For discussion about setting up your studio and advice on the gear and equipment within it.
User avatar
By peterpiper Wed Nov 25, 2020 6:48 pm
I said CLEAN HiHats :)
Out of curiosity I just sampled a pitched up loop (+12 semitones which is equivalent to 66RPM) at 10000 and the hihat sounded quite nice (crunchy but nice, at least to my ears :) )
But it depends what kind of hihat sound (or other instruments with high frequencies) is there in first place. Some samples benefit from it others just sound bad.

peace
By Scrawny Wed Nov 25, 2020 6:51 pm
peterpiper wrote:I said CLEAN HiHats :)
Out of curiosity I just sampled a pitched up loop (+12 semitones which is equivalent to 66RPM) at 10000 and the hihat sounded quite nice (crunchy but nice, at least to my ears :) )
But it depends what kind of hihat sound (or other instruments with high frequencies) is there in first place. Some samples benefit from it others just sound bad.

peace


Yeah, like I said lol. There's going to be a lot of experimentation if I decide to get it. Gonna bust out the white lab coat and beakers every time I power it up. :smoker: :lol:
User avatar
By NearTao Thu Nov 26, 2020 1:07 am
If you want a bit more info on the S950 (not sure about the S900 off hand)...

The sampling rate is 2.5 times the bandwidth...

So a bandwidth setting of 19,200hz, equals a 48khz sampling rate. It's a bit wonky... but this means that...
Max Sampling Rate is 48khz or 19,200hz bandwidth
Min Sampling Rate is 7.5khz or 3000hz bandwidth

A byte is 8 bits, so a 12-bit sample is 1.5 bytes.

The S950 reserves 18,720 bytes for program and configuration data.

Each memory bank is 750k on the S950, it comes with one by default, but can have up to three.

So the math for maximum sampling time is:
Max Sample Time = ((# of memory banks * 1024)-18720)/(bandwidth*2.5*1.5)

This leaves you with:
750k memory
* Max Sample Time at 19200 bandwidth = (1 * 750 * 1024) - 18720)/(19200*2.5*1.5) = 10.4 seconds
* Max Sample Time at 3000 bandwidth = (1 * 750 * 1024) - 18720)/(3000*2.5*1.5) = 66.6 seconds

1500k memory
* Max Sample Time at 19200 bandwidth = (2 * 750 * 1024) - 18720)/(19200*2.5*1.5) = 21 seconds
* Max Sample Time at 3000 bandwidth = (2 * 750 * 1024) - 18720)/(3000*2.5*1.5) = 134.8 seconds

2250k memory
* Max Sample Time at 19200 bandwidth = (3 * 750 * 1024) - 18720)/(19200*2.5*1.5) = 31.7 seconds
* Max Sample Time at 3000 bandwidth = (3 * 750 * 1024) - 18720)/(3000*2.5*1.5) = 203.1 seconds

I believe that the bandwidth for the S900 follows the same rules, and I recall it capping out at 16000hz, so I think it caps out at a sample rate of 40khz... but I may be wrong here... suffice it to say, that I believe that the math will work out similarly, and explains why at the maximum bandwidth it is able to eek out an extra second, because if I recall correctly it is sampling just a bit more often on the S950 at maximum.

As pointed out above, it is also true that based on what the Nyquist (sp?) theory tells us, that to sample a sound accurately, you need to be sampling at least twice as much as the frequency of the sound you want to reproduce. This does not mean that you cannot capture parts of a frequency at a lower rate... because obviously you are catching the sample while it is being played, the problem is that you don't have enough points to properly recreate the signal. This is entirely a sample rate issue, and while bit depth does play some factors here, they are harder to explain due to how samplers manage aliasing to smooth out the recreation of the sound.

So... you can think of a reduced bandwidth/sample rate as a kind of awkward low pass filter, because the less you sample, the less you are able to recreate those higher frequencies.

Similarly, as you have asked about increasing the playback speed, and taking peter piper's response... you can think about playing a sound faster as pushing those frequencies higher... which again, reduces your effective bandwidth/sample rate and more or less results as imparting a low pass filter when you bring those sounds back down to their original playback speed. The reason that I am saying it imparts a low pass filter instead of saying that it is a low pass filter is because there is so much more going on under the hood, even on these old school samplers.

* Bit rate does impart a character to the sound
* How bits are setup to represent significant bits matters... hard to explain, but is it optimized for loud sounds, quiet sounds, or in between
* Aliasing and anti-aliasing are things that these older samplers handling in very unique and interesting ways
* The ADC and DAC in these older systems impart a good amount of sonic character that means you don't get modern digital artifacts
* This gear is OLD, and things like capacitors that used to be brand new will impart interesting character today that they did not years ago

And the list can keep going on and on... but suffice it to say that these devices are super hard to recreate, but can certainly be simulated to some extent, but this is also why you see people focus on things like the Low Pass filter or bit depth, versus recreating how variable sample rate/bandwidth works. I'm not devaluing development work that people make for filters, but a lot of it is marketing calling filters S900/S950, and just simulating parts of how it works... usually the easiest parts... instead of focusing on the things that the hardware actually does to make the sounds and how they uniquely approach sampling with limited memory and cpu resources.
User avatar
By richie Thu Nov 26, 2020 4:32 am
I enjoyed reading this post, thanks for sharing your well thought out ideas man.

The S900 highest sample rate I remember is 19200. It is pretty much identical (with boot disk OS 4.0) to the S-950 but without the real-time time stretch and minus the option to upgrade the memory.

There are some that argue that the "low pass is better" on the S-950 but without sounding condescending, it's usually by the same kind of people that believe purchasing a piece of gear because their favorite producer used it is going to make them create the same quality of music.

Regarding the caps thing, I completely agree with you but in the same token (I've had some debates with some engineers about this very thing when we were working on gear together) that in the case of something like an Akai series sampler, or any old piece of gear, I see no reason not to recap the psu as there is an issue with Akai samplers blowing motherboards due to failing psus. It's worth paying $40 to get higher end newly produced legit caps to extend the life of their gear.

On a side note - There was something interesting I remember trying out and was reminded about it last year from a YouTuber vinyljunkie I think, that when you keep cutting the sample rate down on a S-900/S-950 sample save it, and load it on a newer S series (so S1000-S3000/XL sampler) all the higher end frequency range that was lost from chopping the sample rate down, ends up showing up again but sounds aliased. Something to look into again.
User avatar
By NearTao Thu Nov 26, 2020 2:06 pm
Heh... the caps is something I always roll my eyes at when somebody brings it up... myself included. If it is impacting the sound it definitely is not as intended, and absolutely is a sign that they are not functioning properly. I haven't looked at the S950 board in a while... so I don't remember if they have caps on the outputs or not... but that is frequently in place to help prevent voltage from going the wrong direction... ie... don't plug an audio output into another audio output... I usually see busted caps on the output chain just making the signal quieter... which is not something you want.

I just checked the S900 manual on page 21, and it does say it only goes to a bandwidth of 16000hz. What I do not know is if this still equates to 48khz sample rate or is it the same 2.5 factor as the S950. The manual states that at 16000hz bandwidth you have 11,878ms of sample time... Assuming for a 48khz sample rate, this would mean the bandwidth works out to a 3x factor... but yes... this starts making the math funky. If I feel like it I might work on trying to figure it out later. To note, the numbers I posted above are with the Tone program/sample that load by default... and it's possible that the S900 (and probably S950) manual are ignoring the reserved memory for programs and the tone sample *shrug*.

I'd suspect that the S1000-S3000/XL range differences are likely due to anti-aliasing tech, and possibly rebalancing how the bits are applied. Plus don't forget that there is value in going from a 12 bit sampling engine to a 16 bit sampling engine... there will be more accurate (for various definitions of accurate) calculations, which could help explain why it sounds better... there might just be less noise in the system resulting from the 12 bit calculations.
User avatar
By richie Fri Nov 27, 2020 5:19 am
NearTao wrote:I just checked the S900 manual on page 21, and it does say it only goes to a bandwidth of 16000hz. What I do not know is if this still equates to 48khz sample rate or is it the same 2.5 factor as the S950.


I wonder if I'm seeing 192000 because because I auto boot to OS 4.0? I'll have to look into it when I have my S-900 reconnected.

NearTao wrote:I'd suspect that the S1000-S3000/XL range differences are likely due to anti-aliasing tech, and possibly rebalancing how the bits are applied. Plus don't forget that there is value in going from a 12 bit sampling engine to a 16 bit sampling engine... there will be more accurate (for various definitions of accurate) calculations, which could help explain why it sounds better... there might just be less noise in the system resulting from the 12 bit calculations.


What I'm saying is that all the higher end frequency lost when cutting down the frequency, comes back when loading it on a newer S series sampler. I don't believe it to be a particular bonus option but a glitch of some sort when translating the S-900/S-950 sample format. If I had the energy to, I'd compare if there are any differences with how this sounds with a S1000 vs S3000 vs S3000XL but I don't have the patience to right now.
User avatar
By peterpiper Fri Nov 27, 2020 4:33 pm
richie wrote:On a side note - There was something interesting I remember trying out and was reminded about it last year from a YouTuber vinyljunkie I think, that when you keep cutting the sample rate down on a S-900/S-950 sample save it, and load it on a newer S series (so S1000-S3000/XL sampler) all the higher end frequency range that was lost from chopping the sample rate down, ends up showing up again but sounds aliased. Something to look into again.


:shock: strange. But I can imagine that they added some 'enhancements' into the newer series import. Aural Exciter maybe? But then, I never heard or read about that and why would AKAI not promote such a feature?


@NearTao: The s900 have 40kHz not 48kHz (at least thats what the manual says. The graph on an spectrum analyzer tells another story or they implement a way to low filter on the output stage.)

peace

peace
User avatar
By NearTao Fri Nov 27, 2020 4:46 pm
@peterpiper - yeah... I recall the reason they called it bandwidth had something to do with a filter that was applied after playing the sample... but it's not totally clear from what I recall what it was doing... perhaps it was just a lowpass filter with no resonance, but I'm not entirely certain.
User avatar
By NearTao Fri Nov 27, 2020 9:20 pm
Just dropped a blog post on the recordings for the S950 and S1000... S950 and S1000 White Noise and Drum Break Comparison

For those just looking for the direct links
S950 S1000 - Noise and Drum Comparison Project Files
S950 White Noise and Drum Break Disks

Interestingly peter piper is right. There's clearly something going on with the S1000 that it is letting more of the high frequencies through. I suspect there is some kind of LPF on the S950 outputs, but not entirely sure.