Digital Organ versus Pipeorgan

In the latest issue of the German expert journal " Ars organi" youn can read a remarkable scientific article in digital Organ versus pipe Organ. It explains why experts don't hear a difference between digital and pipeorgan, when there are only a few registers, but as the number of registers increases the sound becomes more an more artificial. Perhaps this article can be translated. My intension: perhaps Martin can create an appropiate updating, that simulate these connections.

Sorry, digitalorgan versus Pipeorgan

In the latest issue of the German expert journal " Ars organi" youn can read a remarkable scientific article in digital Organ versus pipe Organ. It explains why experts don't hear a difference between digital and pipeorgan, when there are only a few registers, but as the number of registers increases the sound becomes more an more artificial. Perhaps this article can be translated. My intension: perhaps Martin can create an appropiate updating, that simulate these connections.

I'm interested!

I think there are 2 big factors that determine the success of a virtual pipe organ when using large registrations.

1. The quality of the wind model. In my opinion, a majority of sample sets do not have a proper wind model implementation.

2. Surround setup. In a real church, the sound of a full organ is a very immersive experience. I think this experience can only be approached with a proper surround setup.

Dear Josq,
unfortunately, my English is not good enough to translated this specialist article . It is mathematically calculated, how the error quotient increases with each additional Register. Here it is shown each pipe in the Pipeorgan interacts with the others, even with a solo instrument or human singing, and this cannot happen with the digital version. Hence the liveiness of the Pipeorgan and the "sterility" of the digital organ of registration increases. Of course the windmodel and the suround effect contribute to the sound quality, but the main cause is the lack of interaction. Jiri Zurek is already trying to imitate this.
My queston to Martin is: isn't it possible in HW to simulate this using appropiate algorithms? That would be better for me than updating too many technical features. Thanks. ( sorry, using Google translater)

[mdyde]

My queston to Martin is: isn't it possible in HW to simulate this using appropiate algorithms?

Hello münsterorganist,

I haven't read the article you mention (and my German wouldn't be good enough anyway), but we do have logged as a long-standing enhancement request that it would be useful for Hauptwerk to attempt also to model the 'pitch-pulling' effects that real pipes exert on each other to some extent (especially on physically nearby pipes) when sounding together. Real pipework interacts in more complex ways than could fully be modelled with sample-based systems such as Hauptwerk, e.g. affecting the magnitudes and pitches of pipes' harmonics individually and dynamically. Hauptwerk's wind supply model also helps, in that when multiple pipes sound simultaneously the effect won't be quite the same as when they sound individually, as modelled pressures vary with their wind demands.

Dear Martin,
many thanks for your answer. The journal article calculates exactly, how high the error quotient will be as the number of Register increases - lots of numbers. But I'm glad you've already got your eye on the problem. I'm looking forward to a corresonding update in the hopefully not too distant future . Thanks for your great work, and of course I already have HW 8.

[mdyde]

Thanks very much as always, münsterorganist.

I would like to comment on the article by the author Martin Ehlbeck. To introduce myself: I am a professor of physics and am familiar with the physics of coupled oscillators. When we model the properties of two or more coupled pipes, we do so by describing them as coupled nonlinear oscillators.

The author presents a quantitative argument for the difference between the electronically sampled sound of a pipe organ and its original sound. He does this by counting all possible pairs of sounding pipes with close frequencies. He develops this argument from the well-known phenomenon that the frequencies of two nearby pipes, slightly out of tune with each other, attract each other. In the language of physics, we would rather call this phenomenon synchronization. It is a well-known phenomenon that can be observed in many physical systems.

The author more or less correctly names the underlying basic phenomenon, that is the coupling of two pipes in close neighborhood and with close but slightly different frequencies - he calls synchronization a merging process. However, he seems not to be aware that, first, the coupling process is a threshold process, i.e., it works only above a minimum amplitude and coupling strength. Secondly, he neglects that the coupling strength decreases rapidly with the distance between the pipes. Thus, synchronization occurs only between pipes of close frequency and sufficiently close spatial distance from each other. Therefore, his argument to count the number of errors in the digital sound image is nonsensical.

The whole subject is well covered in the profound references cited by the author. I wish the author had digested their content better.

Dear Orgelton,
thanks for the comment. It would be nice, if you would pass on your statement to "Ars Organi" editorial team. I'm very excited about the dispute. The fact is, that with all digital organs the sound fidelity decreases as the number of registers increases. As a layperson, I cannot judge why. I hear the difference every day on our large pipe organs, that I play and in the afternoon at home at my HW console, where I really enjoy practicing. Thanks for the contribution.

A very interesting topic indeed! I do agree that simply counting the pipes in the entire organ that are close in frequency cannot be the answer, the interaction between pipes I would think though is definitely one item out of many that make digital organ sound different from the pipe organ especially for large registrations.

Being an engineer of course I pondered quite a bit about what might be the main culprit, and my attempt at an explanation is based on one of the biggest differences between the real thing and the digital organ: In the real organ each pipe is a unique sound source, and as a result the mix between them is different at any position in the room. It is different between the left and right ear, and even when you move your head only very slightly there will be difference in the mixing between at least some of the pipes sounding. When the mixing is done electronically, in the worst case into just a single stereo pair (e.g. headphones), then clearly this effect is gone. The mix is stationary, phase cancellations are therefore more static and not as dynamic (especially for higher frequencies with a wavelength of just a few cm), and my feeling is that here is the major cause for the difference. The more channels you add (with their signals then again being mixed dependent on the exact listening position), the better the result gets.

What does this mean for the home user?
I've a dream of an unobtrusive speaker bar on the ceiling, where I can align all speakers to one listening position without interference at the touch of a button with a cell phone app

When I read through the article about LineArrays on https://de.wikipedia.org/wiki/Line_Array, such systems avoid interference. Are there also solutions that are suitable for rooms with 25-30 qm?
How or with which system can I currently achieve the best possible sound result?

Hallo Rainer,
it is true, that multi channel operations brings a significant improvment. We have a HW installation in our Musicschool. At first we just wanted stereo. The object seemed to be fail, because the sound in the room had nothing to do with an organ. Now we have 8 channel radiation front, top, middle, back. It's ok. Still, I'm missing the final "kick" and I was wondering, whether HW could solve this with certain algorithms. Thanks for your contribution.

Deleted by author.

Surely one of the most important factors for realism is random detuning. When that isn't applied to a sufficient degree (which, sadly, is so often the case) then that goes quite some way to explaining why the realism of a virtual pipe organ is inversely proportional to the number of ranks being played.

I fully agree with this, which is the reason all my sample sets feature a dedicated pseudo-random detuning. The built-in random detuning function of HW is not suitable for multi-channel sample sets, as front and rear channels of the same pipe are not guaranteed to be detuned in a synchronized fashion, which leads to individual notes starting to beat. Hence I implemented a detuning feature where the detuning is by a random amount, but the random nuber is computed during ODF generation and fixed int he ODF and it is assured that samples belonging to the same pipe are detuned by the same amount.

Reiner