@Karlos

Quote:

Karlos wrote: It is interesting that you chose reverb as an example though, since most reverb relies on modelling the absorption of high frequency components of the source sound in the reflection return by attenuation in the air and properties of the reflect surface by... Filtering.

Reverb is not just filtering, not even close, nor is it just high frequencies. It is how those reflections interact/collide with the sounds playing in that mixing time domain, prior to filtering. Did you watch that video link i posted?
Much like tossing a pebble in a pond, if that pond was the source sound. If the frequencies are aliasing, they change the sound differently to how they would if non aliasing. They collide differently and change the sound. That was my point. Thats why oversampling is effective. This dosen't happen the same way in analogue. It just increases in noise.

Last edited by SHADES on 21-Dec-2022 at 11:06 AM.
Last edited by SHADES on 21-Dec-2022 at 11:06 AM.
Last edited by SHADES on 21-Dec-2022 at 11:00 AM.
Last edited by SHADES on 21-Dec-2022 at 10:46 AM.
Last edited by SHADES on 21-Dec-2022 at 10:43 AM.

_________________
It's not the question that's the problem, it's the problem that's the question.

@Karlos

Quote:

I consider myself as an "audiophile", in the truest sense of the word in that acoustics, sound generation, processing and capture are of interest to me. Not in the cretinous sense of paying exorbitant prices for hardware making no objective, quantifiable improvement to the reproduction of recorded audio :)

I don't, or have "golden ears" but i work in the industry. I have use for analogue and digital domains depending on what I am trying to achieve. Each have their own values but to say that analoge cannot give a true representaion of sound is a falacy as it is the analogue instruments that create the sound, and it could be argued, they are never captured "pure" to begin with.
As to the pish posh on silver cables and 500k amplifiers that must be tube amps, I completely agree with you. It's a scam. Just like oxygen free copper cables running the same speakers, sounding more pure. Utterly stupid.

Quote:

An example here https://youtu.be/qr4CvcSNHtM All sound is synthesised, processed and mixed entirely on the fly. Due to performance limitations (I mean it's a language intended for website backends) it's mono 48kHz with no EQ or other post processing.

That's verry cool man. I like that.

Last edited by SHADES on 21-Dec-2022 at 10:58 AM.
Last edited by SHADES on 21-Dec-2022 at 10:55 AM.
Last edited by SHADES on 21-Dec-2022 at 10:54 AM.

_________________
It's not the question that's the problem, it's the problem that's the question.

@SHADES

I did watch, but to reiterate, in the digital domain, the "early reflection" phase is handled slightly differently than the main decay of the reverberation which is usually done using a number of feedback channels going through comb filters. Actual bonafide reverb can only be done using something akin to ray tracing but for audio and still has to model the absorption characteristics of the medium (air), the surfaces of the space and so forth. While this produces the most correct results, let's be honest, it's utterly impractical.

Generally speaking, the more numeric headroom you have the better any audio processing algorithm can sound. 16 bit is fine for end output, but VST signals are typically floating point for a good reason (one of which is the fact that many algorithms make the most mathematical sense when the signal range is -1 to +1).

The main reason, I think, that reverb algorithms sound better at higher processing rates, other than the general statement above is actually due to the filter characteristics and the ability to multisample the poles. I'm skeptical that it has to do with any actual reflection modelling. Even at 48kHz, the shortest representable time is 1/48000 seconds which is about 2.08e-5 seconds. Taking the speed of sound in air to be 330m/s that's a distance of approximately 7mm. If you are modelling any realistic size space for your reflections, you are going to be dealing with at least 2m. If you are honest with yourself, so you think you could hear a difference in the impulse response of a room 2m versus one at 2.007m ?

Last edited by Karlos on 21-Dec-2022 at 12:31 PM.

_________________
Doing stupid things for fun...

So who is this moron that continues to post stupid polls about absolutely nothing worthwhile?
So clogging up the forum with useless polls is your new jam?

Dude shut up

Last edited by outlawal2 on 21-Dec-2022 at 04:16 PM.

@outlawal2

Or just use it as an excuse to talk about something more interesting.

_________________
Doing stupid things for fun...

at the the end of 2022 it is very sad read again: emulation or real.
There will be forever people that prefere the emulation and people that prefere real hardware. Is it realy necessary every time make these polls.

_________________

@Karlos

Quote:

has to model the absorption characteristics of the medium (air), the surfaces of the space and so forth.

Exactly, and no, it's not impractical at all. I use them.
Good reverb plugins do exactly this, across a variety of mediums like steel or aluminium plates, as well as room/padding mic placement, type, distance to speakers, types and sizes of speakers to mic ratios and cabinets Vs back free open-air (if that's the type of reverb used).
Even then, a true analogue plate (Very big floating metal plate with transducers attached in a box in an isolated chamber) is also at my disposal and used. It just has a unique sound, as all real analogue plates do. No two are alike. Room temperature, humidity etc all effect it.
Mind you, the room is controlled well with air-conditioning, it's not perfect though and when recording, you turn that stuff off.
Again, extremely hard to emulate/simulate that physical plate it in digital, and why it is still used.

We have one! Not cheap, big 3 person move, physically heavy type thing, with its own box.
However, again, analogue is very much used and still, very relevant, even in digital mixing.
Do, I still use tape? I wish I had one, but no.
Again, that analogue noise of a real "master tape" machine is, very hard to emulate, it's sound is sort after, expensive to buy and maintain.
I use an Ai plugin for it these days by Baby Audio. It's not perfect by any means, it has its own sound, but it's quite pleasant. Our hearing is very used to having this dither and noise in audio and I believe this is why music can sound more pleasing to us when applied correctly.
A debate for another time perhaps?

All high-res sources sent to it prior to mix down, so those impacts are in the correct effect digital domain, prior to mix-down to 44.1 or 48k.

As processing is applied to sound, it is not as extreme as it is in video however, yes, the processing is up there. Even on the modern CPUs, you try to use sends-> from tracks in to reverb busses, to limit how many instances of reverb you require. Mind you, there are cheap reverbs too, that use filtering and, they too have their place for certain sound requirements where you may want that cheaper sound, as a feature.

Quote:

If you are honest with yourself, so you think you could hear a difference in the impulse response of a room 2m versus one at 2.007m ?

Humid and warm Vs cold and bright rooms play havoc on a variety of instruments. Same with wind instruments. So, yeah, you can.
Anything over 3ms can put people out of feel, or even, put people into "feel" when recording. I often bounce a track +/- 3-5ms if the person is struggling to get into it, or can't figure out how to sit right.

Flipping samples/tracks 180 degrees can stop a phase cancellation, or shifting another instrument by only 3-6 degrees can get it to pop-out or hide in a mix. There are so, so many uses to small adjustments that interact with the entire overall compilation due to the way all these sounds interact with each other. Once it is all done, well yes, most of the time it's "filtered" resampled down to 44.1 / 48 unless, you're mastering to vinyl (send the HiRes masters to the printer) Most good vinyl printers still want the higher rate stuff. Lower noise floor to begin with.

Last edited by SHADES on 21-Dec-2022 at 09:44 PM.
Last edited by SHADES on 21-Dec-2022 at 09:38 PM.
Last edited by SHADES on 21-Dec-2022 at 09:03 PM.
Last edited by SHADES on 21-Dec-2022 at 08:58 PM.
Last edited by SHADES on 21-Dec-2022 at 08:53 PM.
Last edited by SHADES on 21-Dec-2022 at 08:47 PM.
Last edited by SHADES on 21-Dec-2022 at 08:45 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.

@Karlos

Quote:

Or just use it as an excuse to talk about something more interesting.

I love how you and others do this.
I really enjoy an interesting and even technical discussion. I find it fascinating and it's why I still come back here now and then.

Besides, I enjoy learning stuff, even if I can't always follow everything (coding) like you or cdimauiro, matthey and others, write.


Shhhhhh.....*wispers besides.......AMIGA, is still my favourite computer experience. ;)

_________________
It's not the question that's the problem, it's the problem that's the question.

@SHADES

We should spin up a thread for it, I could nerd out on this subject. I defer to your practical experience on using such computationally intensive effects. Certainly for the performance envelope I'm targeting for my uses, I'll be sticking to older school techniques :)

_________________
Doing stupid things for fun...

@Karlos

Quote:

Karlos wrote: @bhabbott Gold plated phono jacks compared to corroded connections? Come on, that's obvious. As for the rest it's absolute bloody snake oil. You think I was joking?

No, I don't think you were joking. We all know those ridiculously overpriced cables are a scam. But your blanket dismissal of 'the rest' is unscientific.

Quote:

The same company make those ludicrously thick cables you mention, only with "actively biased dialectic" with some 72v pack etc. There's literally no credible science to any of it, they are simply taking the piss out of people too stupid, but well heeled enough to to go along with it.

Like Volkswagen took the piss out of customers who believed their diesels were 'clean'? Or Microsoft telling us we must 'upgrade' to Windows 10 in order to be more 'secure'?

Quote:

Here they try to claim that "strand interaction" causes audio degradation. In a network cable linking up digital components.

There's plenty of snake oil in the computer industry too. But tell us that "strand interaction" can't cause signal degradation in a network cable and you will be laughed at. Sending a digital audio signal down an Ethernet cable in real time without degrading it is not easy. "But it's digital", you say "so nothing in the analog domain can affect it!". Wrong. The signal in the cable is analog, and anything that degrades it has the potential to cause errors in the digital domain. Each pair in UTP Ethernet cable has a unique twist designed to prevent interaction between the wires. Untwisting them for more than half an inch can cause sufficient degradation to fail a cable test.

But send the audio data via a standard protocol like TCP/IP or UDP and it gets much worse. Now you have a large randomly varying latency that needs big buffers to even out. Lose a packet and you have a massive dropout. Anyone who has used Skype knows what effect that can have, but more subtle effects can also be audible. Just because the signal is 'digital' doesn't mean it can't be degraded by eg. a cheap poorly manufactured patch cable. But even if the network transmission is 'perfect' there are still effects that could be noticeable - perhaps even problematic in come applications. Hopefully it isn't in the case you mentioned, but that would only be because it was carefully engineered to maintain audio signal integrity.

Quote:

The mentality at work here is the same one that if told they are using the real hardware but are in fact using a hidden, cycle exact emulation will praise the experience and if told they are using a cycle exact emulation, when in fact they are using real hardware will claim they could tell all along. Such is the subjectivity of user experience.

No. You claim that some Amiga fans have the same mentality as 'golden eared' audiophiles, but you offer no proof. And you ignore the wishes of those of us who want real hardware even if we can't tell the difference in a 'listening' test. What's worse is that by arguing there is no detectable difference between a theoretically 'perfect' emulation and real hardware, you are effectively invalidating our choice. This is like arguing that if the Mona Lisa could theoretically be copied 'perfectly', nobody should complain if the original is destroyed and a copy hung in its place.

@SHADES

Quote:

SHADES wrote: Quote: If you are honest with yourself, so you think you could hear a difference in the impulse response of a room 2m versus one at 2.007m ? Humid and warm Vs cold and bright rooms play havoc on a variety of instruments. Same with wind instruments. So, yeah, you can. Anything over 3ms can put people out of feel, or even, put people into "feel" when recording. I often bounce a track +/- 3-5ms if the person is struggling to get into it, or can't figure out how to sit right.

The human ear might not be very good at determining absolute amplitude or phase, but it is very sensitive to relative differences. And our ears aren't shaped that way for nothing. The brain takes in the signals from both ears in real time and correlates their phase and amplitude to determine where a sound is coming from. Anything that disturbs that can affect the sound.

In a room with reflective walls there's a reasonable chance you could tell the difference if it was varied between 2m and 2.007m. With sounds bouncing all over the room there would be multiple cancellation points a few mm apart at audible frequencies. You probably wouldn't notice if the room stayed one size, but vary it abruptly and the effect could be very obvious.

@bhabbott

Replying via phone so expect nonsensical word corrections...

"Like Volkswagen took the piss out of customers who believed their diesels were 'clean'? Or Microsoft telling us we must 'upgrade' to Windows 10 in order to be more 'secure'?"

I don't think these are comparable. Or they are poor comparisons at best. For one thing, the "cleanliness" of diesel emissions is something that can be objectively measured and it's in part how they were caught out in the first place. The attitude behind it can be compared to some extent. However, say VW charged 50x more for their diesels compared to anyone else while making the claim that they were clean, get 5x the milage, improves your suspension and allows you to see further in fog even without your lights on. Then you get the likeness of the audiophile cable manufacturers. As for MS. Well the claim that 10 is more secure than 7 is quantifiably true given 10 is still supported and 7 is not. Of course it isn't the only way, you could choose another OS, or move wholesale to a different platform. It all depends on how technically proficient you are and how amenable to change, whether it's worth your effort etc.

""But it's digital", you say "so nothing in the analog domain can affect it!". Wrong. The signal in the cable is analog, and anything that degrades it has the potential to cause errors in the digital domain."

Well sure that's true at least in theory. As I said when talking about the cable in the first place the reason it uses twisted pairs is that they carry the same signal at mirror levels so that analogue interference induced in the cable can be mitigated. But lest be realistic, cat-5 ethernet is designed for the reliable transmission of digital data up to 100Mbps over many metres. Were not talking about Skype here, with countless hops over the internet with a heavily compressed realtime stream. 20 years ago I worked in VoiP and related applications. All your talk about packet loss, buffering and latency is horseshit in domain we are discussing: a few metres of network cable between a DAC and a storage device/player?

The highest data rate, let's say 192kHz 24 bit 5 channel, uncompressed. What's that? 5x24x192000 = ~23Mbps. Thats 23 % of the total bandwidth. Let's add some TCP packet overhead to that. Based on a typical MTU of 1500 bytes and then the associated network cruft you increase your data requirements a few percent. You still have almost three times the remaining spare capacity than you are using in order to retry any failed packets.

About latency. Yes this will need buffering as do all packet switched networks when dealing with stream data but unless you are synchronising with something else, we are talking about the original audio stream from storage reaching your DAC. You aren't going to notice if there's any latency because what are you comparing it to? Unless the DAC are embedded in the speakers themselves and the speakers are networked and receiving their own streams that run the risk of desynchronisation. That doesn't seem to be the typical arrangement.

Back to reality, most of the time the data being transmitted over your network cable to your DAC will be much lower bandwidth than our extreme example above, and potentially compressed on top of that.

You still want to claim anything beyond your most basic DIY crimped cat 5 cable has a place here? If you do, I know some sellers of the highest quality that are itching to rob you senseless.

Last edited by Karlos on 23-Dec-2022 at 09:46 AM.
Last edited by Karlos on 23-Dec-2022 at 09:42 AM.

_________________
Doing stupid things for fun...

@bhabbott

Quote:

And you ignore the wishes of those of us who want real hardware even if we can't tell the difference in a 'listening' test

I'm sorry, you'll need to clarify that. What wishes am ignoring by suggesting that in a test you would be fooled by a near perfect emulation of something? I'm not telling you to go away and abandon your love of the real thing or to use emulation from now on. I don't care about your subjectivity and I have no wish to ruin your hobby. However when people decry emulation for simply being emulation and that it can never be (subjectively) as good, then there are objective ways to test that subjectivity, e.g. this hypothetical listening test.

I use emulation and can think of many objective benefits and drawbacks over the actual hardware I own that's currently decommissioned. Is it better? That depends on what I'm doing. It's better for development, for sure. Being able to quickly spin up all kinds of different configurations, no risk of damage to difficult to replace hardware while working on something that requires a lot of resetting, in situ debugging, being able to use modern development tools at the same time. Once something has reached a certain degree of maturation and stability it can be tested on the real kit.



Quote:

The human ear might not be very good at determining absolute amplitude or phase, but it is very sensitive to relative differences. And our ears aren't shaped that way for nothing. The brain takes in the signals from both ears in real time and correlates their phase and amplitude to determine where a sound is coming from. Anything that disturbs that can affect the sound. In a room with reflective walls there's a reasonable chance you could tell the difference if it was varied between 2m and 2.007m. With sounds bouncing all over the room there would be multiple cancellation points a few mm apart at audible frequencies. You probably wouldn't notice if the room stayed one size, but vary it abruptly and the effect could be very obvious.

I started a separate thread for this thing but we can get into it regardless. To recap, the ~7mm difference is based on an estimation for how far sound, at standard ~330m/s will propagate in air in the duration of one sample period at 48kHz. Its basically 6.875mm if you care about the full number of digits. In terms of the immediate impact on reflection then, then, it's double that distance since we travel an extra 6.875mm to reach the reflector and an the same additional distance back.

Disclaimer - we are discussing the impact on a single reflection. Of course the total reverberation is the sum of countless additional reflections and it is absolutely undeniable that the overall outcome will be different. The question is, how much that difference is perceptible.

There are two factors at play due to that additional distance. First, we attenuate the signal slightly more in accordance with the inverse square law, and secondly, because of the wave nature of the sound, the return phase will be adjusted too since the distance will be some relative fraction of the wavelength at a given pitch.

The attenuation difference is easy. 1 / (2.006875/2)^2 is 0.9932 The initial echo will be 99. 32% as loud. Humans perceive loudness logarithmically so that difference is basically inaudible, golden ears or not.

As for phase, well it depends on the frequency of the note. The higher the frequency, the shorter the wavelength of the sound and thus the larger the difference in phase will be as a consequence of the additional distance.

The wavelength of a 440Hz (A4, concert pitch) in air at 330m/s is 330/440 = 0.75m. So, the additional distance (2x6.875m) in the slightly larger room is, in duty cycles, 0.006875x2 / 0.75 = 0.018333...

Is that phase difference perceptible? For 440Hz, I strongly doubt it. I mean you can try it out - use Audacity or any similar tool to create a 440Hz sine wave at 48kHz sample rate and see if you can tell any difference to the perceived sound as a function of where you start playback from (aside from the initial click if you start somewhere far from the zero line). For reference, we're talking about at 2 sample difference since the original distance was based on how far sound propagates in 1/48,000th of a second and we are interested in the effect of the total there and back distance (hence 2x). There are about 109 samples per full duty cycle in a 48kHz, 2/109 = 0.01834 - just to sanity check we didn't make a mistake in any of this conversion.

To save you the effort, no, you (almost certainly) won't notice - principally because we aren't particularly sensitive to phase in that way (we are sensitive to the secondary effects it can have). For example, two sine waves at the same frequency played in each ear at different phase offsets can sound off centre at first, but your brain pretty quickly corrects for it and soon you perceive a mono signal.

For higher frequencies, though, we could find one where the wavelength is exactly 2x the additional return path distance and get a near total cancellation. You can rearrange the formula for wavelength easily enough and plumb in 4x6.875mm as the wavelength and it comes out as 12000Hz (quelle surprise, we were dealing with the sound propagation distance in 1/48,000th of a second).

I would expect then that this distance would affect the attenuation of higher frequencies slightly, but as to what difference it would make to the listener, that's hard to say. However, to put it all in perspective, air temperature has a bigger impact. Ignoring the effects it has on the actual instrument itself (let's assume we are using a perfect speaker as an emitter), at 1bar, room temperature the speed of sound is about 343m/s. This has a greater influence on the phase difference of a reflection in a room of fixed size since it literally changes the wavelength of the sound from 0.75m for 440Hz at 0C to 0.77955 at 20C. That's 29.5mm longer.

Shades mentions the impact of differences measured in milliseconds putting musicians off their feel. Well, that is different. at 48kHz, 1ms of audio is 48 samples. Phase delay effects are generally measured in ms.

As for timing, I know from percussionists that a few ms can be a problem, especially things like sequencer jitter. However, I asked one to estimate the jitter in this (no apologies for the quality of the music, it's being synthesizer in realtime by php) https://www.youtube.com/watch?v=vVjLyoBt-bg

He said their clearly wasn't any. This isn't true because the smallest event interval in the above sequencer code is 1 packet. 1 packet is 128 samples, which at the 48kHz rate is 2.67ms. No note event can be placed more precisely than that.

Last edited by Karlos on 23-Dec-2022 at 11:31 AM.
Last edited by Karlos on 23-Dec-2022 at 11:30 AM.
Last edited by Karlos on 23-Dec-2022 at 11:22 AM.
Last edited by Karlos on 23-Dec-2022 at 10:13 AM.

_________________
Doing stupid things for fun...

I get that emulation is not for everyone, but it's one of the big reasons I'm still part of the Amiga scene and is my main experience of Amiga use now, I think it the fact of convenience and how good emulation is that I keep going to it. My Pi400 is easy to setup any where i want it to, and with Pre-config packs like Amikit and Pimiga it really is easy to get a really good solid setup. I also few years ago pulled my original Amiga setup I grew up with over to my PC/winuae and have OS3.9/3.2 setup on my PC too. I don't remember the last time I couldn't run any classic Amiga software on emulation on my PC. The Pi JIT emulation does have a few issues (I hope they can improve this) but emulation is great. OS4 i haven't bothered to setup on emulation yet, I dont think it would match my x1000 anyway. But if it ever did, I would likely make the switch there too.

Last edited by amigang on 23-Dec-2022 at 12:23 PM.

_________________
AmigaNG, YouTube, LeaveReality Studio

@bhabbott

Quote:

In a room with reflective walls there's a reasonable chance you could tell the difference if it was varied between 2m and 2.007m. With sounds bouncing all over the room there would be multiple cancellation points a few mm apart at audible frequencies. You probably wouldn't notice if the room stayed one size, but vary it abruptly and the effect could be very obvious.

The point i was trying to make on this part was in how frequencies mixed together interact with each other.
Like ingredients in a soup. A slight change in reflections can cause other waveforms to have a broad effect on how the rest, sound mixing together, depending on it's amplitude amount (energy)
The way sound, collides and mixes.

I very often invert phases on microphones for drums as an example.
The top and bottom of a snare are out of phase by 180 degrees (a very small time shift in reality, required by inverting a signal - half of that waveform's transition through time) 180 degrees in phase will cancel out inverse similar frequencies, however, generally the skins are tuned differently as well and use different skins. All of this needs shifting in time as well to get a strong sound when summed or added. Then when mixed in with other parts, may also need further overall adjustment. That's without even touching EQ or effects.
Reflections can have huge consequences in time, even in slight adjustments that can create unwanted harmonic effect by summing and making a resident, dominant and audible unwanted frequency.
Understanding that the collision of different, yet similar waveforms can and do create their own frequencies that sometimes are unwanted. It's not just a simple sign-wave tone environment. Those "generated" harmonics, can matter. A lot.

Last edited by SHADES on 23-Dec-2022 at 10:42 PM.
Last edited by SHADES on 23-Dec-2022 at 10:37 PM.
Last edited by SHADES on 23-Dec-2022 at 10:35 PM.
Last edited by SHADES on 23-Dec-2022 at 10:20 PM.
Last edited by SHADES on 23-Dec-2022 at 10:15 PM.
Last edited by SHADES on 23-Dec-2022 at 10:15 PM.
Last edited by SHADES on 23-Dec-2022 at 10:04 PM.
Last edited by SHADES on 23-Dec-2022 at 10:02 PM.

_________________
It's not the question that's the problem, it's the problem that's the question.