?Zero latency? (Page 2 of 3)

Glenn,
I've tested with your torture-test file, and I can easily hear a difference when it is played across the cable. Will upload the recordings later - gotta run.

When it's rendered directly, from the file, it is extremely crisp. Over the cable, it doesn't sound as crisp, due to the ever so slight arpeggiation caused by transmission delay. This is what I was interested in hearing. Thanks for preparing this test.

It is conceivable that some instruments/software might try to hide this arpeggiation, by slightly delaying processing of the notes, if it sees that there are MIDI events pouring through. (complete surmisal on my part)
This would cause other timing problems, though, of course.

Greg.

Ahem... http://www.youtube.com/watch?v=cVtiTyMlqAU

Not to mention the myriad pro recordings.

(BTW, the "latency" you witness in that video has nothing to do with actual latency. It's just Youtube's cockamamie misalignment of the audio and video.)

I got issues with pretty much every digital piano and software piano out there. There's nothing quite like the real thing. My old teacher and I agree that digital pianos are instruments all their own. I've had ideas on making it more realistic, but no one's bitten yet.

But this doesn't mean you can't perform in real time. I don't notice any latency when simply playing/monitoring. I do notice a bit of latency when recording in Pro Tools, but that's not Pianoteq's problem.

Agreed.  If one were playing a layered sound (multiple channels simultaneously) the MIDI timing might be a bit more obvious I guess, but that's quite remote from the issue at hand.

Note that this test I did may not be fully representative of the capabilities of the MIDI protocol, due to possible timing delays on my system. (drivers, OS scheduling, Pianoteq itself, etc)

Greg.

Correct me if I'm wrong, they're glorified player pianos. They remember your playing and then reproduce it, using the actual strings to make sound. But if there's a sound module, I'll bet you a ham sandwich it's triggered by plain vanilla MIDI velocity. And I'm not sure exactly what the thing "remembers" of your playing. Is it 100% translated, or is it transformed into simple velocity?

So for direct recording and patching to a mixer, it's about as expressive as most cheap digital pianos. I'd imagine the new CPs are more expressive, given their modeling technology.

Hello Glenn,

I do have first hand experience with Yamaha Disklaviers, and a good friend of mine shared with me how they work:

First off, the correct spelling is Disklavier (no c's and only one k).  Yamaha Disklaviers are in their fourth generation, although as of this writing, I do not know whether Yamaha is going to introduce anything new to Disklaviers at the 2010 NAMM show later this month.  The first generation was basically your run of the mill player piano of the late 1980's / early 1990's, using electronic relays and solenoids instead of air pressure to record and move the keys as in Ampicos and DuoArts of yesteryear.

I am not sure what was different about the second generation, but the third generation had a sound module attached to it, along with stereo amplifiers and speakers.

The fourth generation Disklavier (the latest as of early January 2010) has a dedicated computer strapped to its underbelly, which contains an XX gigabyte hard drive, sampled sounds running into the hundreds, and the ability to attach a microphone and the instrument's microprocessor will harmonize your voice in real time as you sing (assuming, of course, that you press the correct chords on the keyboard)! 

One can access information about the status of the Disklavier via a handheld device or an optional 10" touch screen (which is standard in Disklavers over 6' --- I believe those are the DC3's and larger, all the way up to a 9' long Disklavier CFIII).

From what I understand, it is the Disklavier's Mark IV incarnation that allows one to use SMPTE timing code off a video camera to synchronize the video with your playing.  Apparently, if you hook up the Mark IV's video output to a television, and hit "Play" on the video camera you have just recorded, the Mark IV will synchronize your piano performance with the video you have just recorded.  Of course, the Disklavier and the video camera have to be connected via the camera's Audio Out (which also outputs the SMPTE sound striping code).

The above is why the Disklavier Mark IV is used for the eCompetition in Minnesota -- contestants from all over the world are able to video record themselves in synchrony to the Disklavier, and send those "auditions" in the form of midi files and videotapes to Minnesota for the preliminary rounds of the contest.  The judges simply gather in front of a television, video reproducing equipment and an attached Disklavier, and "watch" the contestants perform.  I also believe the final rounds may be "experienced live" at selected Yamaha dealerships around the world via satellite transmission as the contestants play through their Finals performances.

Now as to how MIDI is captured while playing, I can tell you from firsthand experience that you would never be able to tell from the piano's action that you were playing a player type piano.  You see, the Disklavier uses fiber optics to watch a small sheet of clear plastic (approximately 1/4" x 3/4" if my memory serves me right) whose "gray scale" varies from completely clear (representing zero note-on velocity) to pure opaque black (representing a value of 127 note-on velocity).  These feather-light slips of plastic are attached to the hammer shanks at a place just prior to (and after) impact with the piano's string.  In other words, the recording system is not hampered by any extra mechanical linkages that would hinder the action's touch.

Interestingly enough, if one were to begin recording -- and then hold a cluster of notes "half-way down" but not enough to strike a string, the Disklavier would reproduce that very same half-way down note position upon playback.  I saw that with my own eyes!

Did I mention that the current generation of Disklaviers can be hooked up to WIFI, and worldwide satellite communications can beam something like 20 different subscription channels to the Disklavier.  For a monthly fee, one's Disklavier Mark IV will play, wirelessly, any type of musical genre you desire -- and not be stuck with listening to the same old files over and over again,

As usual, my threads seem to be overly long.  If you have any more specific questions, please ask and I shall try to find an answer for you from a Certified Yamaha-Trained Disklavier technician.

Before signing off, I would like to state that the Yamaha Disklavier system is NOT a retrofit of a system into an existing production piano, from Yamaha or otherwise.  When a Disklavier is going to be made, the piano that is being built ... is constructed in mind with extra heavy structural components, especially in the vicinity of the keybed, so as to accommodate the solenoid system that pushes "up" on the back sides of the keys (opposite to the front of the key where one's fingers usually press downward).  In addition, the whole piano is beefed up to allow for some sustained loud piano playing afforded by reproducing very high velocity levels.

Cheers,

Joe

Hi Glenn,

The Boesendorfer recording and playback system is their proprietary system that has far more than 128 individual velocity levels.  That's all I know about the latter system.

Perhaps Yamaha purchased Boesendorfer because Samick has purchased Bechstein and are looking at a joint venture in China.

http://en.wikipedia.org/wiki/Samick

Joe

A Yamaha grand whose model number starts with a G is a 4'11" GB1.

But then, ... who asked me?

Joe

Oops, I forgot and spoke out of turn by recalling from memory instead of consulting a good reference.

GB1 = 4'11", GC1" = 5'3, but with the action of lower price than the C1, 5'3" Conservatory model.

I stand corrected; sorry.

Joe

The MP3 of the acoustic grand was dreadful - I didn't realise that it had lost so much quality. I've re-encoded it at a higher bitrate:

http://www.forum-pianoteq.com/uploads.p...atency.mp3

Greg.

What is too much advertising for Yamaha?  They sponsor the Minnesota e-Piano Competition, and provide all the pianos.  "He who pays the piper, calls the tune" as they say.

What I find interesting and unusual is that the performances are recorded, but they are in XP SMFs (standard midi files), so if I'm reading it correctly, the Yamaha doesn't seem to record to a proprietary file type.

http://www.piano-e-competition.com/ecom...ctions.asp

If you're interested in classical piano, you might be interested in checking them out; I have several dozen on my HDD.

Glenn

Which tells me that it doesn't record every nuance of a performance. Besides, unless you play the MIDI file back on a Disklavier, you won't get the performance as originally conceived. (Yes, I played one of those files in Pianoteq, and you could tell that you'd need to futz around with the instrument settings to get a fair rendition.)

I'm sure it's fine for a Yamaha-sponsored competition. In the meantime, modeling technology will get cheaper and be a far more reasonable solution for non-conservatory artists.

It's impossible to achieve zero latency. Even pianos have latency. Electric guitars as well. Mechanical organs even more so. That's all I'll say about this topic.

I played one of the MIDI files back in my Garritan Steinway and can assure you that it does record continuous damper pedal data (half-pedaling).  But what WOULD record EVERY nuance?  Even a microphone is unable to record things that are masked out by other louder sounds.

Take a listen to this.  It's the MIDI file for Lindsay Garritson's performance of Chopin's Nocturne in B Major played thru my Garritan Steinway preset.  Pretty darned identical to the original performance (watch the video of her perfomance!).

Well, isn't "futzing around" the name of the game with ANY virtual piano?   

My point with this post is that the technology is amazing - all of it.  From Pianoteq to the sample-based libraries to the Disklavier.  We here tend to focus on the limitations of certain things and use them against what is quite simply revolutionary methods of making music.  I embrace all of it and look forward to what the next few years will produce - both from the artists that create these tools to the artists that use them.

Curt

Aren't they working on technology to capture every nuance? Again that tells me Disklaviers up to this point have not captured every nuance, and Yamaha (and Bosendorfer) are endeavoring to capture more.

It's a forum on technology development. It's not about using limitations "against" anything. It's about pushing the technology forward.

Hi again,
In the case of the acoustic grand, it is very difficult for me to say what the latency is, because when I listen very carefully, the sound I hear right at the beginning of what LOOKS like the real sound, sounds more like something in the action, more than the string. (it doesn't even sound like the hammer hitting the string, IMHO)  This could be the key bottoming out, or some other sound from the action.  My gut feel is that the real latency is probably more like 30 to 40ms, which would then be dead on the numbers given by Guillaume & Joe.

Greg.

It's a real piano.

Hooray - you're back. ;^)

In all your testing, have you at any stage tested using a native ASIO driver, WITHOUT using the ASIO4ALL driver?
If you're not sure, get assistance from your technical friend.   ASIO4ALL should only be used when you do not have a proper ASIO driver for your audio interface.

EDIT: Sorry, I see that you have tested with ASIO.

HOWEVER - I think I see a potential problem. You have not yet tested with a NON USB soundcard, with a native ASIO driver!  As I reported, my USB interface DOES appear to incur an extra 20ms of latency, when I compare it with my PCI soundcard. I was using native ASIO for both.

I think we are making progress. Please try to test in a configuration that uses a good quality audio interface, that is NOT USB, and that also supports ASIO natively. 

Yes, if you can record your latency using your mobile phone, that would be really helpful. Place the phone right near the key, and play with your fingernail, forte. Move the speakers as close to the phone as you can, but you don't have to go crazy - just report the approximate distance from the phone to the speakers, and we can calculate the delay caused by sound travel through the air.
Turn the volume right up so it's nice and loud.

Greg.

On the other music forum I frequent (past seven years), it's always been said that to minimize latency, native ASIO is necessary.

Native ASIO (not many sound cards seem to have this) is a software driver written specifically for the sound card - by someone that knows what they are doing.  See post number 11 by asreal4/Heinke.

This does not refute what the Dragon says, but native ASIO is an important element that's overlooked - or is something most of us cannot control because our sound cards don't have it.

Glenn

Pianoshrek2 has kindly uploaded a recording:
http://www.forum-pianoteq.com/uploads.php?file=PT.mp3

(rename it to AMR and convert it if you have to - perhaps someone can convert it and re-upload it - I can do that later)

Very quickly inspecting the recording, I saw one at 52ms, and another at 56ms. Not terrible, but, JUST PERHAPS, enough for him to notice!!

I say even more strongly now - try to test without any USB devices at all. Perhaps at a shop that has a good PC setup...............

Once we have established that Pianoteq is CAPABLE of working to your satisfaction on ANY equipment, then we can find a configuration that will hopefully work on your existing computer, without having to resort to buying another computer.

Gotta run........

Greg.

Glenn,

I think you're right about not many (if any) onboard soundcards having native Asio drivers, but there are, as you probably know, several options for pcmcia cards and express slot cards that have Asio drivers for laptops, as well as PCI slot solutions, such as the M-Audio Audiophile 2496, for desktops. 

I've been using an Echo Indigo DJ card for 3-4 years now on various Dell laptops, and it's great. It not only has Asio drivers, but GSIF (Gigastudio) drivers as well, and it's 24 bit 96Khz. I don't push the system on my single core Pentium, but I get 2.3ms latency and the sound quality is excellent.

Michael

And here's a MIDI latency tester:
http://earthvegaconnection.com/evc/prod...l#download

Using only note-on/off events:

FTU (USB)
Message Latency: 7.63ms
Jitter: 7.19ms
Max. Deviation: 11.24ms

MPU401 (integrated on desktop's motherboard)
Latency: 0.38ms
Jitter: 0.01ms
Max. Deviation: 0.48ms

Greg.

Micheal:

On that forum (that has quite a few computer geeks compared to here, and some of them are actually computer guys with a side interest in music), the guys with the M-Audio cards have no problems at all.  They are superb.

Glenn

Greg:

No, I haven't measured total latency because with my ASIO card set to 7.1 ms, I have no issues whatsoever.

The latency thing for me is academic - I'm quite interested in it, but it isn't an issue when I play.  I'm not a great pianist (I'm not even a piano player, my uploads demonstrated that), but I started taking lessons in 1946, and have a pretty good feel for how a piano should react, and mine causes no problems.

If people can sense differences in latency of a few ms, then they will have to learn to adjust - pianists on real grands do - Modartt and Joe Felice demonstrated that quite well.  Twenty to thirty seconds depending on strike force and it varies across the keyboard?  Maybe having played a Yamaha grand for twenty-odd years I just adjusted to it and don't notice it .  I've had DPs since 1990, and adjusted to them to.

Bottom line:  Once one has a decent ASIO card, and has an adequate system, there is going to be latency - if it's a problem, then the user just simply has to learn to adjust to it.  Period.

Yes, that's going to please some, but there is no other solution at present.

Glenn

Greg:

No, in fact I concluded right at the start that Pianoshrek2's system is probably quite inadequate.  My dual core Athlone was purchased five years ago - at the time it's benchmark performance was right at the very top (for whatever that's worth).  It's nowhere near the top today.

I think he just "jerry-rigged" something together, and thought or assumed that Pianoteq was the problem.  If it was, we'd be all complaining about it.  But we as a group aren't - at least I don't hear them.

This afternoon, I visited my piano-rebuilder friend; he's also very much into electronics/music.  We discussed this issue - and without any prompting from me, he specifically said that neither USB2.0 nor Firewire are really going to do the job.  A PCI or PCIe (or Mac equivalent) will be needed.

I get the sense that since so many companies are selling USB and/or Firewire cards (they are so simple to install and are cheap - and are all over the market), that people believe they must be up to the task.  My personal opinion is that this is self-delusion (at least to some extent).  However, if one is not playing live, they may be just fine.

I have a USB external card - anyone want to buy it?

Restating:  those using PCI or PCIe gear don't seem to be having problems.

Glenn