why is higher sample rate changing the sound for the worse? I think its changing the pitch

Question

I'm working on a current song which I posted an example of a while back when asking about why low frequency sawtooth oscillators change the sound. Since then I changed the base of the synth from NI Massive to Ableton's built in Operator synth.

My problem is that when I export the song, the main synth sounds totally different from the way it sounds when I'm working on it in Ableton. The highest possible sample-rate setting I can use is 96000 when in Ableton, so when I export it at that rate it sounds exactly the same. I would just keep it that way if every other synth in the song didn't sound better.

Here is the way it sounds when I play it in Ableton with the main synth exactly how I like it

Here it is with the sample rate raised to maximum

I like the way it sounds much better in the first one. To me, the higher sample rate one sounds lower quality and looses all the punch. The weird thing is that, I've never had this problem with any other song until I upgraded to the newest version of Ableton. This same thing happened to me with another song I'm working on and I just thought it was the synth acting quirky since I used a pretty odd oscillator.

A little more info and things I do differently then normal on this song and the one I just mentioned: The synths are on an instrument-rack which I haven't started using until recently. My other songs with similar synths were created in an older version of Ableton but edited in the new one, but exported at the exact same settings as the 2nd higher sample rate example.

What is the cause of this and can I make it sound like the first one but still have everything else be at the higher sample rate?

EDIT: So I think what it is changing is the pitch

Answer 1

You have happened upon one of the edge cases where the sampling frequency above what humans can hear actually does matter. That edge case is FM synthesis.

There is plenty of anecdotal discussion on forums all over the internet that boils down to people arguing about the original Yamaha DX series of FM gear vs softsynths like FM8. These discussions typically devolve into a shouting match, but along the way a few important things emerge (and are ignored by many): sample rate matters, and bit rate matters. It is up to the individual artist to determine what combinations are better than others, but it does matter.

The reason why this is the case is that the process of frequency modulation generates frequencies that can be well above what a human can hear. How these frequencies are handled can have a significant impact on the sound quality of an FM synth.

Instead of going into the math and technical details of sidebands, interpolation, and aliasing, I'll give you an experiment you can try on your own:

1. Create a new project in live
2. Drop a new instance of Operator into a MIDI clip
3. Change the Level value for operators B, C, and D to -6 dB
4. Add a Redux after Operator
5. Create a 1-bar sequence of MIDI notes, then hit play
6. Mess around with the Downsample value (and bit reduction, if you want) as the MIDI clip plays. Notice that downsampling has vastly more impact on your FM patch than it does on samples or Analog patches.

So, to solve your problem you just need to find a plugin that will allow you to clamp the sample rate of the audio coming out of Operator to 96KHz (or whatever you find pleasant). Redux might not be the best choice, since the Downsample control really controls decimation, which depends on the main sample rate. But it should be good enough for experimentation. You might try bouncing just your operator track at 96KHz and then drop that long sample back into your project and bounce the whole thing (after muting the original operator synth) at 196KHz.

TL;DR - it is the sample rate (and possibly bit depth) that matters here. But these are things that you can mess with via plugins. Have fun.

Edit: my above explanation is rather hand-wavey. I'll try to update it with proper terminology and a better explanation. In the meantime, if you have a solid grasp of the issue, please feel free to edit the above for clarity and correctness.

Answer 2

I would say your hardware or software can't keep up with that sample rate. You really shouldn't need to go that high, but if you do then you really need a pro sound card.

Double the top frequency you can hear is ample, so anything over 40k should do.

Answer 3

This information was given to me by an individual who has three gold selling credits to his name. I was skeptical, and conducted a test.... and he is correct. Here are the details of what I had explained to me.

I got back into the home recording trend shortly after Y2K, and began with a small set up. I had a PC, and some software for recording multitrack. I contacted a manager friend of mine, and we had a discussion about the digital format. I was unsure as to what frequency to record with. My question centered around 16bit vs 24bit. The ignorance coming out of the mouths of music store salesmen was assnine, and they were encouraging you to record at 16bit... and I knew that wasn't correct. My manager friend suggested an individual I should contact, and so I did. The guy has many more years experience than I, and I was highly interested in his opinion. Not only did he confirm my belief (contrary to every music store salesman, and they got in arguments with me about it) 24 bit was the way to go. The other little bit of information he shared with me was to record at 88.2Khz. Now this was odd. He was adamant about that frequency, and explained that if you try and record at 96Khz, and you don't have a sophistocated dithering machine to get you back down to 44.1Khz, you are going to destroy your high frequencies. I was skeptical, and spent an entire day running test burns with square waves, guitar recordings with massive reverb, sine waves of varying frequencies. I dithered in several formats going down to 44.1, and there was triangle (as well as 6 other combinations of options). It took me a good 8 hours to complete the test. I was amazed at how much the reverb was completely granulated when going from 96Khz, and down to 44.1Khz. It made a believer out of me, and so I've always recorded at 88.2Khz, and never at any other frequency going in. Sometimes you have S/Pidif gear that only allows you to track at 44.1, but other than that, Its 88.2 for everything.

If I were you, I would try tracking at 88.2Khz, and then at 96Khz... save both test recordings at 44.1Khz, burn a CD.... and go play the test CDs out in your vehicle, or little stereo blaster. You might find an amazingly not too well known, or understood truth about proper tracking frequency format...

Good luck! Tim

Answer 4

The effect you are getting seems to me like a sound alternation you get at the lower khz setting. Maybe you like this "effect" and maybe not. The higher khz recording represents more the "original" sound you created.

Why is this so?

Check out http://www.mother-of-tone.com/cd.htm for an indepth look at various sampling rates.

I only record at 192khz to prevent sound alternations. If I do not like the sound I am getting at 192khz I can always alter it with plugins. But I can be sure that I am not adding a very hard to control sound alternation by sampling at a too low rate.

Hope this helps.

Cheers, Original GuGa

PS.: When exporting recorded sound to CD format for example, you will of course need to downsample. This is not as simple as it may seem. Just reducing the Bit and khz can again distort the recording. I use iZotope 64-bit SRC™ (sample rate conversion) and iZotope MBIT+ dither (bit-depth conversion) to convert the recording to CD format. Sound Forge features these. Do not know if Ableton does. Products like iZoptope RX also have these built in and can be used. More on this here: http://help.izotope.com/docs/rx1/pages/resample_dither.htm

Answer 5

Some of the manufactures who claim 192 sample rate is really not a true statement, manufactures can use programming to tell a 96Khz chip to emulate a 192Khz recording. This can actually affect the quality of the recording and make a sample sound worse than if it was recorded at 96Khz.

A true 192Khz sample rate AD can be very expensive. As an example think of Apogee, RME, MOTU. Echo Audio has some interfaces that provide true 192khz sample rates also.