I have said this before but because of dulistan’s very important statement:
”Because people's ears and perspectives are always different”.

This dulistan’s little sentence means everything, the most important - because,

people use different sounding headphones and hear different.
Testing with my headphones, Beyerdynamic 990 DT Pro and Grado Lab SR 125x, I can hear differences in sound…

I wouldn't argue with any of that, but I don't think it's relevant to the question of whether different DAWs converting the same VSTi output to analog using the same DAC outputting through the same monitoring system to the same ears will sound different.

Can we disregard the driver's opinion?

If the driver says "It feels slower" while the engineer says "The lap time was lower", then, yes, I would disregard the driver's opinion. ;^)

dulistan heman wrote:

brundlefly wrote:

No offense intended, but I really have no interest in opinions or anecdotes when it comes to this subject, only well-documented evidence. Just throwing up a bunch of renders that might or might not sound noticably different without any details on how they were engineered is not very helpful.

I can't comment on all DAWs, but based on the one's I've tested, the only real differences in this context are in the timing of note transients in the rendered output which is something that happens in every live performance anyway. Beyond that, and the Pianoteq VSTi's inherent variability from one render to the next without re-initialization, I found no meaningful differences in level, dynamics, tone, imaging, ambience or anything else regarding the basic *quality* of the sound.

Now if you want to talk about the possibility of two DAWs sounding different with a complex multitrack mix of virtual instruments and recorded audio with panning, sends, automation, FX plugins, external inserts, etc, etc., that's a different story. But I'd still question whether one DAW can be claimed to sound objectively 'better' than another, only 'different'.

In the absence of an actual defect in the functionality of the DAW software, the quality of the DAC, amplifiers, monitors and the listening environment have a far greater impact on the sound than anything going on in the box.

None taken. Some people don't hear any improvement sound in Pianoteq 8, and that's okay. Same thing with different DAWs and VSTs. Because people's ears and perspectives are always different.

In my opinion, if I only buy Pianoteq using merely the final export, then Pianoteq 6 is already perfect. There's no need to ask for sound improvement or anything. However, because some people use Pianoteq for hours hearing the RAW sounds, then people definitely have different opinions about what they hear.

Another thing to add. There are some differences between car engineers and the drivers. The engineer uses a fact sheet and a computer, while the driver uses feelings that might be called anecdotes. Can we disregard the driver's opinion? Sure, why not?

Hello everyone, coming on-board a bit late…and just for a moment.

This is how I think about it.
I have said this before but because of dulistan’s very important statement:
”Because people's ears and perspectives are always different”.

This dulistan’s little sentence means everything, the most important - because,

people use different sounding headphones and hear different.
Testing with my headphones, Beyerdynamic 990 DT Pro and Grado Lab SR 125x, I can hear differences in sound……..

……..And - with headphones the sound is in our head, and headphones gives left sound channel sound only to left ear, and right sound channel sound only to right ear (unlike standing beside a real piano or using loudspeakers, where the sound come to both ears, maybe not at the same time, and with  that rooms acustics, not at same time to both ears).
And dulistan’s ….ears are different…..People have different shape of heads, different ear canals, different earflaps, and the size and shape of the head have effect on how headphones reproduce the sound (also interesting, in Ptq binaural mode you can increase the the size of the head, for people with big head, my head is a bit model small). 

All above mentioned have effect on how/what we hear listening to music/Pianoteq/daw/vst.
As said:   ”Because people's ears and perspectives are always different”  Thank you dulistan.

All the best, everyone

Stig

I would also like to add, as an addition to my comment above, what I have learned, that to get the right sound/best sound when comparing vst daw  with using speakers, in loudspeaker listening, the loudspeakers are usually placed so that their angle with respect to the viewer is 60 degrees. This creates an equilateral triangle. The resolution of directions is quite poor, at best 10 degrees. This is because in loudspeaker listening, the signal from the left loudspeaker, which is “intended” for the left ear, is also heard in the right ear and such “crosstalk” confuses the hearing of directions.
In order for the listener to have the correct sound image, the right sound, he should sit in a fairly precisely defined area, symmetrically with respect to the loudspeakers, to get the right sound when comparing vst daw. Even a small deviation from this will shift the sound image towards the closer loudspeaker, which change how the sound sounds. Further away, there may be a wider area.

In reality, many listeners often sit wherever they want. If the reflected and direct sound arrive at the ears with a delay of less than 30 ms, the listener cannot distinguish them. Sounds that have traveled longer than this are perceived more as echoes, which ”colors” the sound…..

Oh oh -  despite all the modern technology we have so I always come back to this ”we hear different” :    ”Because people's ears and perspectives are always different”
Now I can't find anything more to say in this post.

All the best, everyone

Stig

brundlefly wrote:

dulistan heman wrote:

Reading that people have different opinions always sparks my curiosity about the digital sounds.

No offense intended, but I really have no interest in opinions or anecdotes when it comes to this subject, only well-documented evidence. Just throwing up a bunch of renders that might or might not sound noticably different without any details on how they were engineered is not very helpful.

I can't comment on all DAWs, but based on the one's I've tested, the only real differences in this context are in the timing of note transients in the rendered output which is something that happens in every live performance anyway. Beyond that, and the Pianoteq VSTi's inherent variability from one render to the next without re-initialization, I found no meaningful differences in level, dynamics, tone, imaging, ambience or anything else regarding the basic *quality* of the sound.

Now if you want to talk about the possibility of two DAWs sounding different with a complex multitrack mix of virtual instruments and recorded audio with panning, sends, automation, FX plugins, external inserts, etc, etc., that's a different story. But I'd still question whether one DAW can be claimed to sound objectively 'better' than another, only 'different'.

In the absence of an actual defect in the functionality of the DAW software, the quality of the DAC, amplifiers, monitors and the listening environment have a far greater impact on the sound than anything going on in the box.

None taken. Some people don't hear any improvement sound in Pianoteq 8, and that's okay. Same thing with different DAWs and VSTs. Because people's ears and perspectives are always different.

In my opinion, if I only buy Pianoteq using merely the final export, then Pianoteq 6 is already perfect. There's no need to ask for sound improvement or anything. However, because some people use Pianoteq for hours hearing the RAW sounds, then people definitely have different opinions about what they hear.

Another thing to add. There are some differences between car engineers and the drivers. The engineer uses a fact sheet and a computer, while the driver uses feelings that might be called anecdotes. Can we disregard the driver's opinion? Sure, why not?

Hello everyone, coming on-board a bit late…and just for a moment.

This is how I think about it.
I have said this before but because of dulistan’s very important statement:
”Because people's ears and perspectives are always different”.

This dulistan’s little sentence means everything, the most important - because,

people use different sounding headphones and hear different.
Testing with my headphones, Beyerdynamic 990 DT Pro and Grado Lab SR 125x, I can hear differences in sound……..

……..And - with headphones the sound is in our head, and headphones gives left sound channel sound only to left ear, and right sound channel sound only to right ear (unlike standing beside a real piano or using loudspeakers, where the sound come to both ears, maybe not at the same time, and with  that rooms acustics, not at same time to both ears).
And dulistan’s ….ears are different…..People have different shape of heads, different ear canals, different earflaps, and the size and shape of the head have effect on how headphones reproduce the sound (also interesting, in Ptq binaural mode you can increase the the size of the head, for people with big head, my head is a bit model small). 

All above mentioned have effect on how/what we hear listening to music/Pianoteq/daw/vst.
As said:   ”Because people's ears and perspectives are always different”  Thank you dulistan.

All the best, everyone

Stig

dulistan heman wrote:

Reading that people have different opinions always sparks my curiosity about the digital sounds.

No offense intended, but I really have no interest in opinions or anecdotes when it comes to this subject, only well-documented evidence. Just throwing up a bunch of renders that might or might not sound noticably different without any details on how they were engineered is not very helpful.

I can't comment on all DAWs, but based on the one's I've tested, the only real differences in this context are in the timing of note transients in the rendered output which is something that happens in every live performance anyway. Beyond that, and the Pianoteq VSTi's inherent variability from one render to the next without re-initialization, I found no meaningful differences in level, dynamics, tone, imaging, ambience or anything else regarding the basic *quality* of the sound.

Now if you want to talk about the possibility of two DAWs sounding different with a complex multitrack mix of virtual instruments and recorded audio with panning, sends, automation, FX plugins, external inserts, etc, etc., that's a different story. But I'd still question whether one DAW can be claimed to sound objectively 'better' than another, only 'different'.

In the absence of an actual defect in the functionality of the DAW software, the quality of the DAC, amplifiers, monitors and the listening environment have a far greater impact on the sound than anything going on in the box.

None taken. Some people don't hear any improvement sound in Pianoteq 8, and that's okay. Same thing with different DAWs and VSTs. Because people's ears and perspectives are always different.

In my opinion, if I only buy Pianoteq using merely the final export, then Pianoteq 6 is already perfect. There's no need to ask for sound improvement or anything. However, because some people use Pianoteq for hours hearing the RAW sounds, then people definitely have different opinions about what they hear.

Another thing to add. There are some differences between car engineers and the drivers. The engineer uses a fact sheet and a computer, while the driver uses feelings that might be called anecdotes. Can we disregard the driver's opinion? Sure, why not?

Reading that people have different opinions always sparks my curiosity about the digital sounds.

No offense intended, but I really have no interest in opinions or anecdotes when it comes to this subject, only well-documented evidence. Just throwing up a bunch of renders that might or might not sound noticably different without any details on how they were engineered is not very helpful.

I can't comment on all DAWs, but based on the one's I've tested, the only real differences in this context are in the timing of note transients in the rendered output which is something that happens in every live performance anyway. Beyond that, and the Pianoteq VSTi's inherent variability from one render to the next without re-initialization, I found no meaningful differences in level, dynamics, tone, imaging, ambience or anything else regarding the basic *quality* of the sound.

Now if you want to talk about the possibility of two DAWs sounding different with a complex multitrack mix of virtual instruments and recorded audio with panning, sends, automation, FX plugins, external inserts, etc, etc., that's a different story. But I'd still question whether one DAW can be claimed to sound objectively 'better' than another, only 'different'.

In the absence of an actual defect in the functionality of the DAW software, the quality of the DAC, amplifiers, monitors and the listening environment have a far greater impact on the sound than anything going on in the box.

CircleOfFifths wrote:

Hi there. I was wondering the same.. In Logic Pro and GarageBand Pianoteq 8 sounds more clear/vibrant, raw, cleaner/deeper bass, sparkling highs etc, etc than the standalone which also sounds beautiful with just more muddy bass and somewhat less expressive purely is sound processing; same computer, velocity curve headphones and speakers, sounds way more present through the DAW..

Have you read through the entire thread? I think we pretty much debunked the idea that there are meaningful audible differences between DAWs or between Pianoteq Standalone and the VSTi when all possible variables are eliminated.

Hi, Modartt forum.

I'd like to share a little background story about sound in my community. Many people refused to believe that there's a difference between any DAW sounds. Unfortunately, my Jamaican friend disagrees with the majority. He has both Cubase 14 and 8, while I have Cubase 10.5. However, in his studio, he keeps using Cubase 8 during recording sessions for any reggae songs. This raised my curiosity: why did he do that? His answer surprised me. He said that Cubase 8 has the best sound for reggae compared to any new Cubase update.

We all knew that Cubase already has so many feature upgrades and all the advantages of every new upgrade, including not using a dongle anymore. However, because of ear candy, he prefers using old software even though he has to sacrifice time for audio editing and other stuff. The reason he has to upgrade to every new version, including Pro Tools and Ableton, is only because of his clients.

I believe this topic will never get old, including in the Pianoteq forum. I always use Pianoteq standalone for playing and gigging because of its lightweight use. But for recording and exporting, I prefer Ableton 12, where they have new improved sounds that have been retained since Ableton 11, including the amazing reverb effect.

Reading that people have different opinions always sparks my curiosity about the digital sounds.

Hi there. I was wondering the same.. In Logic Pro and GarageBand Pianoteq 8 sounds more clear/vibrant, raw, cleaner/deeper bass, sparkling highs etc, etc than the standalone which also sounds beautiful with just more muddy bass and somewhat less expressive purely is sound processing; same computer, velocity curve headphones and speakers, sounds way more present through the DAW..

Have you read through the entire thread? I think we pretty much debunked the idea that there are meaningful audible differences between DAWs or between Pianoteq Standalone and the VSTi when all possible variables are eliminated.

Strange but that could be related to some internal and different ways to process the sound.

And what is your opinion on the sound in standalone ?

Hi there. I was wondering the same.. In Logic Pro and GarageBand Pianoteq 8 sounds more clear/vibrant, raw, cleaner/deeper bass, sparkling highs etc, etc than the standalone which also sounds beautiful with just more muddy bass and somewhat less expressive purely is sound processing; same computer, velocity curve headphones and speakers, sounds way more present through the DAW..

I heartily forgive in advance if you're unwilling to bother with testing any subtleties of MIDI Timecode between DAWs...

Okay, so I was just going to do one quick test with a melody only, unquantized, but with a fixed tempo of 100bpm in the MIDI file, and it's already a can of worms with the first two notes of two DAWs!

I compared Cakewalk and Pro Tools. The initial transients from the note at 1:01:000 had the same timing and waveform, but they were not nulling very well and I observed the peak level of the Pro Tools transient was 1.5dB lower. Sure enough, raising the Pro Tools track Volume by 1.5dB had that first transient nulling to about -60dB.

The second note was still not nulling so I checked the timing vs. the MIDI. The Cakewalk transient was two ticks late which is unusual because I know Cakewalk normally renders transients exactly on the MIDI note with most VSTis. But more concerning was the Pro Tools transient which was 17 ticks late (960 PPQ)! I started tabbing through the MIDI notes in Cakewalk, and found many more Pro Tools transients were significantly late, ultimately finding one that was a full 30 ticks (19ms at 100bpm) behind the corresponding MIDI note which is totally ludicrous! I checked the Event List in Pro Tools to make sure I hadn't screwed something up, but it has the same 960 PPQ tick values as Cakewalk.

That's enough for me to write it off without looking any further. The "pros" are welcome to Pro Tools. I'll stick with Cakewalk.

I have done a bunch more testing (all at 48kHz, 24-bit which is native for my interface), comparing direct exports from Pianoteq and multiple renders from Cakewalk by Bandlab, Studio One 5 and the Pro Tools trial. When all variables are accounted for, I conclude that the only differences are due to the variable buffering/timing of MIDI notes to the VSTi from one DAW to the next. Two renders from any of the DAWs with a re-launch in between will consistently null to about -90dB which is as good as it gets with two consecutive exports from Pianoteq directly (same result with or without a re-launch).

But no two renders from different DAWs will null consistently because of the note timing issues. They will all null well on the first transient with a MIDI note at 01:01:000 but after that, all bets are off. Short snippets will null to a very low level intermittently as the transients just happen to line up. I suspect if I re-did the test with individual note events and pedal messages all well separated in time and no tempo changes in the MIDI file the nulling between the different apps might get better.

In A/B/C/D listening, there are abolutely no detectable differences in level, dynamics, tone, imaging, ambience or anything else.

So I'm satisfied you can use any reputable DAW you prefer, and not concern yourself with the quality of the rendering.

As far as the note timing variability goes, I do find it a bit strange that there is so much variability when rendering offline, and my initial impression was that the vaunted Pro Tools is actually worse than either CbB or Studio One with respect to the alignment of audio transients with the MIDI. I'd have to do some more testing to nail that down, but given the infinitesimal differences I'm measuring and hearing, I don't think I care enough to bother at this point. ;^)

Truly fascinating results!  You've done a fantastic job experimenting on this!

I'm no expert, but I suspect that the differences you're seeing are could be between MIDI Clock and MIDI Timecode being implemented differently in different DAWs (apologies for those tautologies).

MIDI Clock is the older standard, and it was designed to allow multiple, independent devices to share universal tempo track data (everything from outboard SFX to click tracks to drum pads to anything else that supports MIDI) in fractions of a quarter note.  MIDI Clock is fairly robust but designed for music only and is tempo based instead of time-based (making digital music production for TV and film a horrifying nightmare in the 90s and early 00s--not that it isn't now but that's for different reasons).

MIDI Timecode is an implementation/extension of SMTPE Timecode (the same global clock technology used in live television broadcast to switch between video feeds live on a vision mixer before everything went digital and got tons simpler there too).  Depending on conditions and if Clock is running in an extended resolution or not (older software and hardware was limited to 24 units per quarter note resolution while more advanced and contemporary implementations are 480 units or--perhaps in recent years better than that, I'm not sure), MIDI Timecode could be more precise in some conditions then MIDI Clock and less precise in others, but it always has a clear sense of what time you're at in the file.  Also Timecode subdivides decimal frames (again with various precision between software and depending on whether you're syncing devices to create an SMTPE pulse or if you're simply using an SMTPE-based timeline at some decimal-limited, arbitrary resolution) while Clock, again, subdivides quarter notes which will change with the tempo track.  My understanding is that you could increase the time precision of Clock by increasing the tempo, such that 200bpm was twice the time resolution of 100bpm, but especially at low tempi, 60bpm or lower, Clock was worse than 24fps with no subframe signals/data.

My theory is that some DAWs are somehow limited in timing precision which is creating rounding errors as DAWs interpret the data from Clock to Timecode or perhaps the reverse.  This is especially likely if you're MIDI file is a human performance versus a quantized or otherwise computer-generated track.  Probably the MIDI files would null if the MIDI file consisted entirely of whole notes or breves and an easily divisible fixed tempo (something that 60 divides cleanly into like, well, 60bpm or 120bpm).  That said, it's possible that even then while DAWs should render that file at 1bps or 2bps, respectively, there's some additional dithering or timing rounding going that might be unavoidable (albeit unlikely).

All of this reminds me of the horrible time that Adobe quietly increased (sometime this past decade) the maximum decimal precision of InDesign internal linear units and broke every one of my older layout templates as the higher resolution resulted in rounding errors when opening older files...nothing in the files had moved, but the software's new parameters meant it read old information differently resulting in objects not aligning the way they'd used to and breaking snapping, etc. in wild and crazy ways: the file was the same, the way it was interpreted by the (otherwise identical) software had changed subtly in way that everything on the templates moved slightly from where that same software previously thought they should go.

Again, great work!

I heartily forgive in advance if you're unwilling to bother with testing any subtleties of MIDI Timecode between DAWs...

More resources on Clock vs Timecode:

https://en.wikipedia.org/wiki/MIDI_beat_clock
https://en.wikipedia.org/wiki/MIDI_timecode
https://en.wikipedia.org/wiki/Linear_timecode
https://en.wikipedia.org/wiki/SMPTE_timecode
http://www.harfesoft.de/aixphysik/sound...itmcn.html
https://en.wikipedia.org/wiki/Video-signal_generator

But no two renders from different DAWs will null consistently because of the note timing issues. They will all null well on the first transient with a MIDI note at 01:01:000 but after that, all bets are off. Short snippets will null to a very low level intermittently as the transients just happen to line up. I suspect if I re-did the test with individual note events and pedal messages all well separated in time and no tempo changes in the MIDI file the nulling between the different apps might get better.

In A/B/C/D listening, there are abolutely no detectable differences in level, dynamics, tone, imaging, ambience or anything else.

So I'm satisfied you can use any reputable DAW you prefer, and not concern yourself with the quality of the rendering.

As far as the note timing variability goes, I do find it a bit strange that there is so much variability when rendering offline, and my initial impression was that the vaunted Pro Tools is actually worse than either CbB or Studio One with respect to the alignment of audio transients with the MIDI. I'd have to do some more testing to nail that down, but given the infinitesimal differences I'm measuring and hearing, I don't think I care enough to bother at this point. ;^)

@tmyoung I used the Pro version of PianoTeq 8. Sorry, I should have made that clear.

Np

I discovered that reloading a project or unloading/reloading the VSTi from RAM within the same session resets it to initial conditions such that a second render will null to -80dB or more with the first.

Excellent observation.

The digital world is indeed full of those tricky variances (worth mentioning.. to be fair, on a per user workflow level, probably nowhere near as full of glitches as in the old days with replacing worn tubes and demagnetizing tape machine playback/record heads and a load of other time-sinks which could be expensive). I guess for a lot of people now, the craziest stuff to deal with when making music with digital equipment is the old 'turn it off, and turn it on again'.. in a way I'm amazed at what glitch fixes have lingered more than 30 years in tech, like that old chestnut - but I guess, also pretty tame on reflection.

Well presented there @DEZ. Each file has a 192kbps bit rate - so instantly that gives each file a good start for comparison at least by ear..

With a quick listen, the Pianoteq MP3 file here seemed the best, to me. Listened a few more times to try and get a better beam on things like how resonances fade and felt at first that the DAW files may add a tiny fraction of boxiness there.. but listened for that a few mins later and didn't seem to be able to confirm that.. so on just listening alone, they seem all good - but with still a sense that I prefer most the Pianoteq file here.

The software kindly linked by dikrek above "DeltaWave" ("DiffMaker" was also good, used it later) showed Cakewalk had a better null match than Cubase (40db correlated null depth, vs. 26db, rounded). Other diffs do seem to include 'clock drift' - but.. that's possibly attributable to how Pianoteq renders each performance in real time (not a static lifeless instrument.. like being played in real time on each playback, with variances in various things like all the random generated tiny diffs built in. - things like that are the 'ghosts in the machine' I hinted above in other posts.. and of course, encoding and other things may apply here too..)

In general, no 2 playbacks of any MIDI with Pianoteq generating output would likely give perfect exactly the same results (perhaps unloading from RAM like brundlefly did would get closest.. that in itself is an interesting ghost in the machines to know about.. I personally just suspect something like random seed beginning at 1 again.. so many 'random' things would likely calculate cleanly.. then maybe the start seed would 'morph' perhaps randomly during performance.. ?? would love to know more tho). But should say.. random stuff Pianoteq does as the player performs or on MIDI playback is not about quality of files etc., but just those above-mentioned variances due to the mentioned randomly generated elements to give life to the instrument. All that may show up in (even WAV) tests, as differences if outputting a Pianoteq MIDI file to WAV more than once (did not have time to do that.. but asap I'll try it, see if ).

Possibly, the thing to remember might be, when exporting audio from Pianoteq, choosing highest settings, you're probably going to get something 'as good as possible' - but still, I'm hoping for 320kbps in future (should add, that's for MP3).

Output from Pianoteq as a WAV and the bit rate here becomes 3072kbps.

Indeed using that WAV file as a start point in a DAW (rather than rendering MP3).. and then we're truly much closer to testing 'the same input' as output when rendering that WAV to an output file from within different DAWs.

I output a little musical phrase from Pianoteq to a WAV file (at 3072kbps.. no user input there, it just outputs at that).. imported that WAV file into Studio One, exported that as mixdown also to WAV and received an output file with a wonderful 300db correlated null depth (as good as it gets??).. which really gets us to the region of 1:1 similarity.

The only small diffs were below -150db and inaudible.. each DAW may do different things in this range - and it could have been 'that' which gave me poor results years ago with a different DAW (when processing audio a lot in complex projects). But, very happy to see those, kind of expected results from Studio One.

Even so.. I think many will always be exporting to MP3 (convenience, ubiquitous terminal file type with portable/streaming purposes etc.) and getting all kinds of differences due to like mentioned, encoding via different algos, having unknown different user settings inside their DAW, different audio units/sound cards etc.. but when just dealing with outputting to WAV, seems Studio One (only one I tested here) does a clean job. Honestly, would expect most DAWs to excel also at this particular WAV only task with no huge diffs these days.

MP3 will be what most will output and share and talk about - and often it will be moot to null test at all.. but fwif, I've enjoyed this whole thread, with thanks again.

Another thanks to dikrek - both those software titles do interestingly make diffing super easy (even with obviously diff files, no fuss. Cheers!).

Just now read your post above @tm!  Bit rates were all equal at 192kbps on the outputted files - in case DEZ misses your question. Not sure of the rest but I would love to know more about your final question (does Piantoeq process 48kHz at ceiling, or downsample from a computed 192kHz. Have a vague recollection of perhaps it being mentioned once - but seems that I forgot, if so.. my intuition would flow towards keeping users on 48kHz to only 48related computing.. for CPU keeping.. but maybe not.. as there's probably less system stress on background processing to output? ++ to find out for sure.. there's every chance, if it's downsampled, is there a bottleneck of sorts there? how ever small/unnoticed though it might be.).