Quantcast
Classical music recording.... - Page 6 - Gearspace.com
The No.1 Website for Pro Audio
Classical music recording....
Old 9th March 2009 | Show parent
  #151
Lives for gear
 
sonare's Avatar
 
2 Reviews written
🎧 15 years
Quote:
Originally Posted by Plush ➑️
Not only that!, but it was said by Tony F.

with thanks for your appearance here, Tony.
Fellow Slutz, I call your attention to the fact that "Tony F" is Tony Faulkner. I do this not to idolize him or his views but to point out what should be obvious-- that Tony has "paid his dues" and what he says deserves respect. If you don't agree, then disagree WITH RESPECT. Ask him, for instance, exactly how to achieve what he describes in a DAW rather than digital mixer.

Well I guess I already did!

Rich
Old 9th March 2009 | Show parent
  #152
Gear Guru
 
joelpatterson's Avatar
 
2 Reviews written
🎧 15 years
Now, if we could just get William Faulkner, then it would really be a party!
Old 9th March 2009 | Show parent
  #153
Gear Addict
 
🎧 15 years
Being honest I like Rode mics too however I must be honest that one of my NT1A matched pair died during a test this weekend, 3 weeks before gig so service dept will get them sorted in 10 days they say.

It is fairly rare for mics to break without obvious physical cause but it did remind me to have 2 pairs set up and on seperate preamp units. Then both pairs can have full volts with no sags to themselves !

I am told there are to be no soloists which means for this gig I can concentrate on getting the angles, spacing and height correct, I look forward to this and I am listening to as much ref material on the DT990 pro which I will be monitoring.
(I do have AKG K240's too much more neutral) the DT 990 are a quiet a smiley FR
set of cans but I like the detail.

I am thinking both pairs of mics should be ORTF.

For some reason to the eye ORTF seem to be angled quite far outwards, would that not just capture the L/R extremities or is this balanced with the fairly short spacing (17cms) to give a decent stereo field to the recorders. I will have to listen
carefully on the day I will have 5 hours prep time with 6 varying sizes of orchestra's performning so i should have loads of time to experiment.
Old 9th March 2009 | Show parent
  #154
Gear Addict
 
🎧 15 years
I am considering 88.2kHz/24 bit from the perspective of "If you don't try you'll never know"
Old 9th March 2009 | Show parent
  #155
Gear Guru
 
joelpatterson's Avatar
 
2 Reviews written
🎧 15 years
Whenever I say "ORTF" I really mean "NOS" (Nederlandsche Omroep Stichting, the Dutch version) which has the mics splayed out at 90 degress and spaced at 30 cm apart.

There's just something that sounds instinctively right about "any sound falls somewhere within this 90 degree square field defined by the mics." I know, I know, your head isn't shaved into a 90 degree point (unless you're into metal, maybe), but it just seems that the set-up would naturally result in pinpoint stereo placement of anything that hit it.

Although you'd have to be way into the esoterica of it all to categorize the difference between the two... which I expect to see in a few posts from now...
Old 9th March 2009 | Show parent
  #156
Gear Addict
 
🎧 15 years
Well I thought ORTF was 110 degrees and that seemed a little wide to the eye
especially with LDC's and the reported tendancy not to be as accurate at the HF as SDC's.

I suppose to the ear is what counts but 90 degrees seems like a logical start point as you mention.
Old 9th March 2009 | Show parent
  #157
29327
Guest
Quote:
Originally Posted by XLR2XLR ➑️
Well I thought ORTF was 110 degrees...
Indeed it is 17cm at an angle of 110Β°. You're right about it "seeming too wide", but it works!

So, speaking of near-coincident mic pairs, here's something everyone might enjoy. Image Assistant 2.1 The Java app on this page allows you to construct a mic array with a variety of polar patterns, angles and spacing and then view the resulting IAD and ITD's, localization, etc. Is it the bible of stereo pairs? Mathematically yes, subjectively no. But is it fun to play with? Absolutely! heh
Old 9th March 2009 | Show parent
  #158
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by XLR2XLR ➑️
I am considering 88.2kHz/24 bit from the perspective of "If you don't try you'll never know"
The vast majority of the stuff I do is at 44.1... That said, what is quoted here is very intelligent. You won't know what is sounds like UNTIL YOU DO IT. Then you make a choice based on knowledge which is always the best way.

My choice was made having done dozens of high sample rate recordings and deciding that what it usually gave me that was good was outweighed by what was bad (ie SRC, extra processing, effects such as reverbs that only go to 48K, etc...). I still will do 2fs recordings on occasion, but they are few and only when the room and the setup benefit from the extra overhead.

And to echo Plush and Rich... Tony- thank you for posting. It is a pleasure to see you here.

--Ben
Old 9th March 2009 | Show parent
  #159
Gear Maniac
 
🎧 10 years
Re. 96k I did not just say 44k1 sounded better than 96k, because it doesn't.

However the main commercial market for recorded music remains 44k1 for CD, downloads, and Flash. If we are in business we can't just make recordings which sound great in the control-room replayed 96k/24 off a hard-drive for us, our clients, and our friends.

I am not alone in believing typical 96k downsampled to 44k1 using conventional brickwall anti-alias filters sounds worse than 88k2 downsampled to 44k1 using a variety of easy methods when the sampling rate is a direct integer multiple. One of my closest colleagues has got so frustrated trying to make 96k originals sound decent downsampled to 44k1for CD release in parallel with 96k downloads that he resorted to upsampling the 96k material to DSD and then resampling DSD back down to 44k1. If your client needs a 44k1 master you need to give him the best he can have to sell to stay in business. If your client wants 96k to sell to customers that is a different matter.

If we just wanted our recordings to sound great and optimally transparent to us in our control-rooms then it would be arguable that a Studer A80, Ampex ATR100 or Telefunken M15 analogue tape machine with a first generation 15ips or 30ips master would sound the cleanest of all. The problem remains - getting the quality to the retail customer.
Old 9th March 2009 | Show parent
  #160
29327
Guest
Quote:
Originally Posted by fifthcircle ➑️
The vast majority of the stuff I do is at 44.1... That said, what is quoted here is very intelligent. You won't know what is sounds like UNTIL YOU DO IT. Then you make a choice based on knowledge which is always the best way.
I heard an interesting quote about this the other day. A friend and I were discussing the topic of sample rates, bit depth, etc., etc., etc. The final quote was "When was the last time you said 'Something's just not right. It's not the mic... It's not the pre... It's not even the mic placement. Let's try a different sampling rate!'" I'm all for increased quality, but you've got to be realistic about it.

Quote:
And to echo Plush and Rich... Tony- thank you for posting. It is a pleasure to see you here.
I'll second that. It's pretty amazing to have such a renowned engineer and innovator posting! Thanks Tony. thumbsup
Old 9th March 2009 | Show parent
  #161
29327
Guest
Quote:
Originally Posted by TonyF ➑️
I am not alone in believing typical 96k downsampled to 44k1 using conventional brickwall anti-alias filters sounds worse than 88k2 downsampled to 44k1 using a variety of easy methods when the sampling rate is a direct integer multiple.
I've found this as well. The amount of processing that need to be done (interpolation and decimation or, in some implementations, multi-stage decimation, etc.) to a non-integer-related sampling rate is quite a bit more distructive than it's n*x:n (where x=an integer) counterparts. I believe there are a few papers on the subject, as well.

Last edited by 29327; 9th March 2009 at 06:35 PM.. Reason: oops!
Old 9th March 2009 | Show parent
  #162
Gear Maniac
 
Mazo Audio's Avatar
 
🎧 10 years
Quote:
Originally Posted by TonyF ➑️
If we just wanted our recordings to sound great and optimally transparent to us in our control-rooms then it would be arguable that a Studer A80, Ampex ATR100 or Telefunken M15 analogue tape machine with a first generation 15ips or 30ips master would sound the cleanest of all. The problem remains - getting the quality to the retail customer.
Right on Tony, nice to have you here! I think you'll find a lot of friends here. I've leaned
some nice tips from interviews with you over the years. Thanks! I appreciate your healthy skepticism of the digital age.

Lance
Old 9th March 2009 | Show parent
  #163
Gear Guru
 
joelpatterson's Avatar
 
2 Reviews written
🎧 15 years
Quote:
Originally Posted by TonyF ➑️
... The problem remains - getting the quality to the retail customer.
But the problem there is that the retail customer wouldn't know quality if it slammed him upside the head with a garden rake.
Old 9th March 2009 | Show parent
  #164
Lives for gear
 
d_fu's Avatar
 
🎧 15 years
Tony,
Quote:
Originally Posted by TonyF ➑️
I am not alone in believing typical 96k downsampled to 44k1 using conventional brickwall anti-alias filters sounds worse than 88k2 downsampled to 44k1 using a variety of easy methods when the sampling rate is a direct integer multiple.
Could you explain how the downsampling differs? Certainly you're not referring to the "dropping every other sample" idea people sometimes come up with...

Daniel
Old 10th March 2009 | Show parent
  #165
Gear Maniac
 
🎧 10 years
With integer resampling to 44k1 you have some simple options like dropping every other 88k2 sample as you say. That boils down to straight 44k1 sampling with no anti-alias filter apart from the chop at half the original 88k2 rate. That is a bit risky unless you know there is no possibility of aliasing. Any electronic music, close miked instruments, sibilants, bright percussion and you'll hear aliasing which is harsh hf intermodulation and sounds like breaking glass when you have problems. Also sometimes on location you can end up recording hf birdie tones anywhere from 12kHz up to 40kHz depending on the junk in the air and in the in-house mic cabling in halls. These birdies can give you intermod products if you get your filters wrong and the birdies are high enough level.

If you add both 88k1 samples for each channel you get a 44k1 end-result with a notch at 22.05kHz in the signal path. The notch is often enough to minimise audible aliasing or eliminate it completely until you have some unusual effect at ehf. There is audio above 15k and 20k in live classical music, but if it is at relatively low level the alias products are only present well below the system noise level and are one heck of a lot less audible than a typical anti-alias filter at 22.05kHz making the sound as dead as mutton.

We do our recordings at 88k2 most of the time and when I am editing I regularly use CEDAR Retouch to remove live acoustic clicks and squeaks from live concerts. That means I look a lot at the CEDAR spectrum screen and the audio above 20k doing classical is mainly clicks, squeaks, sneezes, percussion and occasionally a few harmonics. If we record string music at elevated level because there are no brass or perc to give us heavy peaks, then we see some harmonic content above 15k but the level is not high normally, and if it is you can always scrap the idea of simple filter-light downsampling and go for the old fashioned way of digital filters. Meridian Audio has developed what they call apodising filters and they are much more acceptable audibly than regular DSP brickwall filters. I have heard them in their new digital monitor loudspeakers which sound good. With any luck the technology will turn up somewhere useful for audio engineers like us.

Another advantage of integer resampling is that if you are using physical hardware rather than having all the audio inside a workstation, then everything may be synchronised to one master clock to minimise jitter in the process. Some hardware sample-rate conversion systems will simply do an SRC to whatever is input to the system, jitter and all. Some software which does SRC on the fly will be a bit approximate too, in the cause of convenience.
Old 10th March 2009 | Show parent
  #166
Lives for gear
 
Roland's Avatar
 
2 Reviews written
🎧 15 years
Quote:
Originally Posted by joelpatterson ➑️
But the problem there is that the retail customer wouldn't know quality if it slammed him upside the head with a garden rake.

Whilst this may be true, in many way's things have improved, I remember when the average hi-fi be ing sold in this country was an all in one Amstrad with record deck, tuner, double cassettte. These were truely awful, even by that times standard. When people complain about digital audio and CD it is easy to forget about how much better things are. To compare CD to esoteric vinyl playback equipment that costs several thousands is not indicative of the standard several years ago as but a handful of the public had those systems. For all that we say about digital and I understand well the points that Tony F is making, CD may chnage the sound, but it is very close, very close indeed. This is something that had never been available to the general public until CD was released around 1982. Reel to reel analogue tape, (certainley of the order we in the profession had access too), was never a practical solution for the general public, perhaps 96/24 would be a better solution, however it is easy for us to talk about that now, not an option technologically 27 years ago.

Regards



Roland
Old 10th March 2009 | Show parent
  #167
Gear Addict
 
🎧 15 years
I highly appreciate the SR discussion, a slight deviation related to the upcoming recording.

Given a mic preamp pot of about 60dB gain range if my mic array with a typical LDC mic
will be up 3.6M and approx 3-4 meters from the musical performance and given an approx max SPL of about 80dB SPL at the mic where would be a "I have just set up" gain position be on my preamps. Obviously this can vary to some degree dependent on pre, actual peak SPL etc. etc. but I would imagine it to be 9 or ten o clock position? does that seem about right?

Obviously I will have to find out on the day but a starting point would be handy.

Thanks
Old 11th March 2009 | Show parent
  #168
Lives for gear
 
Russell Dawkins's Avatar
 
🎧 15 years
This is impossible to answer without knowing more specifics, but I find if I must record without a soundcheck, setting the gain so that pre-performance noise from the audience shows roughly -50dB is usually a safe starting point, unless the piece is horrendously loud.

Never make a gain adjustment in the middle of a movement, wait till the interval between the first two movements and make a careful note of when the adjustment was made and how much. This makes post production easier.
Old 11th March 2009 | Show parent
  #169
Lives for gear
 
sonare's Avatar
 
2 Reviews written
🎧 15 years
I would like to steer things back to SRC for just a moment. For years I did my HD work at 88.2 because it was an integer multiple, and we all know that there are fewer numbers involved when dividing by 2.

Then I kept reading stuff by intelligent folk saying that modern SRC upsampled before downsampling so it didn't matter if you started at 88.2 or 96. To top off the confusion I could not get a straight answer from the Sequoia US tech support guy about whether it upsampled 88.2 first or not. e does not know for sure. Perhaps a moot point, as I use R8Brain Pro for SRC.

Would the erudite among us care to weigh in? Inquiring minds want to know!

Rich
Old 11th March 2009 | Show parent
  #170
Gear Addict
 
🎧 15 years
Fair enough I will do it myself, just wondered if there was a starting point
of value. I will do as I initially thought and I have many hours of rehearsal before to set levels up. Thanks for the suggestion on the audience experiences but that seems somewhat impossible to gauge a performance level by audience ambient noise. Thanks though.

I use R8brain but not the pro version.
Old 11th March 2009 | Show parent
  #171
Lives for gear
 
d_fu's Avatar
 
🎧 15 years
Quote:
Originally Posted by sonare ➑️
I would like to steer things back to SRC for just a moment. For years I did my HD work at 88.2 because it was an integer multiple, and we all know that there are fewer numbers involved when dividing by 2.

Then I kept reading stuff by intelligent folk saying that modern SRC upsampled before downsampling so it didn't matter if you started at 88.2 or 96.
That's correct. And therefore there shouldn't really be any difference in the process between 88->44 and 96->44, at least not because of "fewer numbers involved".


Daniel
Old 11th March 2009 | Show parent
  #172
Lives for gear
 
Corran's Avatar
 
🎧 10 years
For SRC I've always done my final bounce in Sonar to 44/16 after working in usually 48/24. I am wondering if I'd have any major improvement using a dedicated SRC program. I never have noticed a difference between the 48/24 and 44/16 bounce.
Old 11th March 2009 | Show parent
  #173
Lives for gear
 
sonare's Avatar
 
2 Reviews written
🎧 15 years
Quote:
Originally Posted by d_fu ➑️
That's correct. And therefore there shouldn't really be any difference in the process between 88->44 and 96->44, at least not because of "fewer numbers involved".
Daniel
Due to the reigning confusion I must ask the source of your info. And for others, lets not confuse SRC (96>44.1) with word-length reduction {24>16}.

This site SRC Comparisons pretty well tells who is naughty and who is nice.

Rich
Old 11th March 2009 | Show parent
  #174
29327
Guest
At the risk of having my head bitten off, I'll give my few cents on the SRC debate (the topic of my master's thesis and an AES poster session).

I ran into a lot of the same problem as are stated above: none of the SRC companies want to tell you what's inside their little black box. So, for my research I had to go back in time a bit; back to the beginning. So, in addition to the straight-up comparison found on the Infinitewave site, I found the following resources to be very helpful.
  • Multirate Digital Signal Processing by Crochiere & Rabiner - a seminal work on the topic of SRC
  • The work of Roger Lagadec (R&D for Studer) circa 1983 - a few AES papers dealing with requirements for a servicable SRC algorithm/processor and various implementation methods
  • Julius O. Smith's Digital Audio Resampling Page

All this is to say that, in principle, there doesn't need to be any upsampling in an integer multiple conversion; only interpolation between n points to be reproduced as a single sample value. In practice, however, this can be achieved by upsampling to the the integer multiple and then taking the samples necessary to fit the new sampling rate. Different terminology and implementation, but the same interpolated sample will be output.

You know, I never thought I'd say it, but I wish GS had equation support in posts...
Old 11th March 2009 | Show parent
  #175
Lives for gear
 
David Rick's Avatar
 
🎧 15 years
It's the filter design, not the downsampling ratio!

I would say that the two best modern books on multirate signal processing are these:

Multirate Systems and Filter Banks, by P. P. Vaidyanathan


Multirate Signal Processing for Communication Systems, by fredric j. harris

Vaidyanathan's book is full of equations, and harris's is full of pictures, but both are written for DSP specialists. Neither is really accessible to practicing audio engineers, unless they happen to have taken a college-level DSP class from a good engineering school.

The idea of upsampling to the least-common-multiple of the input and output sample rates is not wrong in principle, as long as one understands polyphase implementations don't really do this in practice. They switch between the outputs of a bunch of short filters whose outputs are as if the input signal had been upsampled, filtered, and downsampled. But each of these filters is essentially the same length as the (single) filter one would use in the integer ratio case, so there really isn't any more math being done, just more bookkeeping to choose the correct filter for each output sample.

It follows that if there's an audible degradation in downsampling, it isn't because the downsampling ratio is rational vs. integer, it's because either the interpolation filter design, or its implementation, is bad. The correct strategy is to design a better interpolation filter, but "better" needs to be informed by human perceptual models.

I happen to think that when Tony Faulkner reverts to two-sample averaging, or even worse, no anti-aliasing filter, he's throwing the baby out with the bathwater. Somewhere between 2000-tap FIR filters with bad pre-echo problems, and 2-tap "filters" with only 6 dB of alias rejection, there must be a happy medium! But the best compromise undoubtedly depends on the particular music being processed.

Steve Remote: Can we please split this sample rate conversion discussion into a separate thread? If that happens, I promise to post some pictures of what Tony's 2-tap averager is actually doing. (It may be ok in his particular recording situation, but in general it ain't pretty.)

David L. Rick
Old 11th March 2009 | Show parent
  #176
Lives for gear
 
MichaelPatrick's Avatar
 
🎧 15 years
Quote:
Originally Posted by BLP ➑️
You know, I never thought I'd say it, but I wish GS had equation support in posts...
How about saving it in a graphic and posting it as a picture?
Old 11th March 2009 | Show parent
  #177
Lives for gear
 
MichaelPatrick's Avatar
 
🎧 15 years
Yes! I'm also interested in learning more about SR conversion but leaving the discussion in this thread would be to hide it.

Quote:
Originally Posted by David Rick ➑️
Steve Remote: Can we please split this sample rate conversion discussion into a separate thread? If that happens, I promise to post some pictures of what Tony's 2-tap averager is actually doing. (It may be ok in his particular recording situation, but in general it ain't pretty.)

David L. Rick
Old 11th March 2009 | Show parent
  #178
Lives for gear
 
🎧 10 years
Wow, this is an amazing thread. I am learning something new with every post.....
Old 11th March 2009 | Show parent
  #179
29327
Guest
Ah, yes, the Vaidyanathan book! I gave it a peruse, but it was pretty hardcore... My understanding starts to fade somewhere just a bit after the Cochiere, DAFX, etc. I've worked through some of the math, but it took longer than I was willing to spend after my thesis was complete... I'll have to give the Harris book a look, though!

Quote:
Originally Posted by David Rick ➑️
The correct strategy is to design a better interpolation filter, but "better" needs to be informed by human perceptual models.
I think that this is really where things get disconnected. Too many developers focus on specs and metrics to gauge performance; this just not sufficient. Metrics can help us to understand what we're going to hear, but we don't have a full grasp of the ear-brain apparatus, so a graph is still just a graph. In my study, the moderately trained listener showed very little distinction from algorithm to algorithm, but there was a slight preference for a certain algorithm, and it wasn't one who's graph looked terribly pretty...
Old 11th March 2009 | Show parent
  #180
Lives for gear
 
Russell Dawkins's Avatar
 
🎧 15 years
Quote:
Originally Posted by joelpatterson ➑️
But the problem there is that the retail customer wouldn't know quality if it slammed him upside the head with a garden rake.
That's not my experience at all.

They may well not know what is behind the process, but in my experience a recording of genuine quality is appreciated by almost anyone and sounds better played back even on the lowest grade system (I'm talking clock radios).
πŸ“ Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 108 views: 11873
Avatar for d_fu
d_fu 3rd April 2007
replies: 22 views: 22856
Avatar for Jbassist
Jbassist 1st November 2017
replies: 90 views: 7996
Avatar for rakim87
rakim87 16th November 2012
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump