Pianoteq 7 on Raspberry Pi 400 -- with iPad as interface! (Page 2 of 3)

Yes that option just runs dhcpcd in the background so it doesn't block the boot process, but dhcpcd still takes a while to dynamically acquire an IP address.

Hmm good question, that is strange. It may be that VNC is blocking until the network is ready?

I tried to integrate Pianoteq some more into Raspberry Pi OS by writing my first systemd "unit", inspired by eriks, thank you!

/etc/systemd/system/pianoteq.service

[Unit]
Description=Pianoteq

[Service]
Type=forking
User=pi
LimitMEMLOCK=500000
LimitRTPRIO=90
ExecStart=/usr/bin/vncserver
ExecStop=/usr/bin/vncserver -kill $DISPLAY

[Install]
WantedBy=multi-user.target

After creating and editing ...

# systemctl daemon-reload

I can start and stop my Pianoteq/VNC environment now manually with ...

# systemctl start pianoteq
# systemctl stop pianoteq

... and automatically at bootime with ...

# systemctl enable pianoteq

Because the unit file has its own Realtime-Prio, I commented/disabled these options in limits.conf/audio.conf.

I don't use options like 'LimitNICE=-10' and 'Nice=-10', because they seem to have no or no positive effect in my system.

Attached an example of the running processes shown by htop.
I hope they look like Modartt intended for a RTPRIO host.

Funny that you couldn't relax too

But eventually we are searching in the wrong place. How many seconds does it take, before the kernel starts running? Are the missing seconds here?

The Pi has to do a systemcheck, start a bootloader, eventually waits for a hotkey and finally has to find, read and boot a kernel on the USB SD-card.

PS: Isn't your ControlGroup option obsolete? (can't find it in man systemd.directives)

PPS: Because Pianoteq is a systemd unit now, it appears in the plot (large red bar).

Yes, and that's the reason I assume the time is lost outside of kernel- and userspace.

... 2013, could be a bit outdated.

I will take a break, my last changes were finally:

-  purged cpufrequtils (too much SysV-scripts all around)  and replaced it with linux-cpupower. Applied command now:

sudo cpupower frequency-set -g performance

- in my unit pianoteq.service is one more directive now, that places VNC/Pianoteq at the end of the chain, after the login getties:

[Unit]
Description=Pianoteq
After=getty.target

[Service]
Type=forking
User=pi
LimitMEMLOCK=500000
LimitRTPRIO=90
ExecStart=/usr/bin/vncserver
ExecStop=/usr/bin/vncserver -kill $DISPLAY

[Install]
WantedBy=multi-user.target

Result once more:

Top!

SD-Card here:
Intenso 16 GB UHS-I, micro SD 1, HC I, CLASS 10, "up to 45 MB/s".

Cheers

Good to know, thanks!

With raspi-config I changed the behaviour of the power LED "to flash during disk activity".

I see that it starts flashing after 11 s from power-on. The Bootloader seems to be the missing link.

This is extremely cool, Yacob! You have some great tweaks here, such as being able to change instruments directly with your piano! That's really great indeed.

If someone from Modartt sees this, they might consider sharing your write-up with a larger audience via the FAQ. Thanks for sharing!

I just got around to fiddling with this again today. As an update, I got a hifiberry. It seems to work pretty well!

Thanks for showing the rest of your config files, groovy. It saved me a lot of fiddling with vncserver. I was able to replicate this setup and I think I will be using it now!

Here's something I performed and recorded live on the Raspberry Pi 400 (second half):

https://youtu.be/_QCaMsV8gMs

I'm a beginner pianist, but what Pianoteq afforded me here was the ability to create something individually unique, and hopefully, something musically interesting.

The Raspberry Pi 400 never broke a sweat with this piece. It was flawless and invisible.

I ran Pianoteq on it at 48000 Hz, 128 samples, and 256 polyphony (never got this high), in real time. I had two simultaneous VNC sessions open to it to monitor recordings. I controlled switching purely from the Novus touch screen.

General Pianoteq highlights:

• I never lost a recording -- MIDI auto-archiving is priceless.

• Capturing MIDI with the ability to tweak voice settings and re-render extremely high quality audio is priceless.

• Reviewing my performance on Pianoteq allowed me to identify many mistakes, such as missed or silent notes, and diagnose issues with my performance. Even when I wasn't performing on Pianoteq, it was always recording. Truly instrumental to the learning process, when you can no longer have a teacher by your side. It taught me to me listen more intently while playing.

I left the Pi on for many hours and it was rock-solid.

Hui, these are interesting numbers! Great!
Would be nice to see also a screenshot of a normal load with the MIDI file "Blues Demo".

Cheers

Thank you, navindra,
will install the beta-version!

• Backup your SD card. I skipped this, but it is needed.

• sudo apt update

• sudo apt upgrade

• These are the steps you want to think twice about:

  https://www.raspberrypi.org/forums/view...p;t=250730

  After these steps, I found:

  • The new firmware breaks the Power On/Off button on the Pi 400. I reverted the firmware by force reinstalling the regular raspberrypi-bootloader package and that fixed the issue.

  • With the 64-bit kernel, my external USB device occasionally remains powered with "sudo halt" but does power down correctly with the button.

• The next step is simple conceptually but harder to explain here. The Pianoteq 7 64-bit binary wouldn’t load without the 64-bit loader and libraries being present.

  What I did was copy over the binaries from the image at https://downloads.raspberrypi.org/raspi...020-08-24/ — and this is the bit that’s hard to explain. I had Ubuntu running via Multipass so that I could read ext4, then I pointed a loopback device to the Raspberry 64-bit image, and I mounted the ext4 partition on the Ubuntu instance. I made a tar archive of the binaries I needed, scp’ed them over to the Pi, and untar'ed them there.

There are different ways of accomplishing the same thing (you don't need Ubuntu at all), but this isn’t very intrusive to my current installation, and is trivial to revert.

You might also just want to install the Raspberry Pi OS 64-bit beta on a different SD card and start fresh.

Ok, installed the beta-version of raspios_lite_arm64 now ...

cat /proc/version
Linux version 5.10.5-v8+ (dom@buildbot) (aarch64-linux-gnu-gcc-8 (Ubuntu/Linaro 8.4.0-3ubuntu1) 8.4.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #1392 SMP PREEMPT Sat Jan 9 18:56:30 GMT 2021

... and more or less replicated my environment with tigervnc-standalone-server and the remote wireless GUI.

The performance index with pure 64-bit is higher than with 32-bit raspios, but still too edgy to use pianoteq's internal samplerate of 48000 Hz. Blues Demo with NY Steinway D Classical recording for example:

With the standard non-overclocked kernel (1800 MHz) the performance index is 29 on my arm64 system.

To be on the rock solid side I have to reduce to 32000 Hz internal sample rate on the 64-bit system (like before):

Because I have a self-compiled kernel on my 32-bit system (1000 Hz systemtimer) and the arm64 has a few issues with the halt/power-button I probably wait with the migration until arm64 is finally released.

Thanks for the tips everyone, I also did this today on my overclocked Pi4 and while it took the best part of the day to replicate my setup, it was well worth it as the performance increase is quite substantial! The performance index went from 20 to 29

Let me ask what might be a stupid question (please be gentle):

How much simpler would this list of instructions be if we forgo the whole VNC iPad controller idea?  Meaning, you configure the RP4 to boot into PianoTeq, load a preset, and just run like that? 

If this would make the instructions noticeably easier, perhaps we could fork the instructions for such a use case.

While changing all of the presets can be fun, for my kid's piano lessons and practice, we just need one default piano (for us the Steinway D).

Let me know.

@Yacob, excellent instructions!  I think I will try this soon.

A few questions/clarifications (I too am a raspberry/linux newbie):
  1 At the end of Section 3.1, "sudo cpufreq-set -f 2200MHz", Does this just make sure that the freq is 2200 for the current session?  And will be unnecessary if/when we reboot?
  2 How do we reverse the action of Section 3.2?  Meaning, after this reboot, assuming you had a mouse and keyboard, how could you choose to go back to Desktop?

This will not stick unless you put it in a script right before you start Pianoteq:

pi@pianoteq:~/scripts $ cat Pianoteq7.sh
#!/bin/bash

sudo cpufreq-set -c 0 -g performance
sudo cpufreq-set -c 1 -g performance
sudo cpufreq-set -c 2 -g performance
sudo cpufreq-set -c 3 -g performance

# OR if you install cpupower, you can delete the cpufreq-set commands and uncomment:
# sudo cpupower frequency-set -g performance

/home/pi/Pianoteq7/arm-32bit/'Pianoteq 7' --multicore max --fullscreen

Log in and run "startx" from the command line.

You can also also run "sudo raspi-config" to change your desired settings more permanently.

You can also run "vncserver" which will run a new virtual (full) desktop that you can access via VNC.

Yes, I believe that is enough.

Good luck!

I think you are confusing overclocking with setting the cpu frequency to performance mode (leave the frequency to max).  If you are following Yacob's guide, we had already put an overclock set of commands in the /boot/config.txt file.  And in Section 1.8, we created the script to put the cpu's in performance mode.

So again, given the steps in the guide,
"sudo cpufreq-set -f 2200MHz", Does this just make sure that the freq is 2200 for the current session?  And will be unnecessary if/when we reboot?

I am going down this path after being inspired by the folks on this thread  -- I do have some prior experience in linux, etc. but am getting stuck somewhere -- I've created all of the scripts, performance tweaks, perms, etc. downloaded the v 7 and followed Yacob's excellent step by step through 1.9. I have the pretty icon on my desktop, but nothing happens when I launch the icon or the file  in the home/pi/pianoteq folder. The cursor goes to hourglass briefly and then nothing - no indications in task manager. I'm using a 400 with Raspbian installed. Thanks in advance for any tips and advice.

If I click on the extracted Pianoteq_7.so file that is in the arm-32bit folder right after extracting the zip, I get a prompt to "select an application to open "shared library" files"  What file is the executable in this case? I downloaded Version 6 and it actually has an "app" in the download and that launches for me, but I don't see the equivalent binary / executable in the 7 extracted folder.

Really interesting!
I am curious, are you / can you feed the Pianoteq sound back into the Novus to take advantage of its amplification and speakers?

I know the the Pi 400 is faster out of the box. Seem to remember it runs at 1.8Ghz vs 1.5 in the normal pi. So maybe you can get a smaller buffer or more polyphony. On the downside you lose the small form factor.

Hi navindra. Thank you for sharing your experience.

I have a few question and hope you can answer them to help me better understand the general function/purpose of various VSTs.

For now I have no intention of running the PI 400 setup as you described in this thread. I'm not that bothered about the cables etc, so I will want to connect my NV10S (once it hopefully arrives in about a month) to my PC which has a 10700K, 32 GB RAM, PCIe SSD etc. The reason is that I want to run Pianoteq (or similar) from a device that (at least from what you said) will not break a sweat running even the most extreme scenarios.

Thus, my questions are general in nature regarding these types of software and what they do in comparison to running the piano with default out of the box software.

I'll give you a bit of background to help you answer my questions with as little effort as possible.

I bought the piano because I have neighbour troubles when practicing. They don't mind the teaching, but apparently the repetitive nature of practicing is really annoying to them (and I can appreciate that). I can still practice a small amount per day non silently, but I will need to practice mostly with headphones. I went for the NV10S because after trying it out 45mins per day for 4 consecutive days (unlike with the Yamaha N1X and N2) I was fairly convinced that for all intents and purposes it behaves just like an acoustic grand. The sound from the speakers doesn't quite represent a concert grand, but for me it's as good as a small grand and certainly better than my truly excellent Yamaha U3 upright.

I'm sure you've noticed that I wrote "fairly convinced" and that's because in the end I was still left with maybe 0.1% of doubt. Of course it's difficult to make a definitive determination after 4 days of short practice; and once I get the instrument this might quickly be dispelled through regular use. However once I found out about different possibilities (partly through your YT videos:) I have decided to prepare in advance and investigate if perhaps it's possible to dispel that remaining 0.1% of doubt in case it doesn't get dispelled just through regular use of the piano in it's default state.

In the process I worked out what I think is the key difference between the pianos default functioning (+ other sampled software doing essentially the same) and Pianoteq; and which was possibly the cause of the small amount of niggling doubt. Namely, the number of velocity layers on the Pianoteq is 127 vs up to only 16 using sampled sound, giving infinitely better dynamic control an making it in that respect essentially indistinguishable form an acoustic piano to the human ear. This to me is actually the key parameter to be able to convincingly mimick the acoustic grand on a digital instrument. The other is the action and of course the NV10/S has that and then some (which is also why I didn't get the Yamaha's: no proper pedal mechanism and the action is only OK while the Millenium lll is simply amazing!).

So with the background out of the way these are some of the questions I have and would greatly appreciate if you can offer your findings/opinions:

- with Pianoteq you can tell that you get infinitely more dynamic graduation and this in turn allows the instrument to without a shadow of a doubt feel and behave 100% like an acoustic piano? Or is it all about just a better quality of sound? Or both?

- if the answer to above is yes would this be the same with the Garritan software? Or is that still only a sample so the velocity layers are once again limited to max 16? (I ask this because I really like the CFX sound in that software, at least from what I've heard on YT).

- once set up, whichever software I chose, the sound can be reproduced through the speakers and the headphones? And in terms of the sound is it an improvement through the speakers, too, or only through the headphones? I ask this because as mentioned I will use the instrument to teach on, as well as for my practice.

-  how does it all work? This is what I understood so far and you can let me know where I might be wrong or have perhaps misunderstood: you connect the PC using a printer type USB cable with the NV10/S. You also connect from PC line out to NV10/S line in using a dual male headphone jack type cable. So it's 2 cables total running from PC to the piano? Then you install the software and you're pretty much good to go?

- while using these VSTs do you still retain the virtual technician and other functions on the NV10? For example, when trying out I found that unless I set the minimum touch to 1 the piano is a lot less convincing as an acoustic instrument. Also some headphone settings really make things a lot better vs other settings etc.

I tried to condense it as much as possible and I sincerely hope it's not all to much. Any response will be greatly appreciated. Thank you.