home studios equipment staff/friends booking/rates for sale forum contact

16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not be

Moderators: kerble, Electrical-Staff

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby total_douche on Fri Nov 24, 2017 1:16 am

The extra data points only really serve to change the Nyquist frequency. The time between samples is several orders of magnitude less than what the ear can perceive. To be more specific, with respect to your example, a millisecond is .001 seconds. The length of an individual sample at 44.1Khz is .0000227 seconds (1/44100). The shortest delay the ear can detect (according to the great Google) is .1 milliseconds. That's 4.4 times the length of a single sample at 44.1KHz (.0001/.0000227). It's just not going to make a difference.
User avatar
total_douche
"kick a ginger day" victim
"kick a ginger day" victim
 
Posts: 1244
Joined: Mon Dec 29, 2008 10:11 pm
Location: Hastings, MN

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Fri Nov 24, 2017 1:22 am

Re: Time based effects. I remember a few years back using the inputs from two separate audio interfaces w/ no clocking on the same source. Utterly unusable, massive amounts of phasing & digital weirdness.

I know wow is unavoidable with tape, but it doesn't seem to create the same oddness that bad clocking with digital recording does. And that's not the above scenario, merely using one poorly designed interface will result in the audio's transient response sounding noticeably 'off'.

A few rounds of AD/DA will also mush & slew your shit up no end. You're sampling, resampling at different discrete points in time including in-between previous sample points & it sounds shit.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Fri Nov 24, 2017 1:28 am

total_douche wrote:The shortest delay the ear can detect (according to the great Google) is .1 milliseconds. That's 4.4 times the length of a single sample at 44.1KHz (.0001/.0000227). It's just not going to make a difference.


Try sending an audio track to a bus and add a couple of samples delay. Notice anything?
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby Anthony Flack on Fri Nov 24, 2017 1:34 am

Boombats wrote:Anthony.FLAC


Is it boring? I'm sorry if it is, but I find the whole maths side of this quite interesting and it always provokes controversy it seems.

24K wrote:No, not unless you are using specialist mic's & recording actual ultrasonic content. It's not going to magically appear in a recording with none.


Sorry, I guess a better way of putting it is, if you are pitching down, any ultrasonic content that may or not be present at the higher sample rate might become apparent. If you don't, it doesn't.

There's loads of difference. Pitch down an octave & you've halved the transient response of the recorded material. You're going from 44100 discrete data points per second to 22050.

I think it may be the extra data points that make higher resolution audio, sharper & more focused not necessarily the extra frequency content.


Just to be clear, here I'm talking about the opposite - not pitching the sound down but doing a sample rate conversion while keeping the pitch the same.

And what is interesting is, as long as there is no frequencies above the Nyquist frequency the wave is reconstituted exactly the same. The timing of the transients is not affected because the wave is sampled at different points.

The wave shape is not stair-stepped between the samples and it doesn't move in straight zig-zag lines between samples. It is a curve that passes through all the sampled points. The fact that it is frequency limited is the key - by definition the curve cannot change direction any more sharply than a sine wave at whatever is the highest frequency.

There is only one possible curve that obeys this rule and passes through all the sampled points. As long as it's safely below the Nyquist limit, the lower sample rate and higher sample rate describe exactly the same curve. That's why it can be reconstructed at any resolution.

A few rounds of AD/DA will also mush & slew your shit up no end. You're sampling, resampling at different discrete points in time including in-between previous sample points & it sounds shit.


I think the crucial aspect here is that you're going into analogue each time, passing through amplifiers, adding noise and who knows what, there's impedence in every part of the chain and so on. As I say above, extracting accurate sample points in-between other sample points is no problem as a purely mathematical digital conversion.
Anthony Flack
Greatest Man Who Ever Lived
Greatest Man Who Ever Lived
 
Posts: 8626
Joined: Fri Aug 03, 2012 12:27 am
Location: New Zealand

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby regular username on Fri Nov 24, 2017 2:51 am

24K wrote:
total_douche wrote:The shortest delay the ear can detect (according to the great Google) is .1 milliseconds. That's 4.4 times the length of a single sample at 44.1KHz (.0001/.0000227). It's just not going to make a difference.


Try sending an audio track to a bus and add a couple of samples delay. Notice anything?

you mean phase cancellations?
User avatar
regular username
seminarian
seminarian
 
Posts: 71
Joined: Fri Mar 15, 2013 2:03 am
Location: west-germany

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Fri Nov 24, 2017 10:11 am

regular username wrote:you mean phase cancellations?


Yes, "The length of an individual sample at 44.1Khz is .0000227 seconds (1/44100)" Ok, add 1 solitary sample. It's just fucked your shit up, loss of HF, a veritable mess.

No real world conversion process is mathematically perfect. It'll waver - drop samples. Analog doesn't do this. Sure, slower IPS will result in a loss of HF content but your still capturing real time.

Good conversion with a solid clock will get decent results. Providing you're sending it a good signal & not going rounds.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby regular username on Fri Nov 24, 2017 1:17 pm

24K wrote:
regular username wrote:you mean phase cancellations?


Yes, "The length of an individual sample at 44.1Khz is .0000227 seconds (1/44100)" Ok, add 1 solitary sample. It's just fucked your shit up, loss of HF, a veritable mess.

No real world conversion process is mathematically perfect. It'll waver - drop samples. Analog doesn't do this. Sure, slower IPS will result in a loss of HF content but your still capturing real time.

Good conversion with a solid clock will get decent results. Providing you're sending it a good signal & not going rounds.

assumed you were referring to a mix of almost identical signals, where a very short delay on one track actually leads to phase cancellations/comb filtering. if you're talking about a short delay on a single solo track i don't see how that would result in HF loss
User avatar
regular username
seminarian
seminarian
 
Posts: 71
Joined: Fri Mar 15, 2013 2:03 am
Location: west-germany

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Fri Nov 24, 2017 3:04 pm

regular username wrote:assumed you were referring to a mix of almost identical signals, where a very short delay on one track actually leads to phase cancellations/comb filtering. if you're talking about a short delay on a single solo track i don't see how that would result in HF loss

I'm talking about one channel, load say a full stereo mix (44.1k) onto a stereo channel. Send this 100% to a send, delay the send by 1 sample. No effects, nothing - just the identical audio delayed by one sample. Listen.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby regular username on Fri Nov 24, 2017 3:42 pm

interesting! do you happen to have any measurements ready, maybe a frequency response or something?
User avatar
regular username
seminarian
seminarian
 
Posts: 71
Joined: Fri Mar 15, 2013 2:03 am
Location: west-germany

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby projectMalamute on Fri Nov 24, 2017 5:24 pm

24K wrote:I'm talking about one channel, load say a full stereo mix (44.1k) onto a stereo channel. Send this 100% to a send, delay the send by 1 sample. No effects, nothing - just the identical audio delayed by one sample. Listen.


I just made a little video to demonstrate this:

phpBB [media]


You can here the difference pretty clearly, even with just this iPad mic recording.
bandcamp
youtube

Endeavor to persevere.
User avatar
projectMalamute
Supreme Commander at the Forefront
Supreme Commander at the Forefront
 
Posts: 1963
Joined: Sun Nov 18, 2007 10:59 am
Location: Westfield MA

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby Anthony Flack on Fri Nov 24, 2017 5:40 pm

24K wrote:I'm talking about one channel, load say a full stereo mix (44.1k) onto a stereo channel. Send this 100% to a send, delay the send by 1 sample. No effects, nothing - just the identical audio delayed by one sample. Listen.


Say what now? You're saying playing a single stereo track, isolated, with a one sample delay sounds different than without it? I don't know what's going on in yer DAW but that should not be. Or are you playing it alongside the original?

I know what you're saying about transients. Surely, if sample rate A is twice that of sample rate B, then you can catch the transients coming in half a sample sooner at rate A than you do with rate B? Because it's polling more often. That makes intuitive sense yeah? Let me see if I can explain more clearly why that isn't so, providing the frequency limit is observed.

Let's say sample rate A is capturing however many samples per second and sample rate B is exactly half that and only capturing every second sample. The signal we're capturing is comfortably under the Nyquist limit of both.

We start with a bunch of silence - samples all record 0s for a while.

Then the transient hits. Sample rate A catches it and records 0.4. Sample rate B misses it.

Next sample. Sample rate A catches the transient at 0.7 as it continues to rise. Sample rate B also catches that one at 0.7.

Now, when it comes to reconstructing the wave from sample set B, it doesn't jump straight from 0 to 0.7. It's a curve that passes through both points, and what you will find is that in order to pass through the 0 and the 0.7, the curve must also pass though the 0.4 position exactly where the missing sample is. So the transient is captured exactly the same, in the same place.

What about if sample rate A captured the rising transient at 0.4, but then the transient started to fall again by the time the next sample is taken? Sample rate B will not be able to reconstruct this; the bump will be smoothed off. But that can only happen if you break the Nyquist limit of sample rate B - by definition, it's high frequency content oscillating faster than half the sample rate which is being cut off.
Anthony Flack
Greatest Man Who Ever Lived
Greatest Man Who Ever Lived
 
Posts: 8626
Joined: Fri Aug 03, 2012 12:27 am
Location: New Zealand

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Fri Nov 24, 2017 7:13 pm

Anthony Flack wrote:
Say what now? You're saying playing a single stereo track, isolated, with a one sample delay sounds different than without it? I don't know what's going on in yer DAW but that should not be. Or are you playing it alongside the original?



Yes, I purposely said route to a send to illustrate this. Watch the video above, especially the part where the phase is inverted. At 1 sample its quite clear that you have phase fuckery & are losing HF content.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby regular username on Fri Nov 24, 2017 7:19 pm

projectMalamute wrote:
24K wrote:I'm talking about one channel, load say a full stereo mix (44.1k) onto a stereo channel. Send this 100% to a send, delay the send by 1 sample. No effects, nothing - just the identical audio delayed by one sample. Listen.


I just made a little video to demonstrate this:

phpBB [media]


You can here the difference pretty clearly, even with just this iPad mic recording.

that's the phase cancellation i was talking about; you'd get the same result with delaying an analog signal by 20 microseconds or so
Last edited by regular username on Fri Nov 24, 2017 8:15 pm, edited 1 time in total.
User avatar
regular username
seminarian
seminarian
 
Posts: 71
Joined: Fri Mar 15, 2013 2:03 am
Location: west-germany

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby Anthony Flack on Fri Nov 24, 2017 7:26 pm

Ok, so yeah, delaying and recombining the time-shifted signal with the original will result in comb filtering.

But a sample rate conversion will not time-shift the signal.
Anthony Flack
Greatest Man Who Ever Lived
Greatest Man Who Ever Lived
 
Posts: 8626
Joined: Fri Aug 03, 2012 12:27 am
Location: New Zealand

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby total_douche on Fri Nov 24, 2017 8:05 pm

24K wrote:Yes, I purposely said route to a send to illustrate this. Watch the video above, especially the part where the phase is inverted. At 1 sample its quite clear that you have phase fuckery & are losing HF content.

You were talking about latency, and I responded to that. Let's not pretend that 1/44100 of a second is going to make a difference here. That still doesn't change the fact that sample rate only really determines what information can be carried. That you get destructive interference with very small changes in a shifted waveform mixed onto itself is completely irrelevant here.

It's just like the ongoing "debate" that Gibson "collectors" have about ceramic and PIO caps. The sum total of everything else is probably going to make a bigger difference in sound than sample rates for most music, provided you're at least using 16-bit 44KHz audio. Yeah, I get that higher bitrates have a greater dynamic range, but the dynamic range in a 16-bit recording is already the difference between complete silence and turbofan exhaust. Similarly, I could probably negate every hypothetical advantage to using PIO caps by changing my picking position by a few millimeters.
User avatar
total_douche
"kick a ginger day" victim
"kick a ginger day" victim
 
Posts: 1244
Joined: Mon Dec 29, 2008 10:11 pm
Location: Hastings, MN

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Sat Nov 25, 2017 5:47 am

24K wrote: Even a few milliseconds latency can be off putting if playing / monitoring through a DAW. Plus, if using plugins even a few samples can cause noticeable phasing issues if the delay compensation is incorrect.


If you've somehow managed to get your roundtrip latency down to a solitary sample then I doff my cap. The closest I've ever been able to achieve is in the milliseconds realm. And it is noticeable.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Sat Nov 25, 2017 5:56 am

Anthony Flack wrote:But a sample rate conversion will not time-shift the signal.


No, it's not going to move the signal, but the anti-aliasing filter will effect it. We're dealing with the real world & there's always going to be trade offs (foldback, pre-ringing etc). It would be nigh on impossible to have a perfect sample rate conversion, so designers have to implement workarounds.
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby projectMalamute on Sat Nov 25, 2017 12:07 pm

I don't buy it. I play bass through my computer all the time, it's how I practice. My system usually runs at 5.8ms round trip. That's 128 samples each way at 44.1K.

The sound gets to my headphones in about the same amount of time if would take to travel from a floor wedge to my ears through the air. Doesn't bother me at all.

You are claiming a few samples is noticeable, the equivalent of moving the amplifier less than an inch. Best I could do for you is 16 samples, the equivalent of moving your amplifier about 5 inches further away. I'll bet all the money in my wallet that if we got the levels matched and A/B'd the signal direct through the mixer with the signal through the computer you wouldn't be able to tell which had the delay.

Do you perform live with your head stuck in your speaker cabinet because the delay of standing 5 feet away is a problem?
bandcamp
youtube

Endeavor to persevere.
User avatar
projectMalamute
Supreme Commander at the Forefront
Supreme Commander at the Forefront
 
Posts: 1963
Joined: Sun Nov 18, 2007 10:59 am
Location: Westfield MA

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby 24K on Sat Nov 25, 2017 12:44 pm

projectMalamute wrote:Do you perform live with your head stuck in your speaker cabinet


Nope, I strap a wedge to each shoulder & sit on a bassbin (careful to avoid the brown note).

At home through a DAW I get a round trip of 9ms. I should probably upgrade my interface but I put all spare cash into things for work. Mic's pre's etc. I can record at my work place - real amp, not monitoring through a DAW.

When I haven't plugged into an amp for a good while I get a shock & have to adjust. I've developed a bad habit where I'm playing slightly ahead to compensate for DAW practise/recording.

How much do you have in your wallet? :)
User avatar
24K
mayor
mayor
 
Posts: 99
Joined: Wed Aug 22, 2012 1:39 pm

Re: 16 bit 44.1 kHz is just fine and 24 bit 192 kHz may not

Postby projectMalamute on Sat Nov 25, 2017 12:56 pm

Sure, but 9 ms is a lot more than 'a few samples'. At 44.1K that's about 400 samples, at 96K it's almost 900 samples. I find that's the range where you are just approaching noticeable. Like a weird stage plot where your amp is farther away than you'd like.
bandcamp
youtube

Endeavor to persevere.
User avatar
projectMalamute
Supreme Commander at the Forefront
Supreme Commander at the Forefront
 
Posts: 1963
Joined: Sun Nov 18, 2007 10:59 am
Location: Westfield MA

PreviousNext

Return to General Discussion

Who is online

Users browsing this forum: Google [Bot] and 12 guests

cron