Hi all.

A few years ago the theory was that it was better to record for anything that would end-up on CD at 44.1 or an interger multiple thereof. I'm pretty sure that I have read that this is no longer considered to be the rule, and that the benefits of recording at 48, therefore pushing the reconstruction filters further out of the audio spectrum are worth more than the added processing of sample rate conversion? What is the up-to-date view?

I usually record at 44.1 / 24 by the way.


Ta

A.

I tend to record anything I'm 'mastering' at 44.1k and anything a professional mastering house is going to convert at 48.

J

--------------------
www.jackruston.com

I recorded 3 tracks on one weekend at 48 hz and the following weekend the same band, same instruments and amps at 44.1 (24 bit).

The first three songs after mixing and put on CD simply dont have the sparkle or clarity of the three songs done at 41.1

It could be minor differences in mic placement or use of pre-amp settings, but all my recordings since have been at 41.1 and they've all retained crisp clarity when put on CD.

Just my experience.

Back in the mists of time sample rate conversion was tricky and most processes -- hardware and software -- were seriously flawed. Synchronous conversion between integer sample rates was easier, hence the advice you remember.

These days increased processing power and better filter designs mean that sample rate conversion is no longer the weak link in the chain. Even modest SRCs perform better than most D-A converters, and that is true regardless of the conversion ratio.

As to whether there is a benefit in working at 48 and then SRC-ing to 44.1, I would say no. But if you have the power and space, there is a definite benefit to recording and processing at 88.2 or 96 and then converting back to 44.1 at the end. And maintain 24 bits (or better) for as much of the process as you can.

hugh

--------------------
Technical Editor, Sound On Sound

Quote:

---

Hugh: ...there is a definite benefit to recording and processing at 88.2 or 96 and then converting back to 44.1 at the end. And maintain 24 bits (or better) for as much of the process as you can.

---

Is there any advantage in converting existing 44.1 recordings to 88.2 or 96 before processing, then converting back to 44.1 when all processing is completed?

Mike

--------------------
Audio Recording Live

Quote MikeOates:

---

Is there any advantage in converting existing 44.1 recordings to 88.2 or 96 before processing, then converting back to 44.1 when all processing is completed?

---

Interesting question ! Simple theory says there shouldn't be, since any benefits of the higher sample rates are lost when downsampling. But I often put EQ boosts in at around 30 KHz when mastering using our TC System 6000, and they have a clear audible effect even though I'm listening at 44.1. This is presumably because the filters have a broad enough Q for their influence to extend down below 16 KHz into the audible range.

Now in theory ( again ) the same effect should be possible using the right shaped EQ below 22 KHz. However in practise I think finding the right EQ shape to achieve this might be quite tricky. For me, it's much more intuitive to simply boost at 30-40 KHz in the same way you do in the audible range. So the next question is, do applications like Nuendo, Logic etc allow you to do this, as the TC does, or is it necessary to upsample first ?

Ian

--------------------
Production Advice - unlock the potential of your mix

Hi Hugh

I can't find it now, but I think I remember you saying in one of the many discussions about this, that if its for CD release then stick to 44.1 unless project was to be released at higher format then go for 88 or 96?
I believe that working at 24 bit is the more crucial consideration.
Have I got it wrong or faulty memory?

Quote Hugh Robjohns:

---

Back in the mists of time sample rate conversion was tricky and most processes -- hardware and software -- were seriously flawed. Synchronous conversion between integer sample rates was easier, hence the advice you remember.

These days increased processing power and better filter designs mean that sample rate conversion is no longer the weak link in the chain. Even modest SRCs perform better than most D-A converters, and that is true regardless of the conversion ratio.

As to whether there is a benefit in working at 48 and then SRC-ing to 44.1, I would say no. But if you have the power and space, there is a definite benefit to recording and processing at 88.2 or 96 and then converting back to 44.1 at the end. And maintain 24 bits (or better) for as much of the process as you can.

hugh

---

--------------------
Matt
www.krcollective.org

I imagine... that higher sample rates will assist with audio warp features, reducing the amount of frames that need to be extrapolated. But I have no proof of this.

-n.

Personal experiences: it is very convenient to record at 44.1 when going to CD. No conversions, no nothing, what you hear is what you get. At a higher speed the computer is taxed more, more disk space is used and there is one more step needed (the SRC). Seems sort of ill advised to go for all the trouble with the marginal difference at 48. I would say that 96 is probably the choice if 44.1 is not good enough. To me it seems very dependant on the AD-converters, some of them definitely preferrs one or a few speeds more than the other speeds. Not necessarily correlated to price.

Gunnar.

Quote MikeOates:

---

Is there any advantage in converting existing 44.1 recordings to 88.2 or 96 before processing, then converting back to 44.1 when all processing is completed?

---

There can be. High frequency EQ is much more accurate and analogue-like when performed at elevated sampling rates, and in fact many of the better software equalisers upsample internally for this reason. Dynamics processing is also more accurate, and so is metering. So yes, there can be advantages in upsampling for the mixing/mastering processes if you have good quality SRC facilities.

hugh

--------------------
Technical Editor, Sound On Sound

Quote matt keen:

---

I think I remember you saying in one of the many discussions about this, that if its for CD release then stick to 44.1 unless project was to be released at higher format then go for 88 or 96?

---

If your hardware won't support elevated rates because of a lack of DSP power or storage, then it is undoubtedly better to stick to 44.1 rather than use 48 and convert later. But if you can, I'd recommend wortking at 88.2 or 96. personally, I do virtuaklly everything now at 96, and then convert to whatever final rate I need.

Quote:

---

I believe that working at 24 bit is the more crucial consideration.

---

I agree.

hugh

--------------------
Technical Editor, Sound On Sound

96k has a levelling affect on converters: It takes one of the difficult bits further away from what you actually hear. It's not a perfect fix because there's still plenty of circuits there that need to be well designed etc but it will often make a ropey converter perform better than it has any right to sonically.

The problem is that you tend to hammer the processing power quite hard. It's fine for very simple recordings where you may only use a few plug ins, but when you need tonnes of them it gets tricky.

I find that disk space is not so much of an issue any more as you can buy drives so cheaply but also you can pass that expense on to the client.

J

--------------------
www.jackruston.com

Is there an easy way to know if a plug or DAW EQ upsamples? If I have audio that was recorded at 44.1 can I simply set the Cubase project options to 96 or is there more involved (I'm pretty sure that I should know this but still in the office and it's been a long day).

When Hugh suggests that in the case where 88 or 96 is not possible then it's undoubtedly better to stick to 44.1 - does this contradict the statement from earlier that the quality of SRCs these days is so good that it doesn't matter?

As always, thanks to all.

A.

--------------------
Andi, www.thedustbowl.net Mixing, Mastering, Audio Editing at The Dustbowl Audio

Edited by Andi (21/02/07 07:26 PM)

Quote Hugh Robjohns:

---

...if you have the power and space, there is a definite benefit to recording and processing at 88.2 or 96 and then converting back to 44.1 at the end. And maintain 24 bits (or better) for as much of the process as you can.

---

Hugh - for those of us without hardware SRCs, is this still a good idea? My two options for sample rate conversion are using Logic Pro or Bias Peak. Do they do a decent enough job to warrant it?

Quote Barilla:

---

Hugh - for those of us without hardware SRCs, is this still a good idea? My two options for sample rate conversion are using Logic Pro or Bias Peak. Do they do a decent enough job to warrant it?

---

Take a look at http://src.infinitewave.ca/ for some sample rate comaprisons (including Logic and Peak).

Cheers

James.

--------------------
JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.net

Quote James Perrett:

---

Quote Barilla:

---

Take a look at http://src.infinitewave.ca/ for some sample rate comaprisons (including Logic and Peak).

---

Very interesting link indeed, James. Thanks for that. Is it correct to deduce that the best conversions are the 'cleanest' looking on there? If so, Barbabatch seems to be better than Peak and I shouldn't be using Logic for SRC at all, if possible.

Quote James Perrett:

---

Take a look at http://src.infinitewave.ca/ for some sample rate comparisons (including Logic and Peak).

---

Bias Peak wins that one then.

--------------------
John - Sound-Link ProAudio
President - Federation Internationale des Chasseurs de Sons

Quote John Willett:

---

Bias Peak wins that one then.

---

Bit of an eye-opener, John!

WOW! I guess that answers my question then. Fortunately I use Audition, although I have v2 I hope the SRC is the same as v1.5 in those graphs.

Mike

--------------------
Audio Recording Live

Good charts, in the real world though how much difference does noise at -120dbfs make to noise at -170dbfs or -190dbfs make? It's all quieter than my (non-exotic) dacs..

--------------------
Andi, www.thedustbowl.net Mixing, Mastering, Audio Editing at The Dustbowl Audio

Quote John Willett:

---

Bias Peak wins that one then.

---

Actually, I think the very clear winner is the Weiss Saracon.

hugh

--------------------
Technical Editor, Sound On Sound

Someone already highlighted inconsistency here.
First we hear statements like "Even modest SRCs perform better than most D-A converters" and then that there is a clear winner.
I think some explanation would be welcome.
Which converters one has to have to declare the Weiss Saracon a clear winner?

I felt good then I felt bad!

The only SRC's I have access to are within Logic Pro.
Confess that I don't really understand the relevance of the graphs - i.e it looks clear cut but whats that got to do with the real life sound.
What SRC do you use then Hugh, if your allowed to say without upsetting advertisers/suppliers?

--------------------
Matt
www.krcollective.org

Quote George Andoh:

---

First we hear statements like "Even modest SRCs perform better than most D-A converters" and then that there is a clear winner.

---

Single chip converters will struggle to better a signal-noise ratio of 115dB. Looking at those SRC plots, the worst of the spurious artefacts were below 150dB in most cases -- better even than true 24 bit performance. Hence, there is no inconsistency in the statements to which you refer.

Quote:

---

Which converters one has to have to declare the Weiss Saracon a clear winner?

---

It is the clear winner because it quite obviously has the most accurate response with virtually invisible artefacts. Very impressive.

Whether you would be able to tell the difference between that and any of the other SRCs is a separate debate...

Hugh

--------------------
Technical Editor, Sound On Sound

Quote matt keen:

---

Confess that I don't really understand the relevance of the graphs - i.e it looks clear cut but whats that got to do with the real life sound.

---

The Logic SRC plotted on those charts suggets that it produces a lot of artefacts at a realtively high level compared to some of the others. But even so, the worst artefacts were still below -110dBFS, which will be into the noise floor on most systems and all but inaudible.

In perfect monitoring conditions, with excellent monitors, and critical material, you might hear the difference in a direct A/B conparison with something significantly better... but for all practical purposes, there is nothing really to worry about. Moving a microphone six inches or pushing a fader up another 1.5dB will still make a heck of a lot more of a significant difference to your mix.

Quote:

---

What SRC do you use then Hugh, if your allowed to say without upsetting advertisers/suppliers?

---

I take the easy route and use the one provided in Sadie when I'm mastering for CD. It's clearly not the best, but it is clean and benign and is more than good enough for 16 bit CD transfers, which is what it was designed for.

hugh

--------------------
Technical Editor, Sound On Sound

Thanks
24 bit 96k it is then for me and then use Logic Pro SRC after mixing.

--------------------
Matt
www.krcollective.org

Quote Andi:

---

Good charts, in the real world though how much difference does noise at -120dbfs make to noise at -170dbfs or -190dbfs make? It's all quieter than my (non-exotic) dacs..

---

Not to mention, of course the mic, the preamp, the guitar cab, the analog synth, the ROOM ... etc ...

All of these are going to have a noise floor orders of magniture above that!

(Newbie talking out of turn! )

--------------------
New album 'Fantasy Bridge' available now!
Making of Fantasy Bridge Diary

Quote MarkOne:

---

All of these are going to have a noise floor orders of magniture above that!

---

That's certainly true, and in many cases I suspect these noise sources will make it hard or even impossible to distinguish between the subtleties of most SRCs.

However, it is perfectly possible to hear noises well below the average noise floor -- that is one of the magical things about the human hearing process. So if recording in a low noise environment it is possible for the ambient noise to get rather closer to the peak level of the SRC artefacts in poor devices. As you can see from some of those plots, some SRC implementations generate quite complex patterns of side tones and spurious artefacts which, in many cases, are anharmonically related to the source signal.

In other words, unlike normal 'analogue' harmonic distortion, which our sense of hearing largely ignores when present in small qualtities, the kinds of artefacts generated by some kinds of SRC implementation can sound very unmusical and thus tend to draw attention to themselves.

In the past, this was a very real problem. These days, most SRCs -- even the bad ones -- manage to keep the artefacts at a very low level. So low that anyone with budget gear, working in less than perfect circumstances can largely ignore them completely.

But it is worth being aware that there are still quite significant differences between the best and the worst, even if you can't hear it at the moment.

hugh

--------------------
Technical Editor, Sound On Sound

Couldn't sleep last night so read some of the articles by Bob Katz on digido.com. In one article entitled "mixing tips and tricks" Mr K states "Always mix to 24 bit files, even if the source tracks are 16 bit! Do NOT sample rate convert." Emphasis is his. I'm never sure when I read this sort of thing if the advice applies only to million dollar equipment set-ups and 10 million dollar ears. In apparent contradiction, I believe that some of the Massenburg designed EQs do upsample.

Upsampling in EQ's is a slightly different proposition to independent sample rate conversion and doesn't necesarily suffer from the theoretical issues that affect non-interger rate conversions.

I think the point Katz is making is that if you are sending your work to be professionally mastered, it is best to give them the best possible source to work with (hence the 24 bit files idea) and to leave all the complicated processing to them because they'll generally have the best and most appropriate equipment.

Hugh

--------------------
Technical Editor, Sound On Sound

Hugh

That would make perfect sense. Take the rest of Saturday off?

Quote Andip:

---

Couldn't sleep last night so read some of the articles by Bob Katz on digido.com. In one article entitled "mixing tips and tricks" Mr K states "Always mix to 24 bit files, even if the source tracks are 16 bit! Do NOT sample rate convert." Emphasis is his. I'm never sure when I read this sort of thing if the advice applies only to million dollar equipment set-ups and 10 million dollar ears. In apparent contradiction, I believe that some of the Massenburg designed EQs do upsample.

---

I doubt Bob Katz spent that much on equipment even if you count his propane powered gas-turbine electrical generator as audio gear.

Looking through those I notice Logic, Soundtrack Pro and Digital Performer all produce exactly the same results. Implies they're all using a OS/system based SRC rather than their own.

...or SRC software they have all licenced from the same source

hugh

--------------------
Technical Editor, Sound On Sound

Quote matt keen:

---

Thanks
24 bit 96k it is then for me and then use Logic Pro SRC after mixing.

---

I'm with you on this one. I worked at 24 bit 44k until today

Just as well as I bought another UAD card. Everything happens for a reason

Cheers,
Peter

Quote Conz:

---

Quote matt keen:

---

Thanks
24 bit 96k it is then for me and then use Logic Pro SRC after mixing.

---

I'm with you on this one. I worked at 24 bit 44k until today

Just as well as I bought another UAD card. Everything happens for a reason

Cheers,
Peter

---

Certainly does.
Incidently had a fabulous holiday last summer at Budle Bay and Lindisfarne.

--------------------
Matt
www.krcollective.org