What is the base sampling rate?

Sampling rate affects only one thing: the highest frequency that can be accurately reproduced. That's it.

With modern anti-aliasing filters, 48 KHz sampling can accurately reproduce just a little over 20 KHz, which is the highest frequency that the best human eats can hear. Anything above that is inaudible, which makes it a waste of resources in an audio system.

Well, somewhat. They can produce a mangled 22k signal... and I would argue a mangled 12.25k signal... and so on. Go down into the real concepts. They have been feeding us the lie that 44.1k can do up to 22k.... well, it just isn't true in the real world.
 
Yes to the extent that if your buffer size is 256 samples a higher sample rate will reduce the latency time.

But I am pretty sure that the processing latency in all Fractal products is exactly 0. And that means no oversampling.

Latency is not 0. There are buffers. That is how PCM digital audio works. See you interface in your computer. But it is very low and yes, if you wanted to run 96kHz it would require more processing power... but technically the latency is lower. There is, of course, a tradeoff.
 
They have been feeding us the lie that 44.1k can do up to 22k.... well, it just isn't true in the real world.
Who has?
And yes, that's why 48Khz provides a safety buffer due to the filtering necessary - and again, this is why using a single rate and optimising your filters for it (rather than a changeable rate AT THE HARDWARE LEVEL) is the optimum solution.

There is no content above the real cut-off of a good 48Khz conversion that you can hear. Or feel.

Ever.

Again, A/B/X double-blind testing will always confirm this.

You can hear a difference between 96K and 48K, but someone could hear a difference between 384K and 768K. That's the wonder of placebo - the difference is real to the person experiencing it.

It's just not repeatable in a clinical setting because it's not objectively real.
 
I'm 28 years old, and I don't here any differences between 48 kHz and 96 kHz processed material. But I do remember something about filter slopes getting worse near the Nyquist frequency.
If the filter crowds the sampling frequency too closely, you could trim the top end a bit. If the filter is too steep, you can get overshoot. A 48 KHz sample rate gives you enough room to design a filter that does justice to the high end.

Higher sampling rates only add information in the ultrasonic spectrum. For example, a 96 KHz sampling rate lets you capture signal up to 40 KHz. In audio signals, there is only noise up there. That noise is inaudible by itself (it's ultrasonic, after all), but it can intermodulate with the audio to add distortion and noise that falls within the audio spectrum. Small amounts of that distortion can be perceived as fullness or liveliness, much like an overdriven channel strip can make things sound thicker. But it happens at the expense of accuracy.

If I'm going to add distortion to make things sound thicker, I want to do that in the modeling software itself, where I can have control over it. By the time I hit the converters, all I care about is accuracy.
 
I will say, the only people that say there is no difference are people that have not spent a little bit of time listening to it. Forget the B.S. arguments. Just listen. I have > 600 albums in hi-res. Can I still listen to others? sure. Of course I could still have my old Stella guitar instead of a Martin too... same thing, right? they both have wood and six-strings. 192kHz is like the difference of a cassette tape at 1-7/8 ips or a reel-to-reel at 15 ips. There absolutely was a difference then and there is one now. I have yet to not prove it even to the most skeptical if using decent equipment. The biggest proof was when I first recorded a drumset at 176k. I could not believe how awesome it sounded. It was astounding. Every nuance was clear and present.

Every little detail matters when you are wanting to do the best job, IMHO. Why spend money and then only get 80-85%?

I hate to turn this into a sampling rate debate but every CD is only 44.1 anyway. If 48k is good enough for an axe fx then I just can not believe the difference is audible. There is no way the human ear can hear that.
 
So let me get this straight.

We are busting Fractal's chops because they went with 48 KHz, 24 bit sampling rate?

That's "better" than Line 6 Helix 48 KHz, 16 bit rate.

That's "better" than Kemper's 44.1 KHz, 16 bit rate.

Headrush seems to get a little cagey with sample rates as their driver seems to upconvert or down convert on the fly, but it would seem to be no better than 48 KHz, 24 bit, based upon their recommendation of using 1024 sample size ir files for optimal processor performance. .

So why would we want to burn up processor power to run everything at 96 KHz 32 bit through the entire signal chain when, barring DVD audio & SACD, most consumer formats are 44.1 KHz 16 bit or worse if you are talking streaming or mp3?

Not to mention most of the classic DDL stomps and rack reverbs were not remotely at 96 KHz.

If you are that concerned about archiving at the highest possible quality, then run analog out into your apogee or UAD hardware or analog tape machine. Then you can always be on the cutting edge of resolution that you desire.
 
Last edited:
So let me get this straight.

We are busting Fractal's chops because they went with 48 KHz, 24 bit sampling rate?

That's "better" than Line 6 Helix 48 KHz, 16 bit rate.

That's "better" than Kemper's 44.1 KHz, 16 bit rate.

Headrush seems to get a little cagey with sample rates as their driver seems to upconvert or down convert on the fly, but it would seem to be no better than 48 KHz, 24 bit, based upon their recommendation of using 1024 sample size ir files for optimal processor performance. .

So why would we want to burn up processor power to run everything at 96 KHz 32 bit through the entire signal chain when, barring DVD audio & SACD, most consumer formats are 44.1 KHz 16 bit or worse if you are talking streaming or mp3?

Not to mention most of the classic DDL stomps and rack reverbs were not remotely at 96 KHz.

If you are that concerned about archiving at the highest possible quality, then run analog out into your apogee or UAD hardware or analog tape machine. Then you can always be on the cutting edge of resolution that you desire.


.....what is this , reasoned and logical thought?!? we are here for wringing of hands and gnashing of teeth......
 
The fact is that if in my DAW I'm working @96/192kHz I will be forced to perform many AD/DA conversion or resample the 48kHz digital signal from the AxeFX...why don't let the user to choose if use an higher sampling rate (reducing the possibility/instances of effects) if needed? I have an Eventide Eclipse and an Orville and I have that possibility (implemented in that units years ago)...I would have this kind of freedom with the AxeFXIII
 
.....what is this , reasoned and logical thought?!? we are here for wringing of hands and gnashing of teeth......
LoL... also a belated festivus airing of the grievances.

Truly, at the sample rates that people are lamenting about in this thread, the half sample rate harmonics, if they even exist outside of NASA level instrumentation, would be within the dB range of most coveted stomp boxes or tube amp noise floors, so they would be effectively masked anyways.

If anyone disputes this, show me the fft of your 12th fret harmonics through any amp sim recorded at 88.2, 96, or 192. After the initial attack, let's see how quickly the amp sim noise floor eats it up.
 
The fact is that if in my DAW I'm working @96/192kHz I will be forced to perform many AD/DA conversion or resample the 48kHz digital signal from the AxeFX...why don't let the user to choose if use an higher sampling rate (reducing the possibility/instances of effects) if needed? I have an Eventide Eclipse and an Orville and I have that possibility (implemented in that units years ago)...I would have this kind of freedom with the AxeFXIII
One extra ad/da conversion is not going to destroy the integrity of your signal.

In the days of analog tape that many audiophiles revere, every time you used multiple ddl, pitch, or digital reverb units in tracking or mixdown, you had way more ad/ad conversions, at lesser sampling rates, using much lower quality converters than we are talking today, and we made some great recordings. The clock jitter and dc offset on some of those units would make a spec junkie of today cringe, but the recordings stand the test of time for audiences.
 
Last edited:
I'm one of those that clearly hears a difference between a 96/24 recording and the CD version, though I don't know what's the reason and if it has anything to do with sample and bit rates.
Said that, I'm ok with 48k but I can see the "convenience" argument.
Furthermore I'm not sure that the CPU usage would increase as much in an axe fx, we know that the amp block for sure and probably most effects already use higher sampling rates internally, maybe all blocks that increase dramatically harmonic content (drive block) or require precise micro delays (flanger, cab block mic distance, maybe even chorus, pitch, etc..)
 
I'm one of those that clearly hears a difference between a 96/24 recording and the CD version, though I don't know what's the reason and if it has anything to do with sample and bit rates.
Said that, I'm ok with 48k but I can see the "convenience" argument.
Furthermore I'm not sure that the CPU usage would increase as much in an axe fx, we know that the amp block for sure and probably most effects already use higher sampling rates internally, maybe all blocks that increase dramatically harmonic content (drive block) or require precise micro delays (flanger, cab block mic distance, maybe even chorus, pitch, etc..)

If you want to see the difference between sample rate CPU usage, create your usual preset block chain in your DAW and record at 96KHz or 192kHz while having your cpu/memory meters on your computer open. Now run the same chain recording at 48KHz. You will see just how much more taxing on the processor & memory usage it is.

If you want to see what the net gain of such a recording effort is, blindfold someone else, play both recordings, and ask them to pick the clearer sounding one. Most experiments show the average listener is about as successful doing so as a coin flip would be in determining which is the higher sample rate.

Regarding your idea of using different sample rates for different blocks, then you are introducing more ad/da conversion and cpu load using that paradigm, which is what other posters said they were trying to avoid when it was suggested they use analog outs for other sample rate recording.
 
Last edited:
I have yet to see/hear any credible mastering engineer state that they can reliable tell the difference between 48 kHz and 96 kHz. I think, quite a few of them might be able to hear the difference between 16 bit and 24 bit audio. At least, the mastering engineer I know (Holger Lagerfeldt, http://www.onlinemastering.dk/references.html) got a 100% score in a four-part blind test (that I personally failed completely). Holger is very open about not being able to tell the difference between 48k and 96k.
 
If you want to see the difference between sample rate CPU usage, create your usual preset block chain in your DAW and record at 96KHz or 192kHz while having your cpu/memory meters on your computer open. Now run the same chain recording at 48KHz. You will see just how much more taxing on the processor it is.

...

Regarding your idea of using different sample rates for different blocks, then you are introducing more ad/da conversion and cpu load using that paradigm.
Yes I know that changing the sample rate a plugin works changes the cpu usage accordingly.
I'm saying that the axe fx already works like that, it's not my idea. Cliff himself stated the amp blocks uses oversampling and a bit depth of 60 bit in the axe fx 3. I just assumed that probably other blocks work like that too and that most of the "tax" is already charged on the axe fx cpu.

If you want to see what the net gain of such a recording effort is, blindfold someone else, play both recordings, and ask them to pick the clearer sounding one. Most experiments show the average listener is about as successful doing so as a coin flip would be in determining which is the higher sample rate.
I already made some (not scientific) blind tests of that kind on myself, would be interesting to do it on other people though
 
I love the fact my XL even has an audio interface but it was a major pain, when most of my projects were 44.1, to switch to 48. I remember recording, with the XL as the audio interface, on a 44.1 project and it worked... but had some latency issues. Cliff stated that it would up and down convert on the fly but it took processing power to do so.

So perhaps, with the new III, with increased horsepower, the 44.1 auto conversion with be less taxing on the system and at least that convenience factor would be helped.
 
I'm saying that the axe fx already works like that, it's not my idea. Cliff himself stated the amp blocks uses oversampling and a bit depth of 60 bit in the axe fx 3. I just assumed that probably other blocks work like that too and that most of the "tax" is already charged on the axe fx

Ok, there is a difference between oversampling and sample rate increases in terms of cpu load.

If you are oversampling 8 times at 48KHz for each period, that is going to be less CPU/memory intensive than oversampling 8 times at 96KHz or 192kHz by sheer math.

Now if you are saying using an increased sampling rate coupled with lower oversampling to equate the same cpu load, that increases the risk of audio artifacts that the oversampling method is designed to eliminate. Net result is you may end up sacrificing audio quality in order to keep a session in a given sample rate uniformly.
 
Last edited:
Now if you are saying using an increased sampling rate coupled with lower oversampling to equate the same cpu load, that increases the risk of audio artifacts that the oversampling method is designed to eliminate. Net result is you may end up sacrificing audio quality in order to keep a session in a given sample rate uniformly.
Yes I was saying that, 8x48 should be the same to 4x96 theorically.
Anyway what you're saying about audio artifacts goes beyond my knowledge so I will avoid to continue this discussion, but I'd be glad to read more about that :)
 
Yes I was saying that, 8x48 should be the same to 4x96 theorically.
Anyway what you're saying about audio artifacts goes beyond my knowledge so I will avoid to continue this discussion, but I'd be glad to read more about that :)
If I still had my old textbooks, I would recommend them, but they are woefully out of date on a lot of topics anyways.

Just in a regular search, you can find a lot of good writing on the topic, but you can can go down the rabbit hole of crazy formulas rather quickly, depending upon the source.
 
Back
Top Bottom