It is currently Fri Mar 29, 2024 2:21 am


OPEN LETTER TO SAMPLE SET PRODUCERS

Existing and forthcoming Hauptwerk instruments, recommendations, ...
  • Author
  • Message
Offline

josq

Member

  • Posts: 912
  • Joined: Tue Jan 24, 2012 12:11 pm

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 3:04 am

I believe Adri has a point, although the discussion got confused a bit by the mention of 6-channel recordings.

To me, the most clear illustration is tremulants. Basically, a tremulant is a device to fluctuate the wind supply. By now, we all know the difference between sampled and modelled tremulants. It shows that tremulant (= wind supply) behavior can have a far more complicated an characteristic effect than can be modelled until now. One reason is that in practice, first the pipe is affected by the tremulant, and then the acoustics mixes in, while in a model, first the acoustics have been mixed in, and then the tremulant waveform is applied.

In a wind model, other effects should be included, for example the effect of playing certain pipes on other pipes, as explained by Adri. I presume these effects can be included in a model, to some extend. But how to capture these effects in Hauptwerk?

I think wind modeling will work best when applied to dry samples, followed by a convolution reverb. However, until now the drawback is that one convolution reverb is applied to all samples. This makes me wonder: what if the impulse response of each pipe is recorded separately, in addition to the dry sound of the pipe itself? Is Hauptwerk able to process such a recording? How realistic would it sound? And would this indeed enable more realistic tremulant and wind models?
Offline

adrianw

Member

  • Posts: 160
  • Joined: Sun Dec 18, 2011 5:10 am
  • Location: Lincolnshire, UK

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 3:13 am

When I say that organs on CDs sound better than in HW, I was referring only to the behavior of the wind, nothing else


Not something that I nor CD reviewers normally notice or comment on!

For example, let's say I play a Holpijp/Gedackt 8' only, and let's say I'm playing a chord or even just one note with the right hand, and then in the left hand I'm playing up and down passages. Most certainly in baroque and earlier organs with wedge shaped bellows, I am going to hear the differences in the right hand notes; they will be ever so slightly affected. Now that's natural wind behavior. This we don't have yet in HW.


Sure we do. This is exactly the facility that the HW wind model provides. While I am sure Martin could give a fuller answer, my understanding is that HW provides for the wind supply parameters to be given, the reservoir modelled, the wind demands of all the notes played at any instant calculated and the sound engine programmed to give a resulting pitch and amplitude modulation of each of the pipes individually.

I'm with Bach, so I turn it off for all but baroque sample sets, because I think sample set makers generally over estimate wind instability and impair realism with an over sensitive response. And it seems to take quite a bit of CPU. Of course, it could also be left on but dialled down (or up if you like the effect).

The principal wind effect that is not (yet) modelled by HW is nothing to do with wind supply, but to do with sympathetic acoustic effects one pipe to another (a closer analogy to your piano sympathetic resonances). These seem to be related to both air disturbance and wind-chest toe-board transmission. HW has some of the necessary data (pipe positions etc) that in theory might make this possible to model but I doubt whether the fundamentals are yet understood sufficiently to allow a practical implementation. Of course, Martin might surprise us in a future version.

- Adrian.
Offline
User avatar

mdyde

Moderator

  • Posts: 15444
  • Joined: Fri Mar 14, 2003 1:19 pm
  • Location: UK

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 4:06 am

adrianw wrote:
adri wrote:For example, let's say I play a Holpijp/Gedackt 8' only, and let's say I'm playing a chord or even just one note with the right hand, and then in the left hand I'm playing up and down passages. Most certainly in baroque and earlier organs with wedge shaped bellows, I am going to hear the differences in the right hand notes; they will be ever so slightly affected. Now that's natural wind behavior. This we don't have yet in HW.


Sure we do. This is exactly the facility that the HW wind model provides. While I am sure Martin could give a fuller answer, my understanding is that HW provides for the wind supply parameters to be given, the reservoir modelled, the wind demands of left and right hand calculated and the sound engine programmed to give a resulting pitch and amplitude modulation of each of the pipes individually.


Yes -- Hauptwerk actually can and does already model those effects (depending on how the wind supply model is configured in the organ definition by the sample set creator). For example, try holding a high-ish note on the St. Anne's Great Open Diapason Large 8 and then playing short notes in the lowest octave (or on another large-diameter rank on the same wind-chest) and you should hear the pitch, amplitude and harmonic content (brightness) of the higher held note fluctuate subtly.

adri wrote:Before the age of electricity, organs were pumped by humans, whether of the old wedge shaped type or the magazine bellows type. We seem to forget this.

For example, take the sample set of Kreszow with its 32' and plethora of 16' stops in the pedal; could we in real life in the 18th century draw all these pedal stops at once? And all the stops to create a tutti in the manuals? No, the bellows would not be able to handle that at all.

Are we ever going to have the virtual equivalent of human powered winding? An intelligent system that would respond differently according to how many stops were drawn? And increase the virtual peed of pumping when more stops are drawn? And would produce audible windsag when the organist draws too many stops?


Wind-sag and automatically-pumped bellows can be modelled too, and some sample sets have the latter. If you play chords on St. Anne's with all stops and couplers drawn you will hear a small amount of wind sag. Some historical sample sets have automatically-pumped wedge bellows, e.g. from Milan Digital Audio and Sonus Paradisi (and also OrganArt Media, if I recall correctly).

Theoretically, hand-pumped bellows (e.g. harmoniums) can also be modelled, although I'm not aware of any sample sets that do that currently, presumably because of the issue of the physical hardware needed for the user to control the pumping action.

Hauptwerk's wind supply model is actually already very powerful (based on fluid dynamics air-flow physical modelling) in terms of modelling the flows within the organ, and in each pipe. I would say that the most significant improvements could be made by using dry samples with unique impulse responses applied in real-time for each pipe separately (so that the dynamic effects of the wind model, tremulants, swell box changes, etc. would interact correctly with the acoustics, as josq mentioned), and by more detailed modelled of the pipes' frequency/phase changes in response to their air-flow changes (beyond the current pitch, amplitude and brightness modulations).

josq wrote:To me, the most clear illustration is tremulants. Basically, a tremulant is a device to fluctuate the wind supply. By now, we all know the difference between sampled and modelled tremulants. It shows that tremulant (= wind supply) behavior can have a far more complicated an characteristic effect than can be modelled until now. One reason is that in practice, first the pipe is affected by the tremulant, and then the acoustics mixes in, while in a model, first the acoustics have been mixed in, and then the tremulant waveform is applied.
...

I think wind modeling will work best when applied to dry samples, followed by a convolution reverb. However, until now the drawback is that one convolution reverb is applied to all samples. This makes me wonder: what if the impulse response of each pipe is recorded separately, in addition to the dry sound of the pipe itself?


Yes -- that would be a key longer-term goal, as computers become sufficiently powerful to make per-pipe real-time convolution feasible.

josq wrote:Is Hauptwerk able to process such a recording?


Hauptwerk doesn't have convolution capability natively (per-pipe or otherwise) currently, but I do very much hope we will be able to add per-pipe convolution in the future. It would require a huge amount of computing power for organs and acoustics of any size. Plus of course it's a big task to implement, and a big (and probably logistically difficulty) task for sample set producers to record a separate real per-pipe impulse response (FIRs) from every real pipe's physical position. (Per-pipe FIRs generated from modelled acoustic spaces would probably be easier, and could potentially also be very realistic.)
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Offline

adri

Member

  • Posts: 1545
  • Joined: Tue Aug 31, 2004 6:33 am
  • Location: Belgrade, Serbia

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 4:16 am

Adrian wrote:

The principal wind effect that is not (yet) modelled by HW is nothing to do with wind supply, but to do with sympathetic acoustic effects one pipe to another (a closer analogy to your piano sympathetic resonances). These seem to be related to both air disturbance and wind-chest toe-board transmission. HW has some of the necessary data (pipe positions etc) that in theory might make this possible to model but I doubt whether the fundamentals are yet understood sufficiently to allow a practical implementation. Of course, Martin might surprise us in a future version.


The interference between pipes has to do with:
a) layout of the pipes on the windchests (e.g. chromatic vs. another arrangement, like C/CIS, etc.)
b) how compact the organ is; thus how close the pipes at packed together: sardine can or spacious cavern?

Wind behavior also influenced by the type of windchest used, as we have, in old organs, either:
1) spring chests (e.g. Stade Huss/Schnitger organ still has them)
2) upper slider chests (e.g. Zeerijp)
3) normal slider chests

When the normal slider chest was introduced in the western Netherlands, there was a lot of protest against it, especially the remaking of the Hagerbeer organ in Alkmaar by F.C. Schnitger. The conservative folks believed that the wind behavior was much better in Springchests, as the wind didn't hit the pipes so directly in a straight wind line. The unique upper slider chest system found in Zeerijp provides a "curved" path for the wind, which the builder Theodorus Faber in the mid-17th Century seems to have made as a kind of transition stage from springchest to the common slider chest.

While these differences may be more academic than audible to our hears, it nevertheless is worth mentioning here as well.
Offline

OrganoPleno

Member

  • Posts: 652
  • Joined: Thu Dec 03, 2009 4:08 pm

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 9:29 am

mdyde wrote:Plus of course it's a big task to implement, and a big (and probably logistically difficulty) task for sample set producers to record a separate real per-pipe impulse response (FIRs) from every real pipe's physical position. (Per-pipe FIRs generated from modelled acoustic spaces would probably be easier, and could potentially also be very realistic.)


On some future version of Hauptwerk, we'll be able to CHOOSE either the sampled FIR of each pipe (if our new CPU is up to the task), or a modelled FIR (for those with more limited CPU capacity)... just like we do now for sampled vs modelled Tremulants.

adri wrote:Wind behavior also influenced by the type of windchest used, as we have, in old organs, either:
1) spring chests (e.g. Stade Huss/Schnitger organ still has them)
2) upper slider chests (e.g. Zeerijp)
3) normal slider chests.


On some future version of Hauptwerk, we'll be able to CHOOSE which style of Windchest to apply... just like we do now for different listening positions. And all of the physical interaction between the Pipes... how the playing of one pipe "draws" the pitch of another near-by pipe... will be perfectly modelled as well... just as soon as all of the complex Physics involved is properly understood!
Offline

student

Member

  • Posts: 25
  • Joined: Thu Jul 04, 2013 4:27 am
  • Location: Timisoara, Romania

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 10:28 am

mdyde wrote:Hauptwerk doesn't have convolution capability natively (per-pipe or otherwise) currently, but I do very much hope we will be able to add per-pipe convolution in the future. It would require a huge amount of computing power for organs and acoustics of any size. Plus of course it's a big task to implement, and a big (and probably logistically difficulty) task for sample set producers to record a separate real per-pipe impulse response (FIRs) from every real pipe's physical position. (Per-pipe FIRs generated from modelled acoustic spaces would probably be easier, and could potentially also be very realistic.)


Martin, have you looked into using CUDA for this? It is way off my current job requirements, but if I had any spare time not used for organ practice (I barely have time for the organ) I would love to learn more about this topic, but until then it is only an uninformed question. I guess it ultimately boils down to how efficiently can a convolution be implemented on a CUDA core and how well does it scale to thousands of cores processing thousands of pipe signals and their corresponding (different) infinite impulse responses. So a powerful graphics card would make it possible to create a realistic pipe organ!
Offline
User avatar

mdyde

Moderator

  • Posts: 15444
  • Joined: Fri Mar 14, 2003 1:19 pm
  • Location: UK

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 11:15 am

Hello student,

Thanks for the thoughts. I don't think we'd really want to get into detailed discussions about how Hauptwerk might work internally, but we're certainly aware of graphics card (GPU)-based processing possibilities and their advantages and disadvantages.
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Offline

1961TC4ME

Member

  • Posts: 3144
  • Joined: Mon Feb 07, 2011 1:45 pm
  • Location: Lake Minnetonka, Minnesota

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 3:15 pm

I listened to a number of the examples of this set on CCH and out of them all I feel this one sounds the best >>> http://www.contrebombarde.com/concertha ... usic/27195 I think the listening position in this example is quite ideal and is better than the first CD version Adri point to, which to me is more harsh and bright, and not really in the ideal position (too far away) to my ears. I've heard much better CD's I guess and will often opt to purchase, download or what have you based just on this. It might be the greatest organ in the world, but if I don't like the recording position I know it will grate on me and I will not enjoy it, so I don't buy it, same thing goes for sample sets, I'm very picky in that regard.

I guess I would argue in favor of the 6 channel sets, BUT it really depends on how you adjust things. Although these are intended to be wet sets (otherwise the diffuse and surround portions of the set would not be included), one has to be very careful how much of the direct signal is introduced into the mix, as just even a hair too much to me totally ruins the sound. Whereas, a very slight amount of the direct signal adds a benefit in my opinion, especially clarity, but everyone is different. I have the Dingelstaedt set by SP and have both the diffuse and surround signals dialed up to around 92 - 95, anything less or allowing more of the direct signal into the sound, the sound becomes compromised, more unrealistic and synthetic or "robot" sounding as Adri puts it.

Wind modeling and comparing what we get in Hauptwerk to the real thing? I remember back in the day when it was not even available to us here in the U.S. I'm no expert on it I'm afraid, but I do know it can be adjusted in HW quite extensively. I've tinkered with it to some degree, I can hear the results, but I don't know that I can say it has made any huge make or break difference for me one way or the other. Just me, but I think the recording technique / position and the outcome / success of that far outweighs wind modeling.

Marc
Offline

adrianw

Member

  • Posts: 160
  • Joined: Sun Dec 18, 2011 5:10 am
  • Location: Lincolnshire, UK

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 4:29 pm

I hope Jiri Zurek will forgive me quoting a confessional snippet from his site re the SP Rotterdam Marcussen which I have just recalled and seem apposite to the discussion.

The wind in the real organ is rock-stable. This can be modelled perfectly by disabling the Wind Model in Hauptwerk. Hence, one can switch the wind model off for this organ safely. The fact, that modelling the stable wind is so extremely easy in Hauptwerk, I took the liberty to make the wind system freely according to my own liking. There are 11 wedge bellows modelled and the wind is really lively. If you do not like this feature, just turn the wind model off and the sample set will behave as the organ in reality. There is one more reason for setting the wind model off with this big sample set: the CPU usage. I strongly recommend to switch the wind model off on older CPUs (those manufactured prior to 2011), since the wind model eats a lot of CPU power which can have an influence on other parts of the sample set (tremulant model etc.).


(My italics).

I believe Jiri's preference for "really lively" wind over reality is shared by some other sample set producers (as well as Adri, it seems) who perhaps take similar liberties, but without telling us...

- Adrian.
Offline
User avatar

dkoschinski

Member

  • Posts: 89
  • Joined: Tue Jun 07, 2011 11:55 am
  • Location: Netherlands

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostMon Sep 18, 2017 5:48 pm

OrganoPleno wrote:Those CD excerpts sound pretty good, at least on my cheap Desktop PC speakers. What we'd have to know is, WHAT did the SOUND ENGINEERS do to modify and distort the recorded sound, in order to produce this commercially-desirable result?


As a sampleset producer I think the cause for this is quite simple: for CD productions the recordings are modified by EQ, levelling, mastering and such afterwards to obtain a overall better sound. But in most cases no noise/ rumble filtering on the recorded tracks will take place (most CD's I've heard had no adjustments to it).
For Hauptwerk: we put every single sample of the raw recording through a noise filtering/shaping process to get rid of unwanted noise. If you play all those processed samples together, you will obtain an overall different sound as for a CD production for sure. The noise filtering process is quite sensitive. With the slightest deviation you can easily filter out the pipe-specific properties, including the wind. So these processes differ quite a bit and that affects the sound.
Offline
User avatar

giwro

Member

  • Posts: 857
  • Joined: Tue Aug 12, 2003 11:22 am
  • Location: Twin Cities, Minnesota USA

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostTue Sep 19, 2017 10:33 am

Speaking as an American organist, I have to respectfully disagree with Adri. Shaky wind drives me nuts.... heck, even flexible wind bugs me. Most of the pipe organs I have played have rock-steady wind - I've never experienced the vagaries of insufficient wind - even on full organ with everything coupled to everything and all the wind-hungry stops pulled. So, flexible winding is not a selling point for me :lol:

When I set up wind models for the sets I produce, I turn it way down.... because that is the way they sound in real life.

Now.

if Adri likes shaky/flexible wind, and thinks it needs to be better, it doesn't mean he is wrong. To accurately model an historic instrument, it should be enabled. I actually think HW's wind model is very accurate - the problem is that to get it right, one must spend a lot if time carefully measuring, testing, and configuring settings. Most folks don't have the knowledge nor desire to to do that.

I probably will never sample an organ with flexible wind, but if I do, I will want to spend time configuring carefully so that it models reality as closely as possible. That would mean I would need to spend considerable time playing the real instrument, so I can confirm my settings in the virtual instrument.

Cheers
Jonathan Orwig
Coon Rapids, Minnesota USA
http://www.evensongmusic.net
Offline

adri

Member

  • Posts: 1545
  • Joined: Tue Aug 31, 2004 6:33 am
  • Location: Belgrade, Serbia

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostTue Sep 19, 2017 11:30 am

Yes, I do like flexible wind; just as I do like human speech, which is also uneven, non robotic.

But no, I don't like an organ that is truly out of breath and suffers wind lag at every turn. Too much is too much, period. A human who is of breath cannot speak well either.

But yes, I do like the little subtleties of the slightest unevenness, and variations in speech, caused by flexibility in the winding. It is indeed a question of proper adjustment. It is ever human sounding, natural, beautiful and desirable, in my opinion (and in this, I'm certainly not alone, as we all know).

Don't forget that older organs generally had lower wind pressure, and smaller bellows, while the larger magazine reservoir bellows of the 19th century and onwards allowed for higher pressures and thus also, in many instruments, the disappearance of these subtle interferences.

Smaller village organs with smaller magazine bellows still suffer wind lag if abused.

To each his own, yes, but in historical organs from before Cavaille-Coll (more or less) and pneumatic action instruments, slight flexing in the sound is the norm, and such organs will simply sound, well......historic.
Offline

josq

Member

  • Posts: 912
  • Joined: Tue Jan 24, 2012 12:11 pm

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostTue Sep 19, 2017 11:47 am

Hi Adri, just curious: to what extend did the old organbuilders design the effect of the wind supply?
Offline

adri

Member

  • Posts: 1545
  • Joined: Tue Aug 31, 2004 6:33 am
  • Location: Belgrade, Serbia

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostTue Sep 19, 2017 4:35 pm

josq wrote:Hi Adri, just curious: to what extend did the old organbuilders design the effect of the wind supply?


Not sure what you mean by that question. It's ambiguous to me. Are you being ironic? A teasing question? Even a trick question, may be? :-)

Many organ builders today apply again wedge shaped bellows to historically inspired new organs. And are reconstructed, where possible, in restorations when they were lost or replaced by modern types.

To stand closer to the music of the past, I believe very much in original instrumentation, or modern replicas thereof; I am an absolute proponent of period music on period instruments. In organs, that means original type wind supplies.

When I heard the Messiah by Handel played by the early music ensemble under Christopher Hogwood, it opened new eyes in my heart: absolutely beautiful as I had grown tired of massive choirs and orchestras playing that piece. It just sounded boringly bad to my ears. You get my drift.

As I said above, there was great protest against Fans Caspar Schnitger's rebuild of Alkmaar; and this included badmouthing then then still modern slider chests in favor of spring chests because of the way the wind would pass through the wind chest. They really felt the spring chests were absolutely superior in terms of wind behavior.

In closing, another video to make my point: https://www.youtube.com/watch?v=4_pDm1sTTys
Can any HW sample set have a wind behavior emulation like this organ has au naturel?

Another video of this stunning instrument: https://www.youtube.com/watch?v=CIy3XAgglt0
Offline

josq

Member

  • Posts: 912
  • Joined: Tue Jan 24, 2012 12:11 pm

Re: OPEN LETTER TO SAMPLE SET PRODUCERS

PostTue Sep 19, 2017 5:25 pm

Hi Adri, I am genuinely interested, so thanks for your answer!

A few years ago, I was at a concert at the Hagerbeer organ in Leiden. This large organ was hand (foot)-pumped during the concert. This really seemed to have a subtle but beautiful "breathing" effect on the sound. I need no convincing... but I seek to increase my understanding.
PreviousNext

Return to Hauptwerk instruments

Who is online

Users browsing this forum: No registered users and 6 guests

cron