It is currently Thu Jun 04, 2020 9:29 am


Why not use (semi-)dry set in combination with IR reverb?

Speakers, amplifiers, headphones, multi-channel audio, reverb units, mixers, wiring, ...
  • Author
  • Message
Offline

Mindenblues

Member

  • Posts: 29
  • Joined: Thu Nov 27, 2014 10:45 am

Why not use (semi-)dry set in combination with IR reverb?

PostWed Feb 05, 2020 10:58 am

Sorry for a maybe stupid question, anyway:
Instead loading this extremly RAM- and CPU-Power-consuming surround sample set, why not use the (semi-)dry set instead in combination with dedicated different convolution reverb settings for the different output channels?

Beside the primary advantage that the needed RAM and polyphony requirements are significantly smaller, I expect sound improvements. Because in my opinion it is a more realistic model approach if a common reverb engine is applied on the sum of dry samples (for a certain surround channel), instead that every sample in the sample sound mix bears its own reverb. A surround set will tend to sound more muddy this way, in case that many registers are used (all with different and own reverb parts, in opposite to the physics in reality). To have a common reverb (dedicated separately to each output channel) applied to all dry samples which are just used will tend to sound more defined and more clearer (compared with same amount of overall reverb), that is my opinion and also result of - subjective- audio tests.

So the model to have a common reverb model for every output channel (could and should be different ones of course, for the front and rear-channels and whatever other channels you have, subwoofer, body shaker and so on) seems to me the better sounding approach and needs much less PC ressources.

By the way, I checked the (semi-)dry Demo-Sampleset of Groningen. It sounds beautiful to me too, despite the fact that some notes e.g. in the Praestant 8 register are considerably uneven in the voicing and/or in the attack (I assume that is how the original organ sounds). It is up to the user whether that is considered as beeing charming or too uneven to be acceptable. In latter case one could try to fumble with the voicings (would be much effort).

Unfortunately even the extended keyboard range has only 51 keys, only up to d4. Many romantic pieces need more, even for Bach there are works which need e4. Also other famous historical organs, sampled from SP, reach in extended version up to at least f4, like e.g. Zwolle, or even g4 . So I don't see the point to not extend it to the commonly used range only because it is a historical organ (like many others which are extended to that range).

Beside that, really beautiful sounding set (only checked the demoset so far).

/BR Olaf
Offline

Mindenblues

Member

  • Posts: 29
  • Joined: Thu Nov 27, 2014 10:45 am

Why not use (semi-)dry set in combination with IR reverb?

PostFri Feb 07, 2020 3:19 am

IainStinson wrote:I prefer to use the original acoustic (even thought convolution reverb in HW 5 is excellent) - perhaps I should be more adventurous.


If that is your aim, to approach the original acoustic of this specific church room as close as possible by means of a Hauptwerk Sampleset, the question is, which is the best approach. It is not automatically a surround set what I will try to explain:

Iain, if you draw a single register, and play a single note, and use not a blend of microphone positions from the sample set, instead e.g. the direct or diffuse channel alone, I agree with you that nothing can be closer on the original acoustic as this channel of the surround set with the captured wave file of this single sample.

However, that is not what happens when you play the sample set. In reality, you will easily have 1000 or more wet samples which sound in parallel. And each and every wet sample bears its own reverb part. Therefore there is an audible tendency, that the sound of a large bunch of wet samples tend to get more muddy as it happens in the original acoustic. Because in first approach you have one church hall used for all pipes which are just ringing instead 1000 separate church halls (I know that this is simplified, but only to make the problem clearer, I hope you understand that).

Furthermore, the approach of blending different channels like e.g. direct and diffuse channel into a single pair of front loudspeakers is another thing what might loose 'originality', because it is not the same as if you move yourself between the microphone position of the direct or diffuse channel. There might be interferences happen with that kind of blending. But I must confess, that - for my ears - that this blending works surprisingly well with the SP multichannel sample sets this way, so at last for me, that is not the main point.

The main point remains, that if you aim for coming as close as possible for original acoustic and use considerable amount of polyphony, that you probably come closer to this aim, if you use e.g. any of the given IR model for Martinikerk Groningen (depending where you like to virtually listen) and use the (semi)dry set instead. Because the model to have a common reverb room for lots of dry pipes sounding is a closer model to the reality as a model to have lots of wet pipes sounding in parallel.

Even if we let aside in that discussion all the other advantages, like that dry sets need much less RAM, CPU power, are faster to load, are cheaper and have almost unlimited flexibility concering which reverb room you like to model, there is another point to consider:

The farer away from a pipe you record samples, the more you need to remove background noise from the samples.
I guess that dry samples need much less artificial treatment to remove background noise in the post processing of samples as wet samples because the signal to noise ratio is higher. That is also an another advantage.

At least for me the only thing is that I achieved several surround sample sets through the years and I am used to them, and I don't have a big dry sample set so far, only demo sets.

But I guess, in the future, I will not buy anymore a surround set, at least if there is a dry set available, based on what I experienced in the last month with HW5 and convolution reverb.

Sorry that this post here is of general nature and far away from discussion about that great Groningen Sample Set, and maybe an administrator could move this and maybe my post before to a separate thread if it seems interesting for others to discuss, what I wrote.

Best regards, Olaf
Offline

josq

Member

  • Posts: 792
  • Joined: Tue Jan 24, 2012 1:11 pm

Re: World famous organ of Martinikerk Groningen will come

PostFri Feb 07, 2020 4:43 am

Hi Olaf,
I think you made a couple of excellent points.

It boils down to the question: what is the most valid approach: first apply reverb (artificial or natural), and then add the sound of the individual pipes, or first adding, and then applying reverb?

As you say, each pipe has its own reverb. So adding the sound of the dry pipes first, and then applying a common reverb, would not be a valid approach in theory, and in my limited experience this is not a satisfying approach indeed.

On the other hand the sound of each pipe is modulated based using the wind model, tremulant model and swell box model. Modulating after applying reverb/on wet samples, as Hauptwerk currently does, is not valid either.

A promising future approach is to apply per-pipe reverb. The workflow will be to start with (semi-)dry samples, then apply the models, then apply reverb, and finally adding the sound of the pipes. The challenge in this approach will be to correctly record the reverb of each pipe. It is not clear to me how sample set producers might address this challenge.
Offline

Mindenblues

Member

  • Posts: 29
  • Joined: Thu Nov 27, 2014 10:45 am

Why not use (semi-)dry set in combination with IR reverb?

PostFri Feb 07, 2020 5:23 am

josq wrote:As you say, each pipe has its own reverb. So adding the sound of the dry pipes first, and then applying a common reverb, would not be a valid approach in theory, and in my limited experience this is not a satisfying approach indeed.


Hi Josq,

no, this is not what I said, I said just the opposite:
Each pipe has its own reverb in a wet sample. But it has almost no reverb in a dry sample. And in the reality, in the church, at least the farer the distant is to the organ, it is more and more only one room what has a reverb, and that is applied to the sum of all dry sounds, since all pipes are dry at the origin of the sound source. They get wet due to the common church room acoustics.
I know that it is very much simplified, but it shows the main differences between the model of using surround/wet sample sets and a dry set with applied IR reverb (differently to each output channel in a surround loadspeaker setting of course).
Therefore I consider the modeling of dry set in combination with IR reverb model as something what could come closer to the original sound, at least if many samples are running in parallel (means, much polyphony) and not only some. But that is the case normally, you don't hit a single note of a single register only, and wait until the reverb faded out before you hit another key...

And, consequently, it seems also a better approach to reality, if wind model, tremulant, swell boxes and all other things which are close to the point of the sound source, are first added, before the common IR reverb model is applied.

@admin: I don't know how to do, but it could make sense to move the posts which started with "Why not use (semi-)dry..) into a separate thread?
Offline

Mindenblues

Member

  • Posts: 29
  • Joined: Thu Nov 27, 2014 10:45 am

Why not use (semi-)dry set in combination with IR reverb?

PostFri Feb 07, 2020 5:41 am

I just sent a mail to the admins with that content, because an off topic theme has developed (I am sorry for that, but might be of common interest):

Dear admins,
I think that an interesting discussion has started, about the comparision dry and wet sample set usage.
But that has nothing to do with the tread theme of Groningen Sample Set.
If possilble, please move all posts which started with subject "Why not use (semi-)dry..."
to an own thread with that subject.
/BR
Olaf
Offline
User avatar

mdyde

Moderator

  • Posts: 12087
  • Joined: Fri Mar 14, 2003 2:19 pm
  • Location: UK

Re: Why not use (semi-)dry set in combination with IR reverb

PostFri Feb 07, 2020 6:09 am

Hello Olaf,

I've split your comments off to this new thread as requested.

There have been lots of threads on the forum on 'wet vs. dry' sample sets over the years, which you might find interesting. As josq mentioned, dry sample sets with an appropriate separate/different impulse response reverb per pipe could be acoustically/mathematically correct (but mixing dry samples together then applying a common impulse response reverb isn't theoretically correct acoustically, since you would effectively be making them all appear to sound from the same point in the virtual space). With a common overall reverb applied and in a dry stereo/surround listening environment, semi-dry sample sets are likely to sound more 'correct' than fully-dry ones, since they will contain some early reflections in their samples, thus helping to maintain spatial impression. Wet sample sets (if recorded from fixed microphone positions, and without mixing) are also acoustically correct in terms of spatial impression, since each pipe will be heard to emanate from its correct distinct point in the virtual space, with the in-sample acoustics adding correctly. (However, there are other compromises with wet sample sets, such as interaction of tremulants with the acoustic, and possibly lower lower signal-to-noise ratio as you mention.)

(I don't have time to be involved in this discussion myself, I'm afraid, but as above there are lots of other threads on it.)
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Offline

Mindenblues

Member

  • Posts: 29
  • Joined: Thu Nov 27, 2014 10:45 am

Re: Why not use (semi-)dry set in combination with IR reverb

PostFri Feb 07, 2020 6:32 am

Many thanks, Martin, that you moved the posting to that thread here!

mdyde wrote:As josq mentioned, dry sample sets with an appropriate separate/different impulse response reverb per pipe could be acoustically/mathematically correct (but mixing dry samples together then applying a common impulse response reverb isn't theoretically correct acoustically, since you would effectively be making them all appear to sound from the same point in the virtual space). With a common overall reverb applied and in a dry stereo/surround listening environment, semi-dry sample sets are likely to sound more 'correct' than fully-dry ones, since they will contain some early reflections in their samples, thus helping to maintain spatial impression. Wet sample sets (if recorded from fixed microphone positions, and without mixing) are also acoustically correct in terms of spatial impression, since each pipe will be heard to emanate from its correct distinct point in the virtual space, with the in-sample acoustics adding correctly. (However, there are other compromises with wet sample sets, such as a interaction of tremulants with the acoustic, and possibly lower lower signal-to-noise ratio as you mention.)


I believe that you are right that it is a compromise to consider the dry sets coming from one distinct point (well, at least they are recorded stereo), but seen from farer distant in the room that compromise disappears more and more with the distant.

Furthermore, I agree that considering only a single wet sample all the spatial informations are captured more realistic if no IR reverb model is used, instead simply to take what the microphone captured.

However I disagree if we consider what happens if we use many wet samples in parallel which all bear their own reverb parts. All that single spatial information will be simply mixed together and at the end you have a mingle mangle mixed overall sound. Means, that distinct information gets less and less distinct. And that in combination that you mix in a surround set e.g. direct and diffuse channels in order to approcimate several different sound positions. That mingle mangles it even more.
My feeling is that the more registers and polyphony are in use with wet sample sets, the more the overall sound moves away from the real sound (at the position where the samples where captured), tends to get more muddy. And from my first exeriences last month with dry set and IR there seems to be an audible improvement in case many samples are just used (that it tends to sound more distinct with same amount of reverb for comparision).

Best regards, Olaf
Offline
User avatar

mdyde

Moderator

  • Posts: 12087
  • Joined: Fri Mar 14, 2003 2:19 pm
  • Location: UK

Re: Why not use (semi-)dry set in combination with IR reverb

PostFri Feb 07, 2020 6:45 am

Hello Olaf,

I assure you it definitely *is* correct acoustically/mathematically to add multiple (appropriately recorded) wet samples together; the reflections add correctly -- it definitely does work. In fact, it's the main reason that Hauptwerk was developed in the first place, and likewise most sample set producers do what they do because of it.

(Sorry -- I really can't be involved in this discussion further myself, though.)
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.

Return to Amplification

Who is online

Users browsing this forum: No registered users and 2 guests