source.

That may sound better. It may make close-miking easier. It may be useful for a sample library.

All perfectly legitimate.

But it isn’t automatically a more “pure” recording of the original piano.

3. Recent research actually strengthens this distinction

There is now a 2026 study from researchers at the University of Southampton, Politecnico di Milano and Fazioli specifically titled “Cast-iron frame contributions to piano string vibration and sound radiation.”

Their coupled string–soundboard–frame model finds substantial force being transmitted into the iron frame and predicts that the frame can make a significant contribution to sound radiation, particularly in the treble.

Importantly, the authors themselves are careful: the final acoustic-radiation prediction still requires experimental validation.

So I wouldn’t present that modelling result as established measurement.

But it independently reinforces the thing Bork did measure experimentally decades earlier:

the iron frame is dynamically coupled into the piano system. It is not an acoustically nonexistent scaffolding that can simply be removed from the definition of “piano sound”.

Official open source: DAGA 2026 — Cast-iron frame contributions to piano string vibration and sound radiation

So I’d make the distinction very simple:

If you say “I prefer the sound after covering the frame”, fine.

That’s an experimentally testable production choice.

But:

“the frame is structural, therefore it should not be part of the acoustics”

doesn’t follow.

And if the stated objective is specifically to capture:

piano minus room

then the clean experiment is to remove the room while leaving the piano unchanged.

Once we remove the lid, cover the frame and otherwise alter the instrument’s own mechanical/acoustic behaviour, we’re no longer merely isolating the piano.

We’re designing a different dry source.

Which is what all our discussions eventually seem to be about. You need to build your own instrument.

@CyberGene dikrek (if real person) used AI bot, and it is plainly obvious.
Forums are getting bloated by the day. I do not like it. I am not here to teach AIs (they will "read" it anyways).
I am here to share ideas, not "established practices" as alleged facts or rules.

dikrek But there is no special transition at 200 Hz.

Oh, now we have to go on a tangent about low freq energy and how it eats up digital bits (headroom) of digital recordings when summed from signals (tracks) of various microphones. No. No time for that.

@CyberGene
Is there a forum policy about\against using AI in responses? As with your claude_user bot you could see how verbose it is and boring.

dikrek “phase alignment works below 200 Hz, becomes impossible above it, and two omnis pointing 180° apart are the worst case.”

Those specific claims don’t hold up.

This is total garbage. Omnis at 180°. Out of the blue "quotation" which I never wrote.
Phase alignment works mostly for freq up to 200Hz (and that is the upper limit, more likely up to 100Hz) → you cut them out (shelf them down) for sources (mics, pickups) that are not bass\low intended. And, yes above that "freq area" it is impossible to phase align two signals. And it is not necessary.

dikrek But physically rotating an omni microphone through 180° does not invert its polarity.

Ok, but also you place it in an acoustic environment not in vacuum of space or up in the sky. Every position change (not rotation!) means different early\late time reflections and for freq of below 200Hz that will result in phase shift.

F‑off AI.

dikrek But why are those reflections automatically less real?

They occur inside the actual piano.

Because that frame is only there because idiots in the past put more and more strings per key (note) to make piano louder in mid and upper range. That frame is not supposed to be a part of any sound reflections. But inevitably it is. It is of metal to support the increased tension pull. To make it less contributing (undesired contribution) to the sound it should be covered with fluff.

Imagine placing a metal frame over the top and strings of a violin just because it "has to be part of the construction to match that of a piano". Or on a guitar, or wind pipes. It is hard metal. Imagine entering in a wooden room and then in the same room but this time with metal beams and bolts and frames on the floor (ceiling). The acoustic sound will be a nightmare.

PASHKULI I vehemently disagree with a lot of what you write, especially your axiomatic way of saying things are true when in fact a lot is just your opinion.

Was the metal frame a necessity? Sure. Does it affect the sound? Sure. Is that part of the piano sound? Yes. Would removing its effect change the sound? Yes.

You don’t like the piano sound, as we’ve discussed, go build a new instrument.

I fed specifically my points of disagreement to an AI and asked it to produce a Binto Pyramid consultative response based on what I know to be true.

Fair correction on one thing: my final sentence put a paraphrase in quotation marks about the 200Hz deal. It wasn’t intended as a verbatim quotation, and I should not have formatted it that way.

But the paraphrase didn’t come from nowhere. Your exact words were that, with omnis, there are phase issues depending on “pointing parallel, 90° or worst is 180° to one another”.

You now say “position change (not rotation!)”. That’s a different claim — and one I agree with. Moving a microphone changes direct arrival times and the reflection field.

But that happens across the spectrum. There still isn’t a physical boundary at 100 or 200 Hz above which two signals suddenly become impossible to time-align.

A fixed delay produces a phase shift proportional to frequency. Higher frequencies are actually more phase-sensitive to a given timing error because their wavelengths are shorter.

What is true is that two microphones around a piano in a room aren’t simply delayed copies of one another. Different positions capture different mixtures of piano radiation and reflections. Therefore a single delay cannot make their complete waveforms identical.

That’s exactly the distinction I was making.

Filtering unnecessary LF is a perfectly valid way to reduce spectral overlap and recover headroom. But that’s a mixing/filtering decision, not evidence of a 100/200 Hz phase-alignment boundary.

DPA’s own material is quite explicit (and those guys know a thing or two about recording): stereo techniques intentionally use time differences between microphones, and DPA also describes time-aligning support microphones to main arrays to reduce comb filtering — with the appropriate delay depending on the room and reflections, not on a 200 Hz cutoff.

More importantly, though, you haven’t addressed the points I actually find more interesting: the directional radiation of an anechoically recorded grand, or why modifying the lid/frame should count as obtaining a “purer” piano rather than modifying the source.

dikrek Your exact words were that, with omnis, there are phase issues depending on “pointing parallel, 90° or worst is 180° to one another”.

No, my exact words were:

"One mic → no problems, two mics (stereo or binaural) → you can\will get phase issues unless you know what you are doing (pointing parallel, 90° or worst is 180° to one another)." Then I began to speak about omni mics as preferred (by me at least).

In drumming, some mics might get almost 180° degree depending on attachment points and space available. That is why various recording (separately) techniques are used including sampling. Not to mention miking guitar cabinets where guitarist obsess over even an ½ cm distance or a couple of ∡ ° off. The tone changes drastically. Now get two, three mics and it is nightmare. Rarely omni mics are used unless for IRs (for reverb not for the cabinets) but that is different

dikrek There still isn’t a physical boundary at 100 or 200 Hz above which two signals suddenly become impossible to time-align.

It is not a boundary, it is a mixing technique. That is why I prefer shelf down attenuation than low cut → it is a gradual gradient (tautology?) and the phase shift is minimal compared to low cut method. This 200Hz is from a mixing stand point so the engineer does not have to worry about phase issues (but should double and triple check no mater which method is used and either pre‑ or post‑ recording).

dikrek Higher frequencies are actually more phase-sensitive to a given timing error because their wavelengths are shorter.

Ok, so what. Their energy is less in 10s of dB range than those 0 to 200Hz which for recording are much more crucial to get phase aligned and also summed properly when phase aligned for proper final level

dikrek But that’s a mixing/filtering decision, not evidence of a 100/200 Hz phase-alignment boundary.

Aha, so you do understand. Finally. Our ears are naturally more sensitive from to 2kHz to 4.5kHz not to lows. But microphones are not like that unless deliberately calibrated as such. That is why all recordings need proper miking techniques pre‑ and post‑ as mentioned. We get annoyed by low rumble in recordings as it alters our acoustic clear perception of sound\instrument a couple of meters in front of us and directed at us (our ears)

dikrek you haven’t addressed the points I actually find more interesting: the directional radiation of an anechoically recorded grand, or why modifying the lid/frame should count as obtaining a “purer” piano rather than modifying the source.

I have addressed it but in another thread (so in short):
If you imagine the piano (acoustic) as a construction of a cabinet (body of rim and lid, then a frame, no… not the metal frame for the strings, rather the other one underneath that holds the metal frame (see photo\video) through holes in the soundboard, strings, soundboard and that cast iron frame.

So, can you see how with the lid closed it is almost as any other acoustic string instrument but the strings and their board is inside, not outside (to get bowed or plucked) by the hand. And the lid closes the described construction (strigs + soundboard and mrames) which is already inside. It sounds but the actual emitting sound elements — the strings — are now hidden. Hence we open the lid (at an angle towards the audience, mind you).

So, there we go. Audience is listening with their ears in an acoustic space (sometimes on open stages outside too). For recording though only some "audience perspective" mics (typically Binautral head) can be used to record it like that, again in a space. For proper recording of the piano there is absolutely no need for this lid → no audience needed. And binaural perception can be achieved with a set of BRIRs (Binaural impulse responses).

Recording in an anechoic chamber is only needed to get as pure and direct recordings intended for sample libraries, then with combination of BRIRs. You can use whatever mics you'd like in an anechoic chamber. Actually it might be a bit pointless to use omni mic in such acoustic conditions because there it will act as 'directional' mic nevertheless. But it is up to the engineer.

Let me put it this way:

Imagine (I know it is hard thing to do), just imagine we could have a piano instrument where notes could have 3 and even 4 strings per note. And the cabinet, soundboard, rims support frame is all wood and can hold any pressure\tension.

Do you think, all of a sudden people will say: erm, somethings missing… the sound is too wooden, why don't we place a gianormous iron cast frame\grid on top of the strings and see what happens.

The only analog of this I can think of from the guitar world is the so called steel truss rod which they place inside (notice: inside) the neck to… reinforce it (notice: reinforce it) to counteract the tension (notice: the tension) from the strings (incl. metal strings).
This rod though is covered by wood from all sides (except from the tension reg. end).

Another rather unsuccessfully applied analogy of a metal "amendment" to a guitar is the thumb\palm rest on some bass guitars:

• which is as idiotic as it looks! (and the whole excessive metal bridge construct at the back)

Shall we cover the enormous iron cast frame of a piano with wood? Not really. Just some mid to high freq. absorb fluff would sufice.
That is how I think it should be, if that construction is to be presered.

But, ultimately, I think pianos should have one string per note. Therefore this iron frame can be reduced to a flat rim frame around the perimeter (some small ribs there might be needed) and again, covered with fluff. Something like this:

PASHKULI that’s fine. But again you are talking about what you disagree with regarding how pianos are designed. So go make an alternative instrument.

I’m not sure how any of this helps anyone.

PASHKULI what does “proper” recording mean? The better piano VIs already have multiple perspectives (some with lid closed, some with mics under the lid and the lid closed, etc).

Again, I’m not sure how your observations are helping anyone.

I understand you record your drumming and maybe guitars. Go find a grand and try and record that. Different perspectives. Try what you theorise.

DavisB Rach3master is on pianoclack I believe. His recordings (of NY Steinway D) are excellent, I'd ask him for his mic setup and that's all there is to it.

I was under the impression Rach3master was female, but you're right, their setup sounds great. I hear a lot of piano recordings on YouTube done with portable Tascams that sound good to my untrained ears. Those portable devices are around 150 bucks. Navindra (from the other forum) also has a setup that works very well.

God Bless,
David

dikrek

So go make an alternative instrument.
Go find a grand and try and record that. Different perspectives. Try what you theorise.

Aye‑aye, sir. Going tomorrow.
The thing is, I do not have the room, the piano. Anything that would make it possible. When I talk to people who have it, they react as if I am going to ruin their property.

Another possibility with regards to frame is to have the actual frame "sandwiched" between the laminates of the rim with variable width of inner lip., imagine a ring (rim) with an inside horizontal protrusion (radial). Inspired by the guitar truss rod construction.
Yes it will make it thicker, no doubt and the outside layers of wood will have spokes to connect to the inner laminate cover through small wholes in the cast iron rim. Proper tests must be performed initially on FEA software environment. But I prefer to stick to the "one string per note" construction. But this one with a metal rim should work too in this concept.

So many ideas, so little money\time. Yamaha have built a $2B Innovation centre for D&R. And I was not approved to go (three times I applied).

PASHKULI this https://www.estudiolibres.com.ar/media/files/Sound-Radiation-from-a-Grand-Piano.pdf and this https://www.yamaha.com/en/musical_instrument_guide/piano/mechanism/mechanism004.html are worthwhile reading for you.

dikrek

That one from Yamaha is filled up with marketing adjectives such as "attractiveness of the sound".
It is not true. Those only increase the metallic zing of the sound. An it is terrible. And also reflected by the iron cast frame and maybe by the pegs and pins (also metal).

Nevermind.

PASHKULI I’m really curious how you envision your instrument to sound since none of the actual pianos are to your taste.

And even more curious how you think your posts are contributing anything to the topic.

dikrek

Yes, it is off‑topic. Ban me. But also it is related to how pianos are sampled (recorded), so not entirely offtopic.
Actually, I have no idea how VSL sample their pianos. They seem to me like typical "family" type small bizz company.
And pianos are not their main focus which otherwise is the case with @VILabs Audio.

I lost track why I chimed in. Got much more work to do building "my company".

PASHKULI https://www.vsl.co.at/instruments/synchron-pianos/yamaha-cfx

https://www.vsl.co.at/instruments/synchron-pianos/bosendorfer-imperial

https://www.kvraudio.com/interviews/a-sound-history-touring-vsls-synchron-stage-52159

Good background info on how VSL samples pianos.

@PASHKULI ive shared this with you before. Go to 2:37.

If you can create a better recording, you could easily make great money doing that for a living.

If you just don’t like the sound, even in person, then we are back to you needing to make your own instrument.

Is the recording processed? They all are. Enhanced in places? Probably, especially the amazing bass. Does it sound awesome? To me it does.

The question is, can you do better?

If so you’d be doing the world a favour.

dikrek

Good background info on how VSL samples pianos.

I do not know. I think it is horrible. Ok, maybe just bloated. But that is also horrible

dikrek If so you’d be doing the world a favour.

No, I wont. This is not my path. This is not the clavier (klavier) instrument I want to build. It won't sound the same (most likely).
And yes, we do not know how those recordings were made as a final render for the video. Mics and such, you know.
So, I will say it is an enhanced presentation for marketing. Some of the notes sound like a broken glass to me.

Again, a klavier as an instrument is not just this monstrous centuries old (by) now goliath.