It is currently Thu Mar 28, 2024 4:31 am


Questions for Sample Set Developers

Sampling pipe organs and turning them into something you can play in Hauptwerk.
  • Author
  • Message
Offline

NicholasA

Member

  • Posts: 116
  • Joined: Wed Nov 15, 2006 3:50 pm
  • Location: Abbotsford, NSW, Australia

Questions for Sample Set Developers

PostSat Mar 24, 2007 7:20 pm

Note that I've moved this post from the organ sample sets section of the forum to here as I think this is a more appropriate section.

Hello all,

Just a few questions regarding the creation of samples for Hauptwerk 2:

a) When making multiple attack samples per note for Hauptwerk 2 (i.e. samples for different key velocities), should full length samples be created for each sample including their own loops? Or should a short sample be created that just includes the attack phase for that particular velocity and Hauptwerk will crossfade to the loop(s) in the "main" sample at the end of the transient?

b) (semi-related to a) When making relatively wet samples, we are advised to make several recordings of each note with different shorter lengths (i.e. staccato) as to avoid the "bounce" effect that is present when Hauptwerk crossfades from the middle of an attack transient to the start of the steady state release transient. We can then define multiple-release samples that will be crossfaded to dependant on how long the sample has been playing for.

Is there any general advice one could follow to do this recording effectively? How many different samples should be made? One for the long main sample and one staccato? Or one for the main sample and several shorter ones?

Some of the pipes on the St. Stephens Organ (particularly of the 16ft violon rank) suffer quite badly from this effect due to their slow speech and it makes them unsuitable for quick playing as they simply don't sound realistic.

c) When recording the key-noise, is it desireable to have zero key-noise in the actual pipe samples? (i.e. include it all in separate files) And should multiple recordings of the noise actually be recorded using different key velocities and then use these parameters in the ODF to configure the key-noise based on how hard the key was pressed? (I would imagine this would greatly improve realism especially in trills where the keys tend to be depressed faster?). If this is the case, should we remove key noise from the actual pipe samples?

d) Finally, this question relates to the user interface. I really think that the whole photo-realistic console idea is quite interesting, but I am much more concerned with the audio quality and can't justify spending any money what-so-ever on a fancy camera or flash equipment to photograph. I have a standard Olympus 4 megapixel digital camera with a 10x optical zoom (no-idea if that is useful or relevant information) and intend on making the photographs with that. I figure that normal photographs are going to come out with that extremely realistic look and I just can't see it looking that great on a computer (especially when moving components such as drawknobs and expression pedals are introduced). Several sets that i've seen seem to mix both the photo-realism with an obvious computer generated component like the lavender audio sets, I find these quite appealing and am wondering if that's a better option?

---

If we have multiple attack and release samples for each pipe AND multiple samples for the key-noise, we are talking a LOT of recording. Is there a point where one can go too-far in the recording process?

I've been offered to inspect an organ at another local church, it's again only a 2 manual instrument with only has one more stop than St. Stephens (but I consider it having a much more useful swell department) but I think it will be great for my first attempt at a purpose-built HW2 organ. It wasn't the organ I had planned to record, but an offering to record is not something I think I can dismiss!

Thanks,

Nick.
Offline

David Butcher

Member

  • Posts: 151
  • Joined: Tue Jan 24, 2006 12:47 pm
  • Location: Suffolk, UK

PostSun Mar 25, 2007 6:50 am

Hi Nick,

I don't think there's necessarily any one right way of doing things, but I can try to answer some of your questions based on my experience developing the Lavender Audio sample sets.

I chose to remove the key action noise from the start of samples wherever possible. Sometimes doing so affected the sample attack too much, so the odd noise is still present. I then recorded a whole load of key action sounds and set them to sound on a pseudo-random basis. Separating the action noise in this way gives people a chance to vary the level or even disable it altogether. However, if planning a thoroughly authentic recreation of an instrument, warts and all, I would consider keeping the action noise as part of the sample, especially if recording multiple attack samples.

I used a Canon 4 megapixel digital camera for all the organ artwork. The one other essential piece of equipment is a tripod ! The biggest difficulty I found was trying to fit everything into a window that would display on 1024x768 resolution. One tip - carry out all image editing and processing on a much higher resolution (say 3 times bigger) before resizing downwards.

Creating an HW2 organ does take a lot of time and I think you need to decide how far you want to go with all the realism features before starting. For example, although my sets don't feature multiple releases, an awful lot of effort went into getting the effect of the swell box accurate. Although I don't believe the sets suffer too much with just one release sample, if I were doing those recordings again, I would definitely include a short note release.
IMHO no v2 organ is ever truly completed ... there's always something extra that could be added or tweaked. Ultimately you have to draw a line somewhere !

Hope that helps and good luck in your endeavours.
David
Offline

NicholasA

Member

  • Posts: 116
  • Joined: Wed Nov 15, 2006 3:50 pm
  • Location: Abbotsford, NSW, Australia

PostSun Mar 25, 2007 3:48 pm

David Butcher wrote:...However, if planning a thoroughly authentic recreation of an instrument, warts and all, I would consider keeping the action noise as part of the sample, especially if recording multiple attack samples.


If all of the recorded samples contain the tracker action noise, wouldn't this decrease the realism as they would be layered with other stops containing the same action noise effectively doubling it's loudness? From my limited understanding of pipe organs and slider windchests, doesn't each tracker open a common valve into a column of the windchest that allows air into all the pipes which have their slider (row) in the right location? In this case, I would have thought that you would only want one action noise for all samples playing in a particular sample.

I could understand your way with a unified organ with electric action where every pipe has it's own valve.

David Butcher wrote:...Although I don't believe the sets suffer too much with just one release sample, if I were doing those recordings again, I would definitely include a short note release.


I'll utilise your previous experience and get short releases from now for my sample sets. But should pipes that take longer to speak have more release samples? Some of the pipes in the St. Stephens set took almost half a second to reach full amplitude. How many different releases should I take of that sample?
Offline

David Butcher

Member

  • Posts: 151
  • Joined: Tue Jan 24, 2006 12:47 pm
  • Location: Suffolk, UK

PostSun Mar 25, 2007 6:03 pm

Sorry, I didn't phrase that first bit very well. I was thinking that if the attack transients of a pipe sample were an integral part of what you were trying to sample (eg the chiff of a flute pipe) then it might be difficult to remove any associated action noise without compromising the transient you were trying to keep. As I wrote earlier, I would consider it - not necessarily do it ... :-)
It would be interesting to hear the views of other producers who have more experience of this sort of issue.

Some of the pipes in the St. Stephens set took almost half a second to reach full amplitude. How many different releases should I take of that sample?


I wouldn't like to say to be honest ! Ultimately, I think you may have to experiment a bit to achieve the effect you desire.
Offline
User avatar

ReinerS

Member

  • Posts: 869
  • Joined: Thu Oct 14, 2004 11:52 am
  • Location: Braunschweig, Germany

PostMon Mar 26, 2007 6:57 am

a) When making multiple attack samples per note for Hauptwerk 2 (i.e. samples for different key velocities), should full length samples be created for each sample including their own loops? Or should a short sample be created that just includes the attack phase for that particular velocity and Hauptwerk will crossfade to the loop(s) in the "main" sample at the end of the transient?

Hi,

this one is easy to answer. Currently in HW2 attack and sustain sample are one and cannot be separated. This means you will need a full attack and sustain sample (including loops) for each attack. AFAIK there are no plans currently to extend this. The main problem is that a simple crossfade as into the release will most likely not work and there is currently no software available that could compute a "jump" point between two samples (like a loop point).

Best regards
Reiner
Offline

NicholasA

Member

  • Posts: 116
  • Joined: Wed Nov 15, 2006 3:50 pm
  • Location: Abbotsford, NSW, Australia

PostMon Mar 26, 2007 6:06 pm

ReinerS wrote:Hi,

this one is easy to answer. Currently in HW2 attack and sustain sample are one and cannot be separated. This means you will need a full attack and sustain sample (including loops) for each attack. AFAIK there are no plans currently to extend this. The main problem is that a simple crossfade as into the release will most likely not work and there is currently no software available that could compute a "jump" point between two samples (like a loop point).

Best regards
Reiner


So how does that work with multiple releases? If for example I have several realease samples to be played back dependant on the time the sample has been playing. Clearly Hauptwerk cannot know which attack sample to play for that release because it wouldn't have the information necessary to know which release to play back prior to the player actually releasing the key.

This sounds rather bizzare to me. Can Martin Dyde please confirm if multiple attack samples require complete sustaining portions for each attack? It just sounds unnecessary as once the wave enters the steady state, this information would be redundant independant of the different attack transient (would it not?).
Offline
User avatar

ReinerS

Member

  • Posts: 869
  • Joined: Thu Oct 14, 2004 11:52 am
  • Location: Braunschweig, Germany

PostTue Mar 27, 2007 1:10 am

The problem here is that it requires a very clean crossfade. The assumption is that artefacts due to crossfading (e.g. phase cancellation in upper harmonics etc.) are less audible or even negligible at the release point than when going from the attack to the sustain phase. So it is much harder or even impossible to solve that with just a crossfade.

Obviously at the release point there is no chance of finding a "perfect" match point in the signal as for a loop, since one does not know a priori at which point in time the key is released. Hence the only way to get into the release is a crossfade and we all know the artefacts HW1 produced with that. Fortunately that is greatly improved in HW2.

Still I agree with Martin that crossfading from the attack into the sustain might be quite audible and so would not be a good choice.

As for the multiple releases: It is not a question of which attack sample to play for the release, but which release sample to play with the attack! So, when you release the key, the proper release sample is chosen. This also means that the attack phases of the staccato releases are actually never played, just the releases.

Hope this helps
Reiner
Offline
User avatar

mdyde

Moderator

  • Posts: 15441
  • Joined: Fri Mar 14, 2003 1:19 pm
  • Location: UK

PostTue Mar 27, 2007 5:49 am

Can Martin Dyde please confirm if multiple attack samples require complete sustaining portions for each attack? It just sounds unnecessary as once the wave enters the steady state, this information would be redundant independant of the different attack transient (would it not?).


You would normally want to record the full note sound (i.e. attack+sustain+release) for short notes as well as normal long notes, but you would only load the additional 'short-note' releases in Hauptwerk, not their attack/sustain parts.

I.e. there would be one entry for the pipe in the Pipe_SoundEngin01_AttackSample table but more than one in the Pipe_SoundEngin01_ReleaseSample table and you would use Pipe_SoundEngin01_ReleaseSample.ReleaseSelCriteria_LatestKeyReleaseTimeMs to tell Hauptwerk which release sample to play based on how long the player had been pressing the key.

I'd advise just making some test recordings and organ definitions with a single pipe recorded from a nearby organ to practice before recording a full instrument.

As Reiner says, it would be better still to have the full attack/sustain portions loaded to match the release samples (because the relative phases of the harmonics differ to some extent each time a real pipe sounds) but that would require Hauptwerk to know in advance how long the pipe would sound (impossible), hence you would normally just use a single attack/sustain sample shared between the releases for a given pipe.

Best regards,
Martin.
Offline

NicholasA

Member

  • Posts: 116
  • Joined: Wed Nov 15, 2006 3:50 pm
  • Location: Abbotsford, NSW, Australia

PostTue Mar 27, 2007 8:11 am

Hello,

I'm understanding what's being said now. Thanks for your patience with my mind!

Nick Appleton.

Return to Creating sample sets / recording organs

Who is online

Users browsing this forum: No registered users and 1 guest