Page 1 of 2

Dry or wet sample sets

PostPosted: Wed May 30, 2007 9:36 am
by Johannes Sørensen
Dry or wet sample sets

When I some months ago got so interested in Houptwerk as the most advanced virtual organ program, that I decided to buy it, it was a surprise to discover that the most classical sample sets are wet (some of the greatest are wringing wet) and only a few are dry.

Both the use of wet and of dry sample sets can have advantages and disadvantages, and people can have different opinions.
To discuss improvement of somthing already eminent you have to focus details.
The primare purpose of this writing is not to criticise the many great classical both barouque and romantic wet sample sets, but
- in few words to focus some of the problems in the issue to use wet contra dry sample sets in a virtual organ, and
- in view of the unbalance between the many wet classical sample sets and only very few dry sets, to advocate the value of the dry (or moderately wet) sample set.

Why are the most sample sets wet?
The acoustics is said to be the organs most important "pipe".
Indeed the acoustics of the church is an important factor in creating the best possible realism in a digital reproduction of a pipe organ, but still only one factor out of more, and wee can not optimise them all at the same time. Either wee have to optimise some at the expense of others or to make one or another compromise.



The assumption behind the use of wet samples.

Although a simplification, to make a virtual organ from wet samples is implicit and in principle based on the assumption, that the sum of a multiply of virtual pipes recorded inclusive acoustics will give the samme resultant sound as if you record the same multiply of pipes played in the church (or other location).

But, is this assumption true? Is that the case?
The degree in which it is depends on the circumstances.
If the room is not too wet for example a vellage church, and the organ is played without celesta and tremolo, it is a rather good approximation,
but if the room is very wet for example a cathedral, and especially if you play with tremolo, it is far from the case.

The more wet the more serious the problems are - of course.
Unfortunately the greatest organs are often placed in cathedrals and the sample sets are recorded with the very wet acoustics to replicate the ambience.



The kind of sound we want of a virtual organ.

Some times wee here in forum can read praising words like: "sounds like playing a CD recorded of the real organ", and it is indeed impressive that (a recorcing of) a virtual organ can be difficult to distinguish from a recording of the real organ.
But the sound wee want in our room when playing a virtual organ, is it a sound like playing a CD?
Not exactly in my opinion.
In my opinion the goal for the virtual organ is to aim at a sound like if the real pipe were presant in my living room although it isn't, to obtain the best possible illusion.
The question for me is, how do wee best approach this goal? By dry or wet samples?



The attack, the steady tone and the release.

Some problems of the release when playing wet sample sets, and reduction of the problems by multiple releases, have earllier been discussed in forum and shall not be repeated here.
However also the steady tone and the attack are affected by the reverberation in the wet samples.
The attack is an important component of an organ stop, but with heavy reverberation in the sample set the attack of the tone is masked and blurred.
Celesta and tremolo are masked and blurred as well, and in addition the tremolo sounds unnatural when wet samples are trembled.

If the reverberation is exaggerated in a sample set, one can when playing have a feeling that best can be described as a rubber like feeling and a feeling of playing a ghost organ, an organ without body.
Rather moderate the reverberation than exaggerate it, rather too little than too much.



The direction of the native sound and the reverberation and sound complexity.

Now vee come to one af the most important points of the subject. The possibility to distribute the direct organ sound and the reverberation trough different channels and speakers - or not.

Our hearing has an ability to some extent to identify the direction of the sound, which besides of the sounds content is a part of the total sound image. Different directions of sounds besides of other parameters help us to distinguish sounds from each other.
When wee hear an organ in the church the direct sound from the organ is heard in a relatively narrow spectrum of directions, but the reverberation has all possible directions as reflections from walls, floor and vaults etc.

If the dry and the wet components of the sound are mixed together and heard through the same speakers they have the same direction of course, which is quite otherwise than in the church and gives another impression.
When the reverberation are conducted through different channels and speakers than the dry organ sound, and if the reverb spearkers are placed in the room with the purpose in view, wee can to some extent imitate the acoustic situation in the church.

In addition when the dry and wet component are separatet through different channels, the organ is heard not only more realistic but with much more clarity and transparency, and it can for example be possible to hear the attack of a new tone without it nearly drowns although playing with heavy reverberation.

When the direct sound from the organ is mixed with the reverberation as is the case with wet samples, the result is a much more complex signal than the native organ sound.
The more complex the signal is the greater demand on the quality of the loudspeakers, or in other words with the same speakers the greater complexity the greater distortion.

To day you can buy fairly good digital reverberation units with several adjustable parameters from for example Lexicon at a resonable price.
Although the reverberation is not a precise image of the ambience in the church in which the sample set is recorded, the result can be just fine.


When wee deal with electronic organs analogous or digital/virtual wee have to remember, that the electronics produce no sound by itself, only electric signals, and wee have no sound until the speakers (or phones), the transducers, transform the signals to sound.
The loudspeaker system, inclusive not only the type and qualities of the speakers but also the locations and directions of the speakers in the room, is an important factor in relation to achieve the best realism of a virtual or other electronic organ.
In view of how important it is, it is surprising how little loudspeaker systems are discussed in forum, but it is out of the theme of this writing.



I hope that some of the makers of sample sets in future will provide us with high quality dry sample sets of great classical organs as they do now with wet samples.
In cases where it is not practical possible to make close mic recordings of the pipes, I hope that they besides of the wet sample sets also will make semi dry sample sets with reduced reverberation, although the sets then do not include a replica of the ambience in the church.
A practical solution don't need to be either black og white.



These were some thoughts of a simple amateur.
I hope that some other both users and creaters of sample sets will contribute with their thoughts and experiences.


Thanks for reading so far!

Regards

Johannes

PostPosted: Wed May 30, 2007 10:44 am
by mdyde
Hello Johannes,

Although a simplification, to make a virtual organ from wet samples is implicit and in principle based on the assumption, that the sum of a multiply of virtual pipes recorded inclusive acoustics will give the samme resultant sound as if you record the same multiply of pipes played in the church (or other location).

But, is this assumption true? Is that the case?
The degree in which it is depends on the circumstances.
If the room is not too wet for example a vellage church, and the organ is played without celesta and tremolo, it is a rather good approximation,
but if the room is very wet for example a cathedral, and especially if you play with tremolo, it is far from the case.


I agree that wet samples cannot give as good an approximation for tremulants (or other significant fluctuation in the sound of a given pipe whilst sounding, e.g. opening the swell box suddenly whilst holding a key), which is why I would definitely recommend using dry samples for a theatre organ, but I don't agree that summing wet samples cannot give a an extremely close result to hearing the original organ in the acoustic, no matter how wet the samples.

The key point is that a room is a linear, time-invariant (LTI) system in signal processing terms, i.e. it *is* valid to sum the signal after the room acoustic. A pipe organ similarly has almost no non-linearities, which is a major reason that the sound is perceived as being very clear (the other being the large number of sound sources, each with their own separate acoustic, i.e. distinct, recognisable patterns of early reflections for each pipe separately).

I would say that the significant shortcomings of wet samples are caused by using only one attack attack or only one release sample, giving too little or too much perceived reverb when playing fast. However, multiple attack and release samples address that very well indeed, as I think the Vollenhove sample set from OrganART Media demonstrates.

Now vee come to one af the most important points of the subject. The possibility to distribute the direct organ sound and the reverberation trough different channels and speakers - or not.


My opinion is that multichannel audio is pretty much essential for dry sample sets, mainly because the brain is able to identify and separate the pipe sounds primarily by their early reflections. Since dry sample sets have no acoustic, a dry sample set sounds muddy if played through only a single stereo pair.

However, for wet sample sets, I absolutely believe it is best to listen through high-quality headphones so that the original acoustic is preserved, i.e. all early reflections heard as nearly exactly correctly as possible, and there is no interference from the acoustic of the listening room. In my opinion that gives a far more realistic room impression than is possible with any number of speaker channels, and I think the gain in realism outweighs the disadvantages of less realistic tremulants for a classical organ.

If headphones aren't used, then a single stereo pair of studio monitors are the next best option for wet sample sets because they are designed to be heard close-up, i.e. again minimising the interaction with the listening environment.

If you don't have a copy of the OrganART Media Vollenhove sample set, then I'd recommend getting one and playing it using high-quality headphones. I think you would agree that no dry sample set with added reverb or any number of speaker channels could come close to the degree of realism achieved by a top-quality wet sample set with multiple releases and headphones.

However, of course, for a wet listening environment (e.g. a church or large room without studio monitors), a dry sample set is absolutely essential because it almost always sounds very unrealistic indeed to play a recording containing room acoustic within a room with a significant acoustic of its own, i.e. effectively having two room acoustics.

Just my opinions!

Best regards,
Martin.

Dry or wet sample sets

PostPosted: Wed May 30, 2007 12:21 pm
by jocr
mdyde wrote:The key point is that a room is a linear, time-invariant (LTI) system in signal processing terms, i.e. it *is* valid to sum the signal after the room acoustic. A pipe organ similarly has almost no non-linearities, which is a major reason that the sound is perceived as being very clear (the other being the large number of sound sources, each with their own separate acoustic, i.e. distinct, recognisable patterns of early reflections for each pipe separately).


Does this hold true for very resonant cruciform buildings? In real life the transept [the part of a cruciform church that crosses at right angles to the greatest length between the nave and the apse or choir; also : either of the projecting ends of a transept] seemed to garble the organ sound from front chambers designed to hold a higher pressure organ. We solved the problem by reinforcing the balcony and building the new organ in the back.

There are numerous other less than desirable situations I've lived with, the worst being a two level installation in a deep chamber on one side of the chancel with horizontal shutters on the lower division that could only reflect the sound downwards. Of course, these problems may be limited to Southern California.

James Pressler

PostPosted: Wed May 30, 2007 12:31 pm
by mdyde
Does this hold true for very resonant cruciform buildings? In real life the transept [the part of a cruciform church that crosses at right angles to the greatest length between the nave and the apse or choir; also : either of the projecting ends of a transept] seemed to garble the organ sound from front chambers designed to hold a higher pressure organ.


Absolutely. Room acoustics are linear. A different shape room might *seem*more or less muddy than another, but that's just because the pattern of early reflections is different. The brain will have more difficulty picking out the components of a signal if it has been bounced off a lot of walls because there will be a lot more reflected signals that it needs to consider, i.e. more brain processing power required to decide which reflection is coming from which sound source!

Best regards,
Martin.

PostPosted: Wed May 30, 2007 7:54 pm
by Stephen Phillips
Hello all,

A good subject to revisit. As an enthusiastic amateur, I understand there is yet a third option (not yet fully implementable in Hauptwerk):

1) Record samples relatively close to organ case (too dry for 'CD' sound)

2) Apply an appropriate IR convolution process to the audio signal (ideally from that same venue)

With adequately 'present' samples, this works very well, with the possible addition of a release-tail truncation as currently provided in HW2 providing further flexibility.

The observation that 'trapped' reverberation emanating from the same channel/direction as the 'dry' signal results in a loss of clarity is scientifically correct. Martin's advocacy of headphone listening/playback is very good advice, and is easy to demonstrate empirically. Rooms of a domestic scale are highly destructive to sonic fidelity, even those with rather elaborate acoustical treatment.

The suggestion to bring the reverberant signal around and away from the direct sound, for loudspeaker playback, will dramatically increase clarity and spaciousness, but will require some sort of decoder/multichannel reverb to achieve this. The 2-channel expansion algorithms provided on the Lexicon processors, for instance, provide great results and are highly programmable. Similarly, there are some flexible software IR applications offering similar facilities.

Cheers,

Stephen

PostPosted: Thu May 31, 2007 1:13 am
by Stephen Phillips
Forgot to add:

Headphone playback, while providing for dramatically increased resolution and spatiality, cannot usually (ie.unassisted) accurately recreate the soundfield of the original recording venue. This is certainly true of loudspeaker playback as well, with the solution addressed in the various inter-aural crosstalk (IAC) - cancellation/HRTF processors designed for this purpose. While IAC is one problem headphones are inherently free from, they present a variety of special challenges of their own.

For now, the best 'workaround' for this appears to be the Smyth Research Virtualiser, an industrially-heavyweight concept from Steven Smyth (one of the creators of the DTS multi-channel audio algorithms). The design involves analysing and mapping the soundfield of a loudspeaker/room combination, together with the individual frequency response of the user's ear canals and chosen headphones, to derive a transfer function which allows astoundingly successful three-dimensional recreation of the original soundfield. The system detects and compensates for head movement, completing the illusion of an externalised soundfield. Critical listeners report amazement and delight at the integrity of the illusion, with even a sense of the 'visceral impact' sensation of bass much loved by organ fans and usually absent in headphone reproduction. While I've yet to experience the system myself, it's definitely on a short-list of future projects.

Speaking of speaker quality, I ought to mention the leading-edge work of my Australian compatriots from DEQX, a small Sydney-based hi-tech company specialising in minimum-phase processing/crossover hardware which is making an increasing impact internationally, particularly in control rooms (eg. Abbey Rd) where absolute quality is critical. Again, the system relies on individual analysis/calibration, with a resultant signal flat in the time and frequency domains. This I have access to, and will report on forthcoming experiments on my Klipschorn corner horn speakers.

All the best to my fellow enthusiasts,

Stephen.

PostPosted: Thu May 31, 2007 5:58 am
by mdyde
Dear all,

I would like to apologise for my over-enthusiastic comments in support of wet sample sets, with which I inadvertently caused offence to a dry sample set producer. I want to stress that it was absolutely not my intention to denigrate dry sample sets in any way.

I had really just wanted to make the point that a good wet sample set with headphones can reproduce the spatial characteristics of a room in a way that dry sample sets cannot (at least until/if a real-time impulse response per pipe is possible). However, there are of course many common situations in which a dry sample set is indeed highly preferable or essential, e.g.:

- For playing/listening in a reverberant listening environment. Not everybody wants to wear headphones all the time!

- For playing to an audience of more than one (unless headphones or close-up studio monitors and a sound-proofed room were provided for each member of the audience individually!).

- For hearing the close-to-the-pipework sound, which an organist would usually experience when playing a pipe organ, rather than hearing the (much less distinct) sound from the audience's perspective. Hence a dry sample set can potentially provide a more realistic practice environment.

- Similarly, without reverb, a dry sample set shows errors in playing technique more easily, and is thus very good for practising for that reason too.

- Tremulants in particular are much more successful with dry sample sets. The more vigorous the tremulant, or the wetter the sample set, the less effective a tremulant will sound in a wet sample set. For organs where the tremulant is important or strong, a dry sample set will give the best result. Since the success of a theatre organ in particular hinges upon the success of the tremulant, dry sample sets are pretty much essential for theatre organs. However, classical organ tremulants can be important too!

- Other rapid changes/fluctuations in pipe sound (e.g. opening a swell box rapidly or wind pressure fluctuations) are also less effectively modelled with wet sample sets for the same reasons.

- Similarly the effect of tremulants and other non-static effects cannot be modelled very well during the release (reverb) tail with wet sample sets.

- Dry sample sets use much less memory and require much less computer processing power than wet ones.

In summary, there are plenty of very good arguments for dry sample sets, as there are for wet. The best type to choose probably depends mostly on what you intend to use it for. A dry sample set with lots of audio channels (the more the better) and a listening room with good acoustics can without doubt give a more realistic organ model throughout the room than a wet sample set. Wet samples only really work best when heard through headphones or with close-up studio monitors in a dry listening room.

I hope that presents a fairer overview of arguments for and against each type. Once again, it was emphatically not my intention to offend.

Best regards,
Martin.

PostPosted: Thu May 31, 2007 7:07 pm
by Ken
The different needs and preferences of Hauptwerk users justify different designs of sample sets. Martin expressed some of the needs that would steer some users to use “dryish” sample sets (thank you Martin). In addition, here are a few others:

User Control:
With an external reverb effects processor, the amount and type of reverberation can be selected by the user to match personal preferences, the musical genre, and/or the playback environment. Playback in a (dry) home environment of dry sample sets enhanced with external reverberation is a viable option. Dry samples need not be played back “dry”. Actually some dry sample sets have been designed/edited/optimized specifically for playback in a dry listening environment with external reverberation added.

Dynamic Attach/Release Profiles:
The external reverb effects processing unit automatically generates a multitude of different releases, each one corresponding to the sound profile immediately preceding the release of the specific note.

Combining Ranks from Other Sample Sets:
It becomes easier to match and combine samples from other dry sample sets.

Edit-ability:
Dry samples are inherently more easily edited since they do not include as many transient reflective elements.

Using with Other Instruments:
When playing/recording the sample set with other instruments, it becomes easier for the other instruments to share a common acoustic environment with the organ (simply apply the same reverb to all the instruments).

Tremulant and Swell Box Effects in Reverberant Space:
Applying swell enclosure and tremulant effects in a reverberant setting can be effectively modeled by applying these effects to the dry samples and then applying external reverb.


The great thing is that Hauptwerk accommodates all the different sample set designs that can in turn address the different users’ needs and preferences. Marvelous!

Ken



PostPosted: Sat Jun 02, 2007 8:29 am
by mdyde
In the long-term, using dry sample samples with per-pipe impulse response convolution (reverb), with a sample set specifically voiced for the impulse responses (virtual acoustic), should give, as nearly as possible, the best of both worlds. This recent post has some interesting insights by Csaba Huszty into the feasibility of recording per-pipe impulses responses:

http://forum.hauptwerk.com/viewtopic.php?t=2295

Synthesizing impulse responses using room modelling software is another approach (also an approximation, of course).

With existing multi-channel software/hardware convolvers or reverb units you can get the benefits of realistic distinct early reflections (with the sound of particular pipes/ranks being interpreted by the brain as emanating from different spatial locations), whilst still having the benefits of dry sample sets discussed previously.

For example, this Milan Digital Audio Virginia WurliTzer demo were recorded in that way:

http://www.crumhorn-labs.com/Hauptwerk-AudioDemos.shtml

... using dry samples with a set of synthesized impulse responses. Hauptwerk's multi-channel output capability was used to route different parts of different ranks to the various distinct impulse responses, thus positioning them at different points in the virtual acoustic, and then the result mixed back down to a single stereo output pair. This gives a close approximation to using many amplifiers and speakers in terms of providing separate patterns of early reflections for different pipes/ranks to give a realistic spatial impression, especially for home/close-up/headphone listening, but without the need for a large array of physical amplifiers and speakers and with the benefits of dry sample sets, such as the interaction of the tremulants and swell boxes with the (in this case virtual) acoustic.

Multi-channel convolution is most easily achieved at present using a VST host program (e.g. V-STack) so that Hauptwerk's multi-channel output can be fed to several instances of an impulse response convolver VST plug-in and then mixed back down to drive monitor speakers using the virtual mixing desk of the VST host. However, when Hauptwerk has its own native convolver such a configuration will be simpler.

Multi-channel dry output fed to multiple or multi-channel reverb processors can give a similar effect in terms of spatial impression resulting from discrete early reflection patterns.

Until the day that real-time per-pipe convolution is possible in terms of processing power, both wet and dry approaches have their own significant benefits and approximations and shortcomings (as they will even then, but in different ways). However, a dry sample set with convolution/reverb can provide an extremely good model of a room acoustic (as with speakers, the more convolution/reverb channels, the more effective the spatial impression) for home listening.

Ken pointed out to me that the Exemplum Organum sample set is specifically voiced and designed for use with current external reverb processing units.

Best regards,
Martin.

Drying out the St. Anne's sample set for a church.

PostPosted: Tue Jun 05, 2007 12:42 am
by tomg3usa
Is it possible to obtain a 'dry(er)' sample set for the St. Anne's Brindley to use in a nave? Or is this a DIY process?

Regards,

Tom

PostPosted: Tue Jun 05, 2007 3:56 am
by mdyde
Hello Tom,

Is it possible to obtain a 'dry(er)' sample set for the St. Anne's Brindley to use in a nave? Or is this a DIY process?


I'm afraid there are no dry recordings of St. Anne's, and no plans to re-record it, since there are plenty of excellent sample sets from third-parties, of both dry and wet types.

Sorry!

Hauptwerk does have an option on 'Organ | Load organ, adjusting rank audio output routing' to truncate release samples, giving a simple model of dry samples, although it cannot remove the room acoustic from the attack/sustain portion of the samples.

Best regards,
Martin.

PostPosted: Tue Jun 05, 2007 7:35 am
by dhm
mdyde wrote:Hauptwerk does have an option on 'Organ | Load organ, adjusting rank audio output routing' to truncate release samples, giving a simple model of dry samples, although it cannot remove the room acoustic from the attack/sustain portion of the samples.

Best regards,
Martin.


Martin,

If we do this, is it possible to save both a "dry" and a "wet" version of the same organ, for use in different buildings?

Best wishes,
Douglas.

PostPosted: Tue Jun 05, 2007 8:03 am
by mdyde
Hello Douglas,

There is not currently a means to save multiple sets of rank routing settings (although that is planned for the future), so you need to adjust the settings each time you want to switch between wet and truncated.

Best regards,
Martin.

PostPosted: Sat Jun 09, 2007 3:44 am
by Aristide
dhm wrote:Martin,

If we do this, is it possible to save both a "dry" and a "wet" version of the same organ, for use in different buildings?

Best wishes,
Douglas.


To have the same organ with 2 different load options (dry and wet) I do this:
I run Hauptwerk-standAlone to load wet organs and Hauptwerk-MidiSequencer (which I only use as the StandAlone to play with my phisical organ..) to load the same organs without releases (dry)..
It works becuse HW saves the configuration files in 2 different directories.

However, I think we really need a quality symphonic/eclectic organ (3 manuals) DRY sample set in HW2. (to play with an audience in all the situations we cannot use a real pipe organ)
I would pay much for such an organ.. (as someone said in some past posts about a dry Cavaillé-Coll)

Near all the sample sets are WET.. we have wonderful instruments, sure, but they are suited "only" to play in my room or to record CD.
I would like a sample set which I could use for LIVE performance too, as a musical instrument..

Best regards
Lorenzo

Matching HW sets to real pipes

PostPosted: Mon Jun 11, 2007 10:01 pm
by swendello
This brings up a related question, for which I would appreciate any opinions or advice from Martin or others who could comment.
We are planning a HW installation for a moderate sized church, seats about 580, which has an existing 2 manual 25 rank pipe organ. The HW would be installed as a CODM using both HW and the real pipes.
The plan is to have HW add a few missing stops on the great and swell organs, and add some substantial pedal bass. There would also be a new completely digital 3rd manual, and a new 3 manual console.
The room has a reverb time of about 1.5 sec empty and less than 1 second when full.
Should the HW installation use dry or wet sets? Should we strive for a sizeable number of audio channels to bring the sound as close as possible to the real pipes?
What about using dry sets and adding reverb to one or more of the channels?
I have an experimental HW setup in my living room, with capacity for 4 stereo channels. Currently I am using the Lexicon200 for external reverb, but it is difficult to extrapolate from this to the eventual church setup.
-One of the issues will be voicing the HW to be compatible with the scale of the existing organ.
Are there other problems to be aware of. Finally, how likely is this kind of endeavor to be successful in creating a larger 3 manual organ?
Thanks in advance for your responses.

Wendell