Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Question about USB lag and pedal

I'm planning to buy a digial piano/controller to use with a software synth (like Ivory). The salesman I talked to today said that the USB interface will have too much lag time for that, but using the MIDI port will pose no problem. Is that true?

And for pedals, how does half pedalling works in MIDI? If the keyboard support half-pedal, does it mean it can send 0-127? And for keyboard doesn't support it, can it still send the MIDI signal of different gradation? And I suppose software synth like Ivory can support sustain at different level?

No USB is more than fast enough.

I'm not 100% sure about the peadal question but I believe half pedalling depends on the pedal not the board and I think it can use the full range from 0-127

I bought my DP just for this reason. I think the salesman is trying to sell you some more expensive stuff. Ask him to prove his argument.

I have this USB interface and my lag is 2.7ms with ASIO4ALL drivers meaning that I do not have an expensive professional audio card and still my lag is extremely low. To give you a better idea of my lag - there is no difference in delay between the sound which comes from my headphones attached to my PC attached to my DP with the USB cable compared to the original sound coming out from the DP. I can listen to both sounds at the same time and as I press the keys there is no difference in delay.

There are few things to consider though such as the audio chip of your PC/computer, operating system and the configuration of the account you are planning to use the SW, RAM, HDD. Just check the requirements of your piano software.

I have an Edirol external USB Soundcard with ASIO Support and MIDI Ports.
Please note, that "time lag", or better said "latency time" is generally not a problem of MIDI Interfaces, it is the problem of soundinterfaces.

That said, MIDI Interfaces can cause problems, if they are incompatible with Vista or other OS, or if they have crappy drivers. Check this out before buying.

Ok, the abovementioned USB Soundinterface allows for lower latency times than my other solution: the inbuilt Realtek HD soundinterface with ASIO4ALL driver. The disadvantage is, that the USB Interface has a very high CPU load, while the inbuilt soundinterface has almost no CPU load.
So I get crackles with the external and expensive interface and I get no crackles with the inbuild zero cost interface.
With ASIO4All I get a latency of 6ms and that is absolutely ok for me and so I use that.

I believe that inbuilt interfaces and external Firewireinterfaces generate less CPU load than USB Interfaces, because tthe former are DMA capable and USB is not. Dont kow, if others can confim that...

For synthogy ivory: It does not support proportional pedalling, but there is an upgrade announced for April, that will support it.

Not all keyboards that support half-pedalling will support true proportional pedalling. So far I know, all Kawai models that are older than the CAx3 series only support 8-steps and the models before CAx1 support it in a very proprietary way.

HTH,

Peter

Thanks. Good to know that USB will work just as well. My computer is plenty powerful and I'm willing to buy a decent audio card if required. Pianisti, which software do you use btw?

Thank you Peter. For latency time, do you mean that it usually comes from the time required by the software to write to the sound card for output rather than the time for the computer to take the input MIDI signal?

Do you know of any synth that supports proportional pedalling?

If I end up getting a board that doesn't support proportional pedalling that is good enough to my liking, are there separate pedal modules that I can purchase that I could perhaps plug into the computer? I'm a classically trained pianist so the pedalling does matter to me quite a bit. Although the 8 steps you mentioned in the older Kawais appear adequate (as long as the software synth can support it that is)

Latency time: MIDI is generally fast enough for piano playing.
Even a "slow" serial MIDI Interface can handle several hundred keystrokes per second ;-)

Only if you make big concertal multi-track arrangements you can reach the limit.
Edit: Because MIDI-over-USB has much more data througput it should also be suitable for multi-track arrangements.

Latency problems are caused by soundcards with big buffers. The buffers cause a delay, but they reduce realtime requirements. For realtime playing small buffers are neeeded and the realtime requirements are higher. This problem is solved with the ASIO protocol. (Thats for windows products, others have other solutions or problems)

Many yamaha Clavinovas can use the sustain pedal as a pitchbend controller. So these /should/ also support true proportional pedalling.
Learn about my Kawai experience here: http://www.thebadroom.de/galaxypianos/galaxyboard/viewtopic.php?bid=1&tid=121

Even the Kawai MP8 II supports 8 Steps only. Apart from that it is a very good MIDI controller and piano keyboard.
I dont know if 8 steps is enough nowadays. The newest Galaxy piano versions (http://www.galaxypianos.de) supports emulation of pedal noises with proportional pedals and I dont know if this will work with non-continuous pedals.

As a standalone pedal unit, there is the CME GPP3. This is connected via MIDI and does not require a keyboard with proportional inputs. It is out of production now, but some dealers still have it in stock.

Edit: For a true proportional pedal, it is possible that the pedal produces more MIDI events than the keyboard. I believe, there are not much experiences for this problem nowadays.
This /could/ cause problems with serial MIDI, but MIDI over USB should be able to handle the data amount.


regards,

Peter

zxcjason when Pianisti talks about 3 milliseconds latency, to get a good feel for exactly how fast this is, you should know that the speed of sound is 1.1 feet in 1 millisecond.

So once you get below 5 or 6 milliseconds of latency, the biggest source of lag in your sound system will be merely the amount of time that it takes sound to travel from your speakers.

To the human ear, once you get to a certain amount of latency, we hear everything as being simultaneous.

so you do want to make an effort to ensure that your whole system has less than 10 milliseconds latency because the human ear can hear it, but it's not worth it shaving it down from 6 milliseconds to 5 milliseconds.

Human beings evolved to be able to cooperate and play music together in small groups, so if you have a violin quartet, even though they might be 10 or 15 feet from each other, everyone is able to hear each other and somehow the brains of all the musicians go into synch even though they are all lagged by 10 or 15 milliseconds.

You'll notice that with a bigger ensemble, say a symphony or marching band, you get people that are 20 or 30 feet away and at this point the brain can't synch things up so we have to watch the conductor to stay in time.

If you get a digital piano with 20 or 30 milliseconds of latency you will definitely notice it and it becomes impossible to play, it's like you are playing in molasses and everything slows down as you wait to hear yourself.

But below 10 milliseconds, the brain automagically is able to hear the latency as simultaneous.

Pianisti is absolutely right about USB being fast enough, and the salesman is trying to sell you something you don't need. You have less than a single millisecond of latency from the USB.

Of far more importance in regards to latency are the sound cards and drivers, that's where you get most of your latency problems.

It should also not be forgotten that an acoustic piano has latency ;-)
At fortissimo the key travel speed is about 1m/s. For 10mm this gives 10ms. And the hammer will need a little bit of time to hit the string.

I feel most comfortable with 7ms and below, but 10ms is acceptable. More than 10ms doesnt feel responsive.

The other point is: I believe, it is impossible to do a 100% emulation for the pianist. The pianist will always have an experience that is different to a real piano.
The pedal has no taktile trigger point. The keys might have an escapement touch effect, but so far I know, the escapement effect is not functional with a software piano library.


So the pianist must accept that this is a different kind of piano.

The point is: If a recording is made, then the listener will not necessarily know it.

First off, how in the world will you connect those round MIDI ports to your computer? You 'd need a MIDI to USB intrface and then you'd be going over USB. OK there are a few MIDI to Firewire interfaces. But this is a non-issue.
The sales guy is wrong about USB. It is not the source of latency. On a typical Windows PC almost ALL the latency is caused by the Windows audio drivers. This is the software that sits between Ivory and your speakers. You will need to replace this with a third party driver. On a Mac, all of this is a non-issue they are set up for music production at the factory.

Yes if the piano has a continuos damper pedal then it will send in the range of about 0-127.

That said, I would MUCH prefer to have a piano with "real MIDI" connectors then just USB. USB can only connect to a computer whie MIDI is universally compatible and con interconnect with any other MIDI device. Get MIDI but not for the reason the sales guy said.


Also yu will be needing an audio interface. Many of these have MIDI in the same box. Get one of those and save some cables. For audio Firewire is slightly better but hardly matters for just two aduio channels

i like usb better than MIDI because it means there is one less brick on my power strip

Hi,
All good info posted above. I got Ivory as a recommendation from a computer geek friend a couple years ago and really hated it. I looked into it and almost everyone I spoke to said that it had to be run from an external hd or different computer that the one recording audio which made it a bust for me.
As computers get better, some things improve and others get worse.
When I made my album for Yamaha on the Disklavier the midi data stream regarding pedaling was fully supported and at this time Yamaha is the only company I know of that has that part of pedaling really covered. How it ends up sounding will depend on the sound you are triggering.
I still find midi very useful as there are many occasions where taking a midi file to a good Disklavier grand gets the results a client wants.

Hi peter, could you elaborate a bit on the CPU-load-issue? This is the first time I hear that external (USB) Sound-Interfaces are in some point inferior to On-Board-Sound. How does the CPU-load of USB-devices relate to multiple cores? What kind of CPU are we talking of? Do you imply that your Eridol was a bad buy (or did you get rid of the crackling)? Do you have any idea, if other USB-Interfaces behave better in terms of CPU-Load?

I was planning on buying an EMU 0404 USB until I read your post... now I am not quite sure anymore.

With the external USB I get occasional crackles.
This is like a vinyl phono recording, it is not very disturbing.

But if I use the inbuilt soundinterface I have no crackles at all.

The difference can be seen with windows task monitor. The USB Interface imposes up to 30% CPU load and the internal interface, wich is connected to the PCI-bus imposes average 5% and peak 10%.
When I am back at home I will make a screenshot and post it. I would be VERY interested if this is true for all USB Sound interfaces.

I still use MIDI over USB. This does not cause CPU load and works very well, but I believe, sound via USB is not the best solution.

There are professional soundcards for the PCI Bus, for example made by m-audio.
These are cheaper than USB Interfaces. If you can use these, I would prefer it. For Laptops there are PCMCIA cards, but these are expensive, and have bad driver support.

My Computer is an MSI EX600 Laptop, Windows Vista, 4GB RAM, Intel 1,82GHz Core2 DUO Centrino, Realtek HD Audio inbuilt.

Peter

Interesting - I would have expected crackling noises when reaching 100% CPU load (well 30% would mean 60% load on one core though - peaks might hit the full load occasionally I suppose). I myself have a desktop computer with a QuadCore 2,5GHz CPU. I'd still prefer an external sound interface since my PC is some meters away from my DP and I'd like do control the volume while sitting at the piano (also the sound interface is supposed to work as headphone amp when the computer is not running).

Is there a difference between using heavy sample-based virtual instruments and modeled ones?

I believe, if there is 100% CPU load for 10 milliseconds, the task manager will not display it....

I use galaxy II and this - if the internal soundinterface is used - causes very low CPU load but uses about 200MB of memory.
It needs a fast external disk, this can be an USB II disk or a fast SSD or a FAST USB Stick. This all works well with low CPU load, I tried it all.

I /believe/ that the problem would be worse with pianoteq, because this uses the CPU a lot in a timecritical manner.
I dont know that from experience.

Edit:

I have just seen, you use a 2,5 GHz Quadcore. Probably this will be without these Problems, this is more than twice as fast than my computer and because there are 4 cores the realtime audio capabilities should be better.

I have no crackles with my USB Interface when I use more than 10 ms latency, so with a faster computer 5ms should be reachable.

Peter

Yes there is a huge difference both in sound and in how the computer resources are used. Lots of debate on this but the 10 second answer is that Models are more "playable" and responsive while sample can sound more "real". The resent trend in DPs is a hybrid. Models are smaller, using much less disk and RAM and slightly more CPU.

As for the CPU utilization causing "crackling", that is the reason for the buffer in the audio driver. 10 or more milliseconds worth of data are always available even if none are being computed. But likely a few more than "none" are being computed. One problem with using Windows for music is the multi-queue feedback CPU scheduler is non-deterministic. It is designed to do some good things but one of them i NOT short term predictibilty. So the length of the buffer has to cover a worst case. You end up either neededing a computer that is well more powerfull then you really need or a very long buffer. Most PC users solve the problem by doing a bit of both.

Part of the problem is the 1.8hz clock speed. That is slow by today's standards. There my be other problems with your notebook in the way interruts are shared and many other things.

USB does have a problem in that the USB hardware takes a lot of CPU overhead to run. But you should not have problems until you get to 6or more audio tracks. Then most everyone turns to using Firewire interface.

30% CPU of one 1.8Ghz core to drive a USB interface is not unreasonable

It is debatable id ANYTHING made by m-audio is "professional quality" You'd have to go into some studios and see what's in use there to find out. My guess is that you'd see brands like Apogee MOTU, Focusrite, and others mostly using Firewire.

I just realized that you run the virtual piano from an external device that is also run via USB. Could it be that this causes some conflicts with your USB sound interface?

The audio data is two channels of 16 bit samples at 44,100 samples per second or 1.7 million bytes per second. The USB interface runs at 480 million bytes per second. In other words USB is about 200 times faster then required for stereo audio. But, USB is a very in-efficient design and can't move data at anything close to it's full wire speed of 480MB/S. So it works out to be maybe only 100 times faster then it needs to be.

The amount of data that needs to be read off the disk can be much greater than what is sent to the audio interface. If multiple velocity layers are read and interpolated and then you have some polyphony. I think in a worst case a USB disk could be a bottleneck. In fact here is a quote from Ivory's web site:

Firewire is not so in-efficient as USB and can run at near "wire speed" and requires very little CPU overhead.