Loading more than one perspective in Stereo

With the availability of so many sample sets that make available multiple perspectives, and the recent apparent trend of many loading multiple perspectives of the same organ in Stereo, I wanted to ask how many people prefer this over loading just one perspective at a time (when using stereo)?

I ask this because of the following article:

http://www.colinpykett.org.uk/test-your ... References

In theory, loading multiple perspectives in Stereo should also increase intermodulation distortion. I think I hear it, and so I often try to load more than one perspective at a time, but I usually go back and end up loading only one perspective (usually a “front” perspective). To my ears, this decreases what I perceive to be intermodulation distortion, even if only two perspectives are loaded, and the sound to me is much more clear. Does anyone else have this experience? Judging by Colin Pykett’s articles, it appears that some people can hear this kind of thing more readily than others, so that may explain why many don’t appear to notice it.

Of course, in theory, loading an organ with convolution reverb should also increase intermodulation distortion compared to the same organ sampled with the reverb already present. This, however, I don't think I have been able to detect. Has anyone else noticed this? Still, after initially using convolution reverb a lot when I upgraded to Hauptwerk V, I now rarely use it. Generally, the originally sampled files sound better voiced to my ears than the same sounds fit into a new acoustic. Has anyone else noticed this? I haven't tried any sample sets with their original acoustics added, that would be a good test.

Of course, this doesn't apply to anyone that is able to output different perspectives to different speakers. Still, that appears to be a minority of users, however.

Of course, this doesn't apply to anyone that is able to output different perspectives to different speakers. Still, that appears to be a minority of users, however.

There is a phenomena referred as air mixing. Where even tho there is no electrical or electronic connection, these air borne sound waves interact with each other. Beat frequencies for instance?

Sometimes a positive and at times, not. Sometime ago, I listened to an installation that sounded "strange". The issue wasn't intermod, but basically phase cancellation. My ears were feeding info to the brain, but the brain wasn't ok with it. Rather rejected the info and labeling it as weird. But what was even more interesting was when moving just short distance, everything sounded good. The owner didn't except the "issue" because maybe from the bench, his ears were at the correct phase angle. Then there is a given thst the room is a very important part of the sound "stage".. (Early and late reflections.). Had a conversation with Drew about adding speakers at the rear of the sanctuary. Not in his plan for probably several reasons.

Rgds,
Ed

Intermodulation distortion consists of differences between frequencies. If two tones have the same frequency, there can't be any intermodulation distortion.

So playing multiple perspectives at the same time should not introduce any intermodulation distortion. Neither should the addition of reverb. However, the random detuning and wind model in Hauptwerk might apply different amounts of detuning to the different perspectives, resulting in beating and intermodulation distortion even when playing a single note on a single rank.

It is highly desirable that this problem will be solved in a near-future version of Hauptwerk.

Generally, the originally sampled files sound better voiced to my ears than the same sounds fit into a new acoustic. Has anyone else noticed this? .

Convolution reverbs do not only add reverb, but also act as a frequency filter. They will boost some frequencies (because they are boosted in the recorded space), and this boost is easily excessive for some frequencies.

Effectively, with convolution reverb you are undoing the voicing (balance between frequencies) of the original. Unless you use a convolution reverb that was recorded in the same space as the organ.

[mdyde]

So playing multiple perspectives at the same time should not introduce any intermodulation distortion. Neither should the addition of reverb. However, the random detuning and wind model in Hauptwerk might apply different amounts of detuning to the different perspectives, resulting in beating and intermodulation distortion even when playing a single note on a single rank.

It is highly desirable that this problem will be solved in a near-future version of Hauptwerk.

Hello josq,

Yes -- adding more native support for handling playback of multiple perspectives (maintaining exact pitches/phases/loops/etc. between perspectives for any given pipe) is something we do very much aim to implement as a very high priority in the future.

Hello josq,

Yes -- adding more native support for handling playback of multiple perspectives (maintaining exact pitches/phases/loops/etc. between perspectives for any given pipe) is something we do very much aim to implement as a very high priority in the future.

Very promising & very much appreciated!

I still believe that intermodulation distortion applies, because when sound reflects off a surface, it isn’t perfectly reflected at the same frequency. There is some frequency scatter, particularly at the higher frequencies, and this can contribute to intermodulation distortion.

I still believe that intermodulation distortion applies, because when sound reflects off a surface, it isn’t perfectly reflected at the same frequency. There is some frequency scatter, particularly at the higher frequencies, and this can contribute to intermodulation distortion.

That was my (above) point.

Rgds,
Ed

I still believe that intermodulation distortion applies, because when sound reflects off a surface, it isn’t perfectly reflected at the same frequency. There is some frequency scatter, particularly at the higher frequencies, and this can contribute to intermodulation distortion.

I think I find that hard to believe. At least this is not how convolution reverb works. As for real reverbs, when you play a single note on a 1' stop in a reverberant space, you would hear non-harmonic frequencies due to scattering. And as a secondary effect, you would hear additional difference tones due to IMD. Any recording of a high-pitched pipe would contain the scattered non-harmonic frequencies. I don't think that is the case.

Well, actually there are a lot of non-harmonics in pipe samples, but because of tracker noise, wind noise, pipe chiff, etc.

Josc: if you find it hard to believe, then I encourage you to analyze the recording of a pipe recorded in a reverberant space. I use Adobe Audition. It has a frequency viewer, that lets you see the frequency vs amplitude for a given selection of audio. When you do this, it’s actually challenging sometimes to identify discrete peaks for the pipe’s harmonics, because of the undulation of the higher frequencies due to scatter. Any other explanation for this phenomena would be appreciated if you are more knowledgeable about why this happens.

Josc: if you find it hard to believe, then I encourage you to analyze the recording of a pipe recorded in a reverberant space. I use Adobe Audition. It has a frequency viewer, that lets you see the frequency vs amplitude for a given selection of audio. When you do this, it’s actually challenging sometimes to identify discrete peaks for the pipe’s harmonics, because of the undulation of the higher frequencies due to scatter. Any other explanation for this phenomena would be appreciated if you are more knowledgeable about why this happens.

Yes, visualizations (spectrograms) of frequencies and amplitude (over time) suffer from an "uncertainty principle". Either you have high time resolution or high frequency resolution, but not both. Usually it is a compromise: frequency resolution is to some extent sacrificed to get more time resolution. As a result frequencies are smeared/scattered in the visualization.

I don't know about Adobe Audition, but in other programs you can choose where and how much to compromise by defining an IR window.

Note that this is a visualization problem only: the fact that we never able to depict music with very high time and frequency resolution does not mean that anything bad happens with the music itself.

[mdyde]

A quick thought: it's perhaps also worth noting that no pipe will have perfectly stable/constant pitch (especially as it starts and stops sounding), and as pitch changes occur they will be delayed and replayed by the room acoustic, so that they're heard simultaneously with the later/earlier pitches, resulting in the various slightly-different pitches being heard simultaneously, which will also yield some 'smearing' of the pitch when heard or measured.

I’m pretty sure there is a small amount of scatter on some surfaces, and this is part of what gives a certain acoustic their characteristics. Josc, you don’t have to believe me, but you can use the scientific method to disprove or prove me right. Changing the sensitivity on a frequency spectrum smooths the results, but it doesn’t add or remove frequencies. Even if frequency scatter isn’t a thing, when you add two recording sources, you still more than double intermodulation distortion when you play more than one note.

Colin Pykett's experiment is interesting but I would suggest an extension to his test with C in the left channel and D# in the right. Do use equal temperment for best results. Perform this test with speakers small enough that you can easily move them apart at arm's distance and then move them as close together as possible. The sound will be remarkably less pleasant as the speakers are moved close together. What you hear is NOT intermodulation distortion but just plain normal intermodulation where the two sound sources are arriving at both ears with essentially the same relative levels and phase relationships. The sound will change a lot but it is not distortion. Rather, the ears are no longer able to differentiate between the two sound sources and the brain can no longer work its magic at correcting the existing intermodulation (not IM distortion).

This is why multichannel audio in HW works so well. First, essentially, we preserve the stereo presentation of audio signals. Second we do a lot less combining of different pipes in the same audio channels. This is why I do not play back a second audio perspective through the same channels. For example, if the direct perspective is routed to speakers 1, 2, and 3 then I prefer to route the diffuse channels to speakers 2, 3, and 1 so that the multiple samples of the same pipe play through different speakers.

I've offered this same analysis twice before but have not yet seen a response from a person who actually tried this. All the theorizing in the world means little if we don't actually repeat a test and listen to the results.

Thank you for proposing the experiment. I have tried it with 1046Hz and 1244Hz sine test tones in the left and right channels separately (using the free program REW).

My findings are: when the speakers are far apart, the amount of intermodulation is relatively low, but highly variable depending on the exact position of my head. When the speakers are close, the amount of intermodulation is higher, but much less variable.

I think your experiment nicely covers the ground between the 2 extremes in the Pykett demo: our head generates high intermodulation when hearing different tones through the same ear, but intermodulation is very low if we hear them through different ears.

My conclusion is that having many, many ears would be beneficial - if we can get headphones with a corresponding amount of earcups