?Zero latency? (Page 3 of 3)

Now here's an interesting post:
http://forum.cockos.com/showpost.php?s=...ostcount=6

It mentions the concept of a "safety buffer", and I suspect that is what I have run into with my USB interface. If this safety buffer is required for USB, I'd rather see this buffer in the control panel. By all means set a minimum size for this buffer, but allow us to modify this buffer, rather than some "fake" secondary ASIO buffer. With only one, large, buffer, we could at least benefit from less overhead, because the system would not be rapidly servicing the "fake" small ASIO buffer! Another benefit is that we would have a much better idea of the ACTUAL latency, if we could see the buffer size in the control panel.

It also mentions that the best Firewire interfaces have only very slightly more latency than the best PCI/e devices.

Greg.

Sorry to go little bit off-topic but still talking about latency and its negative effects:

I have learned to play with some latency; I notice it but it doesn't bother me anymore that much. When 1-2 years ago I started to play PTQ and other software pianos this was my main problem.

But what I can now see is this: too much latency can have a negative effect (at least to my) piano technique. I can notice that when practicing with "latency piano" with play-alongs for example I begin to play with more tension and natural relaxation is lost. This can be very harmful to technique. It's more evident in faster tempos and faster passages.

We have to take this latency-issue seriously and be careful with it!

Guys, thank you for the discussin again :-)

Glenn: I didn't report or raise the problem is in the Pianoteq. I've reported the latency issue and ask where the problem is. To be honest Pianoteq was the last suspicious part (it doesn't matter I'm not a technical expert).
I don't think I "jerry-rigged" something together but what was free on the board :-).
So the 1st configuration were my friend's Clavinova 370 connected via USB to desktop 3Ghz quad core, 4GB RAM, HW on-board sound card, ASIO4All, PT settings 64 samples (1.5ms). So do you think the problem is in the sound card and ASIO4All? It should be replaced?
2nd configuration was V-Piano connected via USB or MIDI cable (both were tested) Edirol U-101 connected via USB to my notebook, 2.2 Core2Duo (BTW 7200 HDD), 4GB RAM, ASIO, 7.1 ms. Felling of latency was completely same as the 1st configuration.
So what should I try ? Try the system with desktop and PCI sound card and native ASIO?
Anyway I don't think I have to or can adapt or learn to adjust. This is very critical to me and it doesn't make sense to change my playing, feeling and everything what I learned in my life :-) In playing you think in advance and sometimes you need to react/interact on the sound produced by the instrument very quickly. In very quick passages I'm not able to play in "real-time" mode :-) and operate with sound colors. I'm not able to play with sustain pedal also and pedal is very important, it is not only push and release. A lot of aspects against the latency.

Greg/skip: Thank you for the latency measurement you did for me. Maybe I'm too fussy maybe not :-) My friend played first 5 notes with PT and said - heyyyy there is a huuuuge latency mate. This is really bad. I said yes, I can feel that also. After two hours playing and testing he said. Quite good sound, nice dynamic (BTW much better then any Clavinova)...plainly very good simulation but it is not acceptable with this latency. If you resolve the problem and when you buy the system, we can try to record some 2 piano concertos....
I've tested the PT with all possible instruments with buffer size checking and the latency was same :-(
I will try to find some studio with desktop computer, but it will take some time....

I don't know if anyone has addressed the following possibilty before or not, I haven't read every post.

You can have Asio4All, or your onboard soundcard driver, or an external one, or whatnot, telling you what latency there is at certain settings so you THINK you've got 7.1ms latency happening- and the delay is awful- but the problem may be that you're simply not accessing the driver that you think you're accessing.

It may be as simple as your settings being incorrect, but not in an obvious way. Maybe you're accessing a Direct Sound driver instead, or another native driver- these have latency that sometimes exceeds 300-400ms!

7.1ms latency would not bother most people much, if at all. Your experiencing very bad latency, and your friend verifying how bad it is, indicates with a high degree of certainty that you're getting WAY more than 7.1ms. When things are so far from the norm, there's often an easy, and easily overlooked, explanation.

As has been suggested, other hardware and software components could be responsible. Something as simple as a bad midi cable? It sounds as though your processors are plenty adequate.

Good luck!

Michael

FYI, I'm experiencing a bit of arpeggiation when I use ASIO4ALL with my integrated IDT audio interface on my Dell Latitude E6500 laptop, and using the Fast Track Ultra for MIDI input. I'm mainly noticing it when I play just two notes together, however, it feels a bit awkward when I play normally, too.   I tried playing Glenn's torture-chord (across a looped-back physical MIDI connection), but I didn't notice any problems with this.

Moving the audio back to the FTU, the problem disappears.

Pianoshrek2: whatever you do, I strongly suggest NOT using ASIO4ALL for any of your testing at the moment. I know we have said this 100 times already.

Greg.

Oops, it appears I missed a lot. Just couldn't bring myself to read all 100+ posts before replying. 'My bad' as they say

It still seems to me that this is extraordinary, certainly outside the norm, and that there's got to be some really oddball reason lurking somewhere.

Michael

Luc,
I disagree. If the timing of audio were being affected, there would be glitches in the audio, but there are not.
The "glitches" are in the actual timing of the notes. Everything sounds fine, except for the arpeggiation.
It sounds just like I am arpeggiating the chords with my fingers.

A plausible explanation is that for whatever reason, when I use ASIO4ALL on this particular system, it is performing high priority processing that is robbing the M-Audio driver from processor time to the extent that the MIDI timing is being adversely affected.  Note that I tried running that DPC Latency Checker with ASIO4ALL active, but it didn't report any problems.

Greg.

Hello,

I like "skip's method" and tested it with PTQ  on my netbook  Asus eeePC 901. I measure around 32 ms latency between my fingernail hits the keys of my Kawai ES3 stage-piano and the sound is heard from the keyboard-amp.

Using the sound-engine of  the Kawai ES3 directly the latency is 22 ms.

My PTQ-Setup is:

Kawai ES3 -->
MIDI-to-USB Interface "Worlde MidiPort" -->
EEE-PC 901 -->
Debian GNU/Linux, customized for Realtime-Use -->
PTQ 3.0.5, 44,1 kHz 64 samples, C3 Solo Recording
USB-to-AUDIO Behringer UCA222 -->
PA

I did the "measurement" with another eee-pc using its both built-in two micros and the audio-editor kwave.

One personal statement: Good latency is VERY essential to me and cannot be fast enough, it is very important to get "connected" with your instrument.

- groovy

P.S. Forgot to mention the distance of the microphones: 2 m towards the keyboard and 1 m in the opposite direction towards the loudspeakers.  So there is effectivly a latency of  3 ms caused by the distance keyboard to micro. This means I can correct the latency of my PTQ-chain to 29 ms.

Groovy,
Do you know the USB polling rate on your Linux system?

Thanks,
Greg.
p.s Just an aside - my nick name is related to Skippy the kangaroo - I think it's a poor choice, because it's a real name in the U.S.

Groovy,
This "USB Poll Rate" that I refer to may in fact be ONLY related to the mouse poll rate. If the hack that I did only affected the mouse, then that would explain why I didn't notice any change in my latency tests.  It seems that gamers often increase the poll rate, to improve the response of USB mice.

Also, looking at some results from other USB MIDI interfaces here: http://earthvegaconnection.com/evc/prod...sults.html , some of them are actually pretty good.

You appear to have a nicely working system - congrats.

Greg.
p.s tch tch tch tch....

Using the Delta 66 PCI card, I set the buffer size way up, so that my latency roughly matches Pianoshrek's.  I  used the PCI card because it does not suffer from the startup delay that my FTU has. I verified the latency with a test strike.

It is TERRIBLE!  If Pianoshrek's measurement really is representative of his latency, there is just NO DOUBT that this is his problem! 

The reason I hadn't done this a lot earlier is of course because I thought my latency was already the same as Pianoshrek's, but of course we now know that it is only the very first note, from idle, that has that latency - most other notes, once there is activity, have a much smaller latency, and that's why I never noticed a problem when playing!  I'm curious to know what the idle timeout is, actually.

Greg.

Pianoshrek,
It is clear to me now that USB audio interfaces are fine in terms of latency, so I think you should try a USB device for your audio. It is possible that both your latency and sound quality will increase. (have you tried that Edirol USB interface on your laptop, after the tuning?)  Regarding sound quality, I think the main risk with the internal audio interface of your laptop would be a lack of bass. (I'm not saying that it will definitely have a lack of bass, but it's possible)
Can you describe your sound quality problem? Does music sound ok? How about solo piano recordings?
Are you using headphones directly connected to the laptop, or an external amp/speakers?

In case you didn't ready every post carefully - I have a serious problem when I use ASIO4ALL on my laptop - I experience arpeggiation.  I'm not necessarily blaming ASIO4ALL itself - it could just be a compatibility problem.

Note that I assume you have disabled CPU throttling on your laptop - is that right? Verify with the command line utility: POWERCFG /Q   The Processor Throttle lines should read "NONE". (this is on Windows XP, anyway)

For audio, don't forget that you could get a PC Card interface, which is (I think) equivalent to a PCI card - e.g that Indigo interface that was mentioned earlier. (assuming you have a PC Card slot) The other option is Firewire, which it appears will give you very slightly better latency than USB, in general.

What did you tune in order to decrease the latency?

Greg.

skip: I also experienced problems wit ASIO4ALL since I "introduced" it in my system. With an old version of Logic (5.5) for PC, it even "crashes" my MIDI inputs! Looks like there is something here...

I've not used the Indigo, but it looks like it would be fine, and we have a report here from someone who is very happy with it.  You're right though - these types of cards don't seem to be nearly as popular as USB and Firewire.
PCI interfaces still seem to be popular. (but that won't help you on your laptop)

If you have an "Always On" profile, that will probably be better - it will probably COMPLETELY disable processor throttling. That's what I use.

Maybe you could make another recording, to see how much your latency has improved?

Regarding the sound quality, that sounds like it could be clipping.  Have you reduced the volume in Pianoteq itself? The default volume in Pianoteq is often too high, and does produce clipping.  If that doesn't work, try reducing the volume of your soundcard (the main "Master" output volume). If the harshness disappears, that means that the soundcard cannot drive your headphones loud enough for you.

Greg.

OK, will try to find some studio with Indigo and will try the Edirol with USB again...
I'm not able to do the latency test again in the following 3 weeks with my friend's system, he will travel abroad and me too.
Volume was reduced (without blue and red lights - this is not explained in the manual, what exactly the colors means - maybe limiter and too high volume). Will see about the another sound card and the quality.

Pianoshrek,
If your output meter is going into the red, that's too high and it's probably clipping. Reduce the output down to about -6 and try again.  You could try turning OFF the limiter, to make any clipping more obvious, in fact.  (the limiter will hide the clipping to some extent)

Also, you should have your Wave volume on 100%, in your soundcard's volume control settings.

EDIT: Oops, I see that I have mis-read your post, Pianoshrek - you have already reduced your volume in Pianoteq.   Try reducing the main volume in your soundcard mixer, as I said, just in case you are overloading your soundcard.   It is still possible that Pianoteq is clipping, though - sometimes I hear a bit of clipping, even though I can't see any blue or red. If you haven't tried it on -6 yet, give it a go. However, if your limiter is on, it is probably unlikely that a small amount of clipping would be audible, because the limiter does greatly reduce the effect of clipping.

Greg.

Groovy,
Thanks for the info on the system timer. I have no idea whether something like this is available on Windows. 
If it is available, though, I would hope (perhaps naively!) that the manufacturers of audio/MIDI interfaces would know about it, and if it helped, would either configure the system optimally during install, or advise the user on how to do this.   

Greg.

Hvaatsen,
Keeping the sample rate constant (44.1kHz), I get very different results than you do!   In standalone mode, my audio is glitch free down to 704 samples (16ms) In Cantabile Lite, I'm up to about three THOUSAND so far, and it's still producing a bit of glitching.

Note however that I am using my M-Audio Fast Track Ultra for MIDI input, which MAY actually be defective. (M-Audio have suggested this). So I can't make any conclusions about anything at the moment, when this component is used.

I get arpeggiation with DirectSound too - the delay due to the large buffer sizes does not only cause a pure delay - it causes arpeggiation.  I have a theory that this could be due to the fact that the audio interface does not actually have a physical buffer, so perhaps the driver writes the buffer to the hardware in a high priority process, during which processing of MIDI is blocked. We could contact the authors of ASIO4ALL - they would have a deep understanding of all this.

Greg.

My recent tale of latency problems and the resolution:

I have a Roland PCR-800 midi-controller hooked up to a Creative X-Fi Platinum (the one with the 5 1/4 front panel with midi connections, etc) sitting in a quad-core (Q9300) PC running XP SP3.

Using ASIO4All 2.9, Pianoteq 3.5.2 allows me to use 128 samples at 48khz with 256-note polyphony with no crackles and with the cpu hardly breaking a sweat. Performance index is around 30+. Note: Using the Creative ASIO driver introduces crackles unless sample count is bumped up to 256.

So far so good.

Yet I was still experiencing latency. Not half a second, but just enough to be clearly noticable. I had not used the system for a while (I have a Roland RD700SX which I often use with a notebook instead) but I knew that this had worked with far less latency in the past.

I tried moving the midi-controller from the USB PCI card to one of the USB ports on the motherboard. No change.

I then tried switching to the 5-pin DIN midi connector on the controller, connecting it to the mini-DIN of the sound card. No change.

It finally struck me that a while ago, I had performed a complete reinstall of this system.

Turns out that the Creative X-Fi card upon reinstallation had defaulted to its "Entertainment Mode". Once I brought up the Creative control panel and switched it to "Audio Creation Mode", all was well again.

Checking with ASIO Latency Test Utility 3.7, the latency reported for the Creative ASIO drivers (this utility would not work with the ASIO4All driver) went from 70ms down to around 15ms. Feels great now/again.

I know it was a while ago when this link was posted, but I seem to have missed it.

Although the results are interesting I don't like the micro switches and metal foil approach, it probably changes what it is designed to measure.
I think fiber optics and high speed video cameras could provide better accuracy while disturbing the mechanism a lot less.

JMAO, etc.

Given that PNOscan has been installed on some fairly "upper tier" pianos I might research measurements that have been made.

On faith alone - - I believe SOMEONE must have used their PNOScan installation to figure such things as hammer flight time and delays in the escapement.

With PNOscan II, you get MIDI High Resolution Velocity Prefix control change number eighty-eight (88)?  While I was looking at the website, I was unable to find it mentioned.