The hardware items in my system offer various options for sampling frequency.

My digital mixer offers 44.1, 48 and 96 Hz options.

The audio interface I use with my desktop PC offers 44.1, 88.2, 176.4, 48, 96 and 192 Hz.

The audio interface I use with my laptop offers 44.1, 48 and 96 Hz.

I have read that it's best to use 44.1, 88.2 or 176.4 for projects whose end destination is a CD, whereas 48, 96 or 192 is advisable for DVD work.

However, only 44.1, 48 and 96 options are available on all my gear. 88.2, 176.4 and 192 are only offered by my main audio interface.

I know that 44.1 is CD quality, but I understand that it's preferable to use a higher sampling rate when working on a project prior to the mastering stage, i.e. keeping the highest possible quality until the last moment.

Which sampling rate would therefore be best for me to use when working on audio projects where CDs are the destination?

My software DAW is Cubase Studio 4.1 and my Windows Vista Home Premium system uses an Intel Core 2 Duo E6600 2.4 GHz processor with 2 Gb of RAM installed.

--------------------
Steve Cooke
http://soundcloud.com/stevecooke

I record at 96khz when I'm doing any serious work and benefit from extra headroom albeit at the expense of plug in counts etc. The rest of the time whether writing or just recording rehearsals I stick to 44.1khz.

48khz is DVD standard but I've never heard of using specific multiples of rates depending on destination? Maybe someone could explain the theory?

You can record at 44.1khz and benefit from smaller files and lower system overheads and not have any problems. It really depends on the quality of your equipment.

I think that a lot of professionals record at higher sampling rates to 'future proof' their work against advances in technology.

My Christmas morning pennies worth before the madness starts.

Quote SixStringzzz:

---

48khz is DVD standard but I've never heard of using specific multiples of rates depending on destination? Maybe someone could explain the theory?

---

I read it in the manual for the Alesis iO|26 (page 48), which I purchased the other day.

The manual says:

"If you’re unclear on what base sample rate to use, consider this guideline:
• If your recordings are slated for release on CD, MP3,
Casette, Vinyl, etc. set your sampling rate to 44.1, 88.2,
or 176.4k for best results.
• If your project is slated for DVD, film, or television, set your sampling rate to 48, 96, or 192k for best results."

I hadn't encountered such a formulation before either. A curious one too, considering that most of the audio interfaces I've seen offer 44.1, 48 and 96k options. If it was that important to use multiples of the destination sampling rate, you'd have thought that 88.2 would be provided as standard on most equipment.

--------------------
Steve Cooke
http://soundcloud.com/stevecooke

i heard somewhere that if its going to be released on CD at 44.1 its best to record at 88,2 rather than 48 or 96, is this true? Is it because when it comes to down sampling it to 44.1 its just half?


--------------------
Live Sound and Studio Engineer


Edited by Paul Supersonix Studios (25/12/07 02:49 PM)

I think it used to be advisable to use direct multiples but not any more.

Maybe when Mr Robjohn has finished his mince pies, he may have something to say on the matter!

--------------------
Website / Music Lab Machines / Blog

Quote SixStringzzz:

---

I record at 96khz when I'm doing any serious work and benefit from extra headroom albeit at the expense of plug in counts etc.

---

There aint any extra headroom in 96Khz comparing to any other sample rate. You do get increased frequency response (beyond hearing range), and possibly some extra distortion, but no extra headroom.

Quote SixStringzzz:

---

I record at 96khz when I'm doing any serious work and benefit from extra headroom

---

Are we thinking of Bit Depth?

For CD work the process of converting from 48kHz to 44.1 is a more complex mathematical process than converting from 88.2kHz which is a more simpler ratio (half). The artifacts to the audio added in converting from 48kHz can subjectively make the audio worse than just recording at 44.1kHz.

For DVD as stated this is different as the audio standard is 48kHz in which case using 48kHz upwards is key.

But as pointed out most people record in higher sampling rates for better quality during processing and the important future proofing for new medium.

Remember when choosing a sample rate the ease of use (if transporting across several machines) and disk space (if this is an issue).

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Quote Nuno_:

---

Quote SixStringzzz:

---

I record at 96khz when I'm doing any serious work and benefit from extra headroom albeit at the expense of plug in counts etc.

---

There aint any extra headroom in 96Khz comparing to any other sample rate. You do get increased frequency response (beyond hearing range), and possibly some extra distortion, but no extra headroom.

---

Just beat me to it!

Also it allows the Low Pass Filter used in Analog to digital converters to have a smoother roll off.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

As Mr Threlfall says, 48kHz is standard for DVD - ever noticed how TV programmes use tunes that you know inside out, but they sound up-pitched? Listen to the original of Massive Attack's 'Teardrop' and compare it with House season 3...you'll see what I mean.

Theoretically, I find very little gain in mixing at 48kHz whatsoever for CDs. For one, the numbers apply to a cycle of 1Hz - as in, 44.1kHz will sample a cycle of 1Hz 44,100 times. Take it up into the audible range and think of 100Hz, which has a wavelength of one hundredth of that, so it'll sample it 44.1 times per cycle instead of 48. Go up to 1kHz and...you catch my drift. It soon becomes an incredibly negligible number. Plus, I don't know anybody who can actually hear the difference!

--------------------
"He that hears music feels his solitude peopled at once." - Robert Browning

Okay, but are there any insurmountable problems caused by recording material at 96kHZ when you intend that a CD will be its eventual destination?

My audio interface offers 44.1, 48, 88.2, 96, 176.4 and 192 as options, but my digital mixer only offers 44.1, 48 and 96 (as also does my laptop's audio interface, where I might want to work sometimes). I presume, therefore, for the sake of compatibility between my gear, that it would be best to confine myself to one of the sampling rates common to all the equipment in my set-up, which means one of 44.1, 48 or 96.

If I want to record at a so-called 'professional grade', it seems that the best option is 96kHz? Supposing I record at 96kHz, would I have any difficulty converting that to 44.1kHz when mastering to CD? I have Sony Sound Forge 9 at my disposal for such a project.

--------------------
Steve Cooke
http://soundcloud.com/stevecooke

Quote:

---

Okay, but are there any insurmountable problems caused by recording material at 96kHZ when you intend that a CD will be its eventual destination?

My audio interface offers 44.1, 48, 88.2, 96, 176.4 and 192 as options, but my digital mixer only offers 44.1, 48 and 96 (as also does my laptop's audio interface, where I might want to work sometimes). I presume, therefore, for the sake of compatibility between my gear, that it would be best to confine myself to one of the sampling rates common to all the equipment in my set-up, which means one of 44.1, 48 or 96.

If I want to record at a so-called 'professional grade', it seems that the best option is 96kHz? Supposing I record at 96kHz, would I have any difficulty converting that to 44.1kHz when mastering to CD? I have Sony Sound Forge 9 at my disposal for such a project.

---

In a word, no. You won't experience any problems with bouncing it down to 44.1kHz from 96kHz, the programs tend to handle this very well. It will mean, as someone else above pointed out, less plug-in counts as it loads a lot more on your processor, but if that's not a problem I say go ahead.

Personally I've never had a problem with the quality of recording at 44.1, as it's gonna end up there anyway. This is my colours nailed firmly to the mast in a long-fought debate! Yes, theoretically, running at 96kHz gives you more than double the resolution on your recording. Give it a whizz doing a recording in 44.1 and another from the same source / signal chain at 96 and see which you prefer.

--------------------
"He that hears music feels his solitude peopled at once." - Robert Browning

Quote TurboD:

---

As Mr Threlfall says, 48kHz is standard for DVD - ever noticed how TV programmes use tunes that you know inside out, but they sound up-pitched? Listen to the original of Massive Attack's 'Teardrop' and compare it with House season 3...you'll see what I mean.

---

Sorry to shoot down you agreeing with me, but that pitched up is to do with changing the frame rate from 24fps to 25fps (PAL) which involves speeding up the audio by 4.1%. Although if you were to play a 44.1 track at 48kHz it would result it a speeding up of the audio, and thus the pitch.

Quote TurboD:

---

Plus, I don't know anybody who can actually hear the difference!

---

I would agree. Differences i feel can be heard when using 96kHz (Especially due to the ability to have a less steep LPF) but the difference between 44.1 and 48 is very difficult if not impossible on most material to tell the difference.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Ah - apologies and thanks for the correction. Crossed wires somewhere along the way...!

--------------------
"He that hears music feels his solitude peopled at once." - Robert Browning

Quote TurboD:

---

Personally I've never had a problem with the quality of recording at 44.1, as it's gonna end up there anyway.

---

I agree.

It can be argued that recording at übersonic sample rates gives you better quality/resolution and I am led to believe that it gives more numbers for processors to crunch with better accuracy but I am not sure that *in practice* there is much to be gained : is there a great audible difference given our ears, our ADC/DACs, our mics', amps' and speakers frequency response, our (typically) less than ideal monitoring facilities (untreated acoustics in 'home' and 'project' studios), etc.? And if you're working with modern synths that operate at a 'nominal' sample rate of 44.1kHz, there is even less reason to use high sample rates, especially with the type of material many of us produce.

Given all these factors, I see no real benefit for most of us to be working at anything other than 44.1 especially for eventual CD release ... or MP3 ... or at least, most of us needn't be losing sleep over which hypersonic sample rate to be using - 9 times out of 10 in most situations for most people in most genres, 44.1 is just fine. The only thing that is guaranteed using higher sample rates and bit depths is (massively) increased file sizes, a requirement for more storage and processing power (and/or a reduction in the amount of plugs, channels whatever, available to you) and so on. It can also be argued perhaps that the side effects and artefacts generated by less than ideal resampling/dithering algorithms are possibly worse than simply recording at 44.1/16 in the first place (IYSWIM).

So weighing these things up, I always use 44.1/16 and leave the higher rates and bit depths to those who might genuinely need them in their 'professional' recordings ... or are quite happy to blow their storage and processing on higher spec'd recordings for no appreciable audible benefits.

And maybe this pragmatism of mine has its roots in the fact that I grew up on 'domestic' reel-to-reels, then open reel 4-tracks (with no NR), 1/4" 8-tracks, noisy analogue desks and so on - modern 44/16 satisfies (nay - exceeds) my recording requirements! And arguably, 44/16 exceeds the spec and performance (on paper at least) of many 'cutting edge' studios' recording gear of yesteryear.

Just MHO of course - YMMV. I guess all I am trying to say is don't sweat over sample rates - it's just not worth it. Focus more, perhaps, on the *content* than the specs

--------------------
Website / Music Lab Machines / Blog

Some very interesting points there hollowsun.

However just to clear up Sampling Frequency in no way increases the resolution. That is the bit depth. Nuno_ 's post above explains the effect sampling frequency has.

While I would agree that using a higher sampling frequency is many cases pointless. Using 24bit while recording I feel is a very good idea, as it does increase resolution, and allow reduced quantization noise and much increased quality at low levels. The effect of a well chosen dither, performed at the correct stages will cause few artifacts compared to the benefits.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Quote Ian Threlfall:

---

However just to clear up Sampling Frequency in no way increases the resolution.

---

Ok. "Resolution" might not have been the best word to use but I take your point.

--------------------
Website / Music Lab Machines / Blog

Nyquist's theory my friend.

The sample rate can only produce frequencies up to half of the sample rate.
For example:
44.1kHz sample rate can only produce frequencies up to 22kHz (approximately). So if loosing a little bit of top end isn't an issue, 44.1kHz should be just fine.

--------------------
"How many drummers do you know of that kick their drums?!"

Quote Casiokeys:

---

44.1kHz sample rate can only produce frequencies up to 22kHz (approximately).

---

In theory.

In practice, with filters, etc., more likely to be 20kHz.

But in practice as well, our adult ears tend to roll off at around 17kHz or less as we get older and most mics aren't much better (looking at two extremes, the Shure SM57 and the Neumann U87 both roll off by around -10dB to 20kHz after ~13kHz) and most monitor speakers don't extend much above 20kHz.

Worth bearing in mind for some perspective perhaps.

--------------------
Website / Music Lab Machines / Blog

Quote hollowsun:

---

Quote Casiokeys:

---

44.1kHz sample rate can only produce frequencies up to 22kHz (approximately).

---

In theory.

In practice, with filters, etc., more likely to be 20kHz.

But in practice as well, our adult ears tend to roll off at around 17kHz or less as we get older and most mics aren't much better (looking at two extremes, the Shure SM57 and the Neumann U87 both roll off by around -10dB to 20kHz after ~13kHz) and most monitor speakers don't extend much above 20kHz.

Worth bearing in mind for some perspective perhaps.

---

An excellent point. Plus 2 sample points per frequency is the minimum to avoid aliasing, but an increase in the number of samples will provide a better quality digital representation of that frequency. But again one of the main advantages is that the anti-aliasing filter can be less steep and be out of the audible range.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

OK, I feel there are now enough slight inaccuracies in this thread to merit some form of pendantic response...

Quote Ian Threlfall:

---

For CD work the process of converting from 48kHz to 44.1 is a more complex mathematical process than converting from 88.2kHz which is a more simpler ratio (half).

---

Well, yes, if only it were that simple. An downsampling SRC is essentially a lowpass filter- or at least any half-decent, up to date design is. It's not as simple as throwing away half the samples to arrive at 44.1- because obviously in that situation any components above 22kHz-ish would alias. All SRCs work (I'm led to believe) by padding out the existing samples with zeroes up to a common multiple, applying a lowpass filter, then decimating back down to the destination frequency. As the turnover point of this lowpass filter is related only to the destination frequency, it shouldn't matter whether the conversion to 44.1 is taking place from 48, 88.2 or 17GHz. It might take longer to SRC from 96k-44.1 than from 88.2-44.1, but seeing as this only needs to take place once in the entire process, and only on the final 2-track mixdown, the time differential should be virtually negligible.

Quote Ian Threlfall:

---

Using 24bit while recording I feel is a very good idea, as it does increase resolution NO, and allow reduced quantization noise YES

---

To be clear, neither increased sampling rates nor increased bit depths improve the 'resolution' of a digital recording system. Increased sampling rate gives more bandwidth. Increased bit depth improves the signal/noise ratio of the system. Resolution is a misleading term to use when discussing the recording or playback elements of a system.

Quote Casiokeys:

---

Nyquist's theory my friend.

---

Theorem - it's proven.

Quote Ian Threlfall:

---

Plus 2 sample points per frequency is the minimum to avoid aliasing, but an increase in the number of samples will provide a better quality digital representation of that frequency.

---

More than anything else on this thread- NO! Absolutely, unequivocally not. You need >2 sampling points per cycle to accurately reconstruct the waveform. That's it. No more. The rest are mathematically redundant. It doesn't matter if you have a thousand points per cycle of the waveform- the reconstruction of that waveform is no more accurate than with the bare minimum.

In response to the original poster, I would personally just record at 44.1 and not give it a second thought! Enjoy the benefits of reduced processor strain and smaller file sizes.

Adam

Quote Ian Threlfall:

---

Sorry to shoot down you agreeing with me, but that pitched up is to do with changing the frame rate from 24fps to 25fps (PAL)

---

That was in the old film days and hasn't been a factor since US post-production went to tape (and now disk).

You'll probably find the music track was sped up by the editor to fit the scene!

one thing to bear in mind is that your audio device automatically upsamples your material in order to get much less steep AA filter. we are talking in rates of 12,24,48 x sample rate here.
the best thing is to work at default sampling frequency of your final medium. i work with vinyl mostly and therefore my sampling freq. of choice is 44.1.
if i was to use 48 or 96, i would have to downsample, which is a process that mathematically predicts the new positions of samples. it is not a real representation. same thing happens when you convert from 44.1 to 96 e.g.

in the end, ask yourself - can you hear the difference btw different sample rates?
44.1 uses the least processing power and harddisk space and quite frankly sounds darn good if you know what you are doing.

Quote ceejay:

---

Quote Ian Threlfall:

---

Sorry to shoot down you agreeing with me, but that pitched up is to do with changing the frame rate from 24fps to 25fps (PAL)

---

That was in the old film days and hasn't been a factor since US post-production went to tape (and now disk).

You'll probably find the music track was sped up by the editor to fit the scene!

---

An excellent point, and quite probably true. I was thinking a bit too much me thinks, but we can both agree it wasn't caused by the wrong sample rate.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Quote Ian Threlfall:

---

Using 24bit while recording I feel is a very good idea, as it does increase resolution NO, and allow reduced quantization noise YES

---

To be clear, neither increased sampling rates nor increased bit depths improve the 'resolution' of a digital recording system. Increased sampling rate gives more bandwidth. Increased bit depth improves the signal/noise ratio of the system. Resolution is a misleading term to use when discussing the recording or playback elements of a system.

---

I understand completely what your saying, ie Yes you don't get any more actual audio level, but simply more bits to quantize the value. And I understand if taken wrongly the word 'resolution' can be mistaken for this. However in computers and the digital world, the meaning of resolution is just that.

Quote http://www.answers.com/topic/display-resolution?cat=biz-fin:
:

---

he number of bits used to record the value of a sample in a digitized signal."

---

This could be pictures, video or audio.

Quote adam miller"]

Quote Ian Threlfall:

---

For CD work the process of converting from 48kHz to 44.1 is a more complex mathematical process than converting from 88.2kHz which is a more simpler ratio (half).

---

Well, yes, if only it were that simple. An downsampling SRC is essentially a lowpass filter- or at least any half-decent, up to date design is. It's not as simple as throwing away half the samples to arrive at 44.1- because obviously in that situation any components above 22kHz-ish would alias. All SRCs work (I'm led to believe) by padding out the existing samples with zeroes up to a common multiple, applying a lowpass filter, then decimating back down to the destination frequency. As the turnover point of this lowpass filter is related only to the destination frequency, it shouldn't matter whether the conversion to 44.1 is taking place from 48, 88.2 or 17GHz. It might take longer to SRC from 96k-44.1 than from 88.2-44.1, but seeing as this only needs to take place once in the entire process, and only on the final 2-track mixdown, the time differential should be virtually negligible.

---

I can take your point on that, that is indeed how SRC work. The difference would be very minimal, however manufactures such as Cakewalk state that changes from 48k to 44.1k can cause extra noise added, which a conversion from a higher frequency wouldn't.

http://www.cakewalk.com/Tips/TechTipMay02_3.asp

Quote adam miller:

---

Quote Ian Threlfall:

---

Plus 2 sample points per frequency is the minimum to avoid aliasing, but an increase in the number of samples will provide a better quality digital representation of that frequency.

---

More than anything else on this thread- NO! Absolutely, unequivocally not. You need >2 sampling points per cycle to accurately reconstruct the waveform. That's it. No more. The rest are mathematically redundant. It doesn't matter if you have a thousand points per cycle of the waveform- the reconstruction of that waveform is no more accurate than with the bare minimum.

---

Really? I am not doubting you, I would have just thought that more sampling points would allow for a greater interpretation of the wave, relying less on averaging.

Slightly like this crude graph.



I take your point on board though, I don't want to be misleading people!

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Quote Ian Threlfall:

---

Really? I am not doubting you, I would have just thought that more sampling points would allow for a greater interpretation of the wave, relying less on averaging.

Slightly like this crude graph.



I take your point on board though, I don't want to be misleading people!

---

No. The problem is there is no averaging at all going on. There's no guessing going one regarding the curve that joins the two samples. Assuming a proper bandwidth limited signal there's only one possible curve and having more samples does not create a better representation of the wave. That is why it is so important the use of low pass filters, if any content from frequencies beyond the nyquist frequency are captured then this doesn't hold true anymore.

Edited by Nuno_ (28/12/07 07:30 PM)

Makes perfect sense. Cheers for that Nuno_ / Adam.

--------------------
Ian Threlfall
magicgenie.co.uk web hosting

Quote Ian Threlfall:

---

Makes perfect sense. Cheers for that Nuno_ / Adam.

---

No worries


I have to say that I'm not an expert in this at all, but I did study some stuff that overlaps with it in college and that plus a few articles by Dan Lavry in his site (worth reading BTW) were enough to realise that a lot of the arguments used in favour of higher sample rates in most forums really don't hold much water.

Quote Ian Threlfall:

---

I can take your point on that, that is indeed how SRC work. The difference would be very minimal, however manufactures such as Cakewalk state that changes from 48k to 44.1k can cause extra noise added, which a conversion from a higher frequency wouldn't.

---

From the Cakewalk site:

"Please note you will achieve better results when converting audio from 96 kHz to 44.1 kHz as apposed to 48 kHz to 44.1 kHz. You may find, depending on your source material that going from 48 kHz to 44.1 kHz may introduce noise into your final mix. This may need to be further processed through some form of noise reduction, be it an external editor, a DirectX plug in, or a hardware-based unit."

So SRCing from 48k down to 44.1 with Cakewalk introduces such severe noise that it requires de-noising afterwards? Time to get a new bit of software I think....

Quote Ian Threlfall:

---

Makes perfect sense. Cheers for that Nuno_ / Adam.

---

That's the trouble with sampling theory... it really doesn't seem make sense on an immediate, intuitive level! The books by Pohlmann / Watkinson on digital audio do a pretty good job of explaining the concepts without reverting to the core maths- which I certainly couldn't explain to you!

Cheers,

Adam

Quote adam miller:

---

Quote Ian Threlfall:

---

I can take your point on that, that is indeed how SRC work. The difference would be very minimal, however manufactures such as Cakewalk state that changes from 48k to 44.1k can cause extra noise added, which a conversion from a higher frequency wouldn't.

---

From the Cakewalk site:

"Please note you will achieve better results when converting audio from 96 kHz to 44.1 kHz as apposed to 48 kHz to 44.1 kHz. You may find, depending on your source material that going from 48 kHz to 44.1 kHz may introduce noise into your final mix. This may need to be further processed through some form of noise reduction, be it an external editor, a DirectX plug in, or a hardware-based unit."

So SRCing from 48k down to 44.1 with Cakewalk introduces such severe noise that it requires de-noising afterwards? Time to get a new bit of software I think....

---

Well Yes not the best example!

--------------------
Ian Threlfall
magicgenie.co.uk web hosting