|
Post by svart on Apr 5, 2017 6:23:04 GMT -6
So are there any modern converters (AD/DA) that don't use anti-aliasing filtering? If we are going to talk about implementation I'm all for it, I'd just love to know what the solution to Nyquist looks like in 2017 or are we still using a sort of filter system there? Truly have no idea, would love to know. Tried to Google it before, not a very openly discussed topic, so I'm pulling up at the Cantina Mos Eisley to find out. -L. Some designers are doing DSP filtering instead of hardware filtering. They use extremely wide bandwidth converters and then digitally filter to get steeper filter rolloffs. That being said, I don't personally hear much difference between those and a properly implemented hardware filter, but others swear that there is a huge difference. To each his own I suppose.
|
|
|
Post by Johnkenn on Apr 5, 2017 8:24:40 GMT -6
I would throw my hat on the the AD side... once tracked you can swap out DA's but you are stuck with what you have got... having said that its all important though... we all know that. cheers Wiz But how do you know what you're tracking actually sounds good? I'm kidding, but it IS the chicken and the egg argument.
|
|
|
Post by Johnkenn on Apr 5, 2017 8:28:09 GMT -6
And we are nerding out, of course, but it's an interesting philosophical discussion.
|
|
|
Post by NoFilterChuck on Apr 5, 2017 8:36:53 GMT -6
There is a lot to be said for folks that produce music without any recorded audio. Sample Libraries + synths only. in those situations, your DA is all that matters because you aren't using your AD at all lol
|
|
|
Post by Bob Olhsson on Apr 5, 2017 9:57:24 GMT -6
I started out thinking A to D was more important but at this point believe the monitoring D to A may be even more important. The real test of an A to D is what happens when you apply digital signal processing such as boosting the top end.
|
|
|
Post by mjheck on Apr 5, 2017 10:00:58 GMT -6
There's a Simpsons where Homer is digging for "Lincoln's gold." Finding no gold, he complains that he followed the treasure map perfectly. Marge exasperates that, while he did follow the map, he continually started from some arbitrary point, like where he happens to be standing.
This is kind of how I view monitoring.
Sure I can make all the right moves, but if my starting point - what I am listening to - is flawed, then who cares what I did to get there?
I guess this is not just about DA, but it does seem to be right there with monitor or headphone selection. I've never had more than four DA options in the studio at one time, but the differences in just those was kind of shocking. I thought they all sounded good, but they did all sound different enough to where it changed mixing decisions.
For example, I had my 2192, an Antelope Zodiac and two different BLA modified Apollos (an 8 and a twin). I loved the sound of the 2192 and the Antelope, but frankly, the mixes done through the BLA modded Twin translated better. That became my dedicated DA, and the others were used for round trip analog stuff.
MJH
|
|
|
Post by illacov on Apr 5, 2017 12:49:28 GMT -6
So are there any modern converters (AD/DA) that don't use anti-aliasing filtering? If we are going to talk about implementation I'm all for it, I'd just love to know what the solution to Nyquist looks like in 2017 or are we still using a sort of filter system there? Truly have no idea, would love to know. Tried to Google it before, not a very openly discussed topic, so I'm pulling up at the Cantina Mos Eisley to find out. -L. Some designers are doing DSP filtering instead of hardware filtering. They use extremely wide bandwidth converters and then digitally filter to get steeper filter rolloffs. That being said, I don't personally hear much difference between those and a properly implemented hardware filter, but others swear that there is a huge difference. To each his own I suppose. So based on this answer, we still filter information above a certain frequency to make audio friendly to converters? Thanks -L.
|
|
|
Post by Johnkenn on Apr 5, 2017 13:28:02 GMT -6
Thanks to everyone for a very civil discussion! I should have put a "?" at the end of the thread title...meant it that way.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 5, 2017 21:12:36 GMT -6
That's a really interesting--and loaded--question. In one way it's just like your choice of studio monitor: if you trust your gear and understand the limitations, then you know that your mixes will translate. But it really is a tricky question. If you're cranking out a lot of commercial mixes, then the argument might skew toward D/A. What's recorded is recorded and your biggest concern probably centers around hearing the same thing the mastering engineer hears.
But I'll use photography as something of a counter-argument. Before I found myself running a company, I actually had time to do serious amateur photography. As soon as I knew one end of the camera from the other, I began shooting everything in the camera raw format. That turned out to be a good choice. As my skills improved (and Photoshop became more powerful) I found there was more color, more detail and more contrast in those old shots. I was able to get much better prints from the same old files.
That could mean that the preamp+A/D choice might be more important if you're doing anything you consider to be archival. I record a different kind of performance than most of the folks on this list (thanks for humoring me) and they're performances that might not happen again for years and years. So I record at high sample rates and the purest input chain my budget allows. Sure, I deliver 44.1/24-bit, but I know I've got a lot better in the vault if there's ever an opportunity to use it. So I might skew a little closer to the 'capture' side of the equation.
But it really is quite an interesting question, isn't it? Luckily for us, even budget gear is far better than top-of-the-line from just a few years ago.
|
|
|
Post by massivemastering on Apr 5, 2017 23:39:26 GMT -6
Every single decision you make (tracking, mixing, mastering, etc.) is based on your monitoring chain from the DA on down. Imma go with "really important" on this one.
|
|
ericn
Temp
Balance Engineer
Posts: 15,015
|
Post by ericn on Apr 6, 2017 9:53:24 GMT -6
I'm sticking to equal I need quality AD to get it . I need quality DA to know I got it. Useless to get it and not know it . Just as bad to not be able to get it but know it .
Equal
|
|
|
Post by illacov on Apr 6, 2017 10:12:42 GMT -6
I'm still trying to get my head wrapped around why there's less open discussion on how to improve the implementation and development of digital conversion.
Is there any research on anti-aliasing filtering and its impact on harmonics?
Its just bizarre to me that we have all this gear that generates varying degrees of harmonic distortion which to our ears offers more detail with than without and yet we use converters that somehow zap a large portion of that same information.
So what gives? Even our DA uses this so, it's not like you can escape it LOL Is it technically impossible in 2017to use digital conversion without the filtering (using analog components or DSP?). And why not?
Thanks -L.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 6, 2017 12:46:30 GMT -6
Uhm, AFAIK (and by no means i am a specialist in this field) the anti-aliasing filters are there to prevent aliasing artifacts -in the audioband- that appear when frequencies above the nyqist limit are present. This is due to the sampling process itself. It is a bit like when you look at wheels in a film that has 24 pictures per second (Hz). If the wheel turns faster than nyquist limit, the wheel appears to be turning *slower* backward or forward, but of course this is NOT the real speed/frequency, it is the aliasing in the critical band. (Therefore lowpass is not optional) Hm. I don't know if the answer is a)understandable and b)if we talk about the same filters?
I would be happy if some of our specialists could chime in here...
|
|
ericn
Temp
Balance Engineer
Posts: 15,015
|
Post by ericn on Apr 6, 2017 13:12:55 GMT -6
I'm still trying to get my head wrapped around why there's less open discussion on how to improve the implementation and development of digital conversion. Is there any research on anti-aliasing filtering and its impact on harmonics? Its just bizarre to me that we have all this gear that generates varying degrees of harmonic distortion which to our ears offers more detail with than without and yet we use converters that somehow zap a large portion of that same information. So what gives? Even our DA uses this so, it's not like you can escape it LOL Is it technically impossible in 2017to use digital conversion without the filtering (using analog components or DSP?). And why not? Thanks -L. My guess more $$$ in cheap and tiny than improved quality!
|
|
|
Post by indiehouse on Apr 6, 2017 13:35:21 GMT -6
So, this is an interesting topic that I've actually struggled with myself.
I'm using a (highly modified) Ross Martin PCM4222 AD and a (highly modified) Ross Martin Superbeast II into a BF UAD Apollo 16 MKII.
When I'm tracking drums or full band, I'm using the Apollo's converters and monitoring through the Superbeast. After that, though, I'm only using 1 or 2 channels for overdubs, tracking through the RM AD and monitoring through the RM DA.
But here's where it starts to break down for me. I do a lot of hardware printing come mix time. Do I setup my HW insert through the Apollo's AD/DA converters and monitor through the Ross Martin DA? Do I setup my HW insert through the Ross Martin AD/DA and monitor through the Apollo's DA?
What's the correct answer here? It seems that I'd be nulling any benefits of tracking through the RM converters by then using the Apollo's converters when printing HW. But the latter option leaves me monitoring through the Apollo's DA.
I've also tried setting up HW prints using the Apollo's DA but coming back in through the RM AD, while monitoring through the RM Superbeast. However, I can't use this combo as a HW insert. I have to route it out through the Apollo's DA then back in through the RM AD on a separate track, which induces latency. That means I can't play back the track in the context of the mix, which leaves me EQ'ing/compressing the track in solo, which is no bueno. It seems it's an either or situation. I either use the Apollo's converters on both sides, or the Ross Martin converters on both sides.
Maybe I'm overthinking it. I've got the BF Apollo 16, so unless I change platforms, I've got the best converters in the Apollo line. But, if I'm going through the trouble of tracking through the RM AD, would I be doing damage by printing HW using the Apollo's roundtrip AD/DA converters?
I suppose the Apollo's DA is pretty good, so I could probably monitor through that and print through the Ross Martin converters. But there is a difference. There is more depth/3D to the RM DA. I wouldn't want to lose that by printing through the Apollo's DA. Not that the Apollo's DA sucks or anything. It doesn't. It's a friggin' $3K interface.
Which is more important in my case? What would you do?
|
|
|
Post by Guitar on Apr 6, 2017 13:44:14 GMT -6
I would do either one and not worry about it. It's easy to get tied up in this stuff, tangled even. In the end you're tweaking stuff nobody else will hear, for the most part. Like you said, you've got two top of the line boxes so it seems like not much to lose.
Anyway, this is a really fascinating thread. The question seems like sort of a litmus test that shows surprisingly a lot about how people work differently.
|
|
|
Post by sozocaps on Apr 6, 2017 14:27:28 GMT -6
According to Apogee
|
|
|
Post by Bob Olhsson on Apr 6, 2017 15:56:18 GMT -6
Really clean filters are longer and have higher latency than people like for overdubbing so there's a significant tradeoff.
|
|
|
Post by Guitar on Apr 6, 2017 17:34:14 GMT -6
No offense to the Apogee product, but that was painful to watch.
|
|
|
Post by illacov on Apr 6, 2017 18:35:23 GMT -6
Uhm, AFAIK (and by no means i am a specialist in this field) the anti-aliasing filters are there to prevent aliasing artifacts -in the audioband- that appear when frequencies above the nyqist limit are present. This is due to the sampling process itself. It is a bit like when you look at wheels in a film that has 24 pictures per second (Hz). If the wheel turns faster than nyquist limit, the wheel appears to be turning *slower* backward or forward, but of course this is NOT the real speed/frequency, it is the aliasing in the critical band. (Therefore lowpass is not optional) Hm. I don't know if the answer is a)understandable and b)if we talk about the same filters? I would be happy if some of our specialists could chime in here... There are actually digital cameras with no AA filtering. This is an old article but it illustrates my thinking. They increased the resolution and offered the end user the option of dealing with aliasing in post rather than at the camera level. www.outdoorphotographer.com/photography-gear/cameras/can-you-go-no-low-pass/I'd love to know what the frequency response of our beloved analog gear actually is. Sure it's beyond the range of human hearing but that's like saying you don't get impacted by solar radiation because you can't see it. It's a wave right? Thanks -L.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Apr 6, 2017 23:36:59 GMT -6
Uhm, AFAIK (and by no means i am a specialist in this field) the anti-aliasing filters are there to prevent aliasing artifacts -in the audioband- that appear when frequencies above the nyqist limit are present. This is due to the sampling process itself. It is a bit like when you look at wheels in a film that has 24 pictures per second (Hz). If the wheel turns faster than nyquist limit, the wheel appears to be turning *slower* backward or forward, but of course this is NOT the real speed/frequency, it is the aliasing in the critical band. (Therefore lowpass is not optional) Hm. I don't know if the answer is a)understandable and b)if we talk about the same filters? I would be happy if some of our specialists could chime in here... There are actually digital cameras with no AA filtering. This is an old article but it illustrates my thinking. They increased the resolution and offered the end user the option of dealing with aliasing in post rather than at the camera level. www.outdoorphotographer.com/photography-gear/cameras/can-you-go-no-low-pass/I'd love to know what the frequency response of our beloved analog gear actually is. Sure it's beyond the range of human hearing but that's like saying you don't get impacted by solar radiation because you can't see it. It's a wave right? Thanks -L. Oh, photography is not really an analogy to audio recording. The series of snapshots/samples is, what makes the aliasing a real problem, therefore i used the film analogy...
|
|
|
Post by illacov on Apr 6, 2017 23:47:29 GMT -6
There are actually digital cameras with no AA filtering. This is an old article but it illustrates my thinking. They increased the resolution and offered the end user the option of dealing with aliasing in post rather than at the camera level. www.outdoorphotographer.com/photography-gear/cameras/can-you-go-no-low-pass/I'd love to know what the frequency response of our beloved analog gear actually is. Sure it's beyond the range of human hearing but that's like saying you don't get impacted by solar radiation because you can't see it. It's a wave right? Thanks -L. Oh, photography is not really an analogy to audio recording. The series of snapshots/samples is, what makes the aliasing a real problem, therefore i used the film analogy... Last I checked I thought anti-aliasing filtering was related to digital sampling of data not strictly related to analog to digital conversion for audio, however my inquiry/pointed statement revolves around letting us decide how we want to handle aliasing in post. But it is important to note that there are filtered sampling tech for cameras, so why not for sound is my point? They are talking 50 megapixel sensors for crying out loud! That's beyond the resolution of the lenses itself. That's some promising specs in my book. Let me sample audio like that please and thank you LOL. Thanks -L.
|
|
|
Post by donr on Apr 7, 2017 1:41:00 GMT -6
Great thread. I'd intuit a/D, but how you going to hear it without d/A? Faith, to a degree, and user reviews, although you can judge a good d/A with reference recordings you know. Probably a good idea to confidently get to there before choosing an a/D.
I think much of what we're sonically chasing here on RGO is the euphonics of analog. Recording music exactly like you would hear it live is a wonderful goal, but even if you could, it's not practical as a playback medium for 99% of the way recorded music is consumed. I can't actually come up with a listening situation where I want to hear music with as much dynamic range as it was created live. I'm totally comfortable with peak to RMS DR 16-18 up to about 8, depending on the material.
When digital came in, early CD's would have a three letter code telling you how it was recorded. D or A, at the initial recording stage, the mix stage, and the delivery stage, CD. Buying some early DDD classical records, other than the super quiet noise floor, I was underwhelmed by the sound of those recordings.
A particular disappointment was the DDD CD version of Telarc's "Frederick Fennel, The Cleveland Symphony Winds-Holst/Handel/Bach" which I had ADORED on the vinyl pressing. In that case, the orchestral recording was digital, but the initial disk mastering and delivery was analog, it was a vinyl record. Just gorgeous. When the CD came out a while later, I bought it and thought it sounded no where near as good as the vinyl, in fact it was pretty awful.
In hindsight, I can only assume the euphonics in that case was the disk cutting and my humble analog (Shure v15, Dual TT, Fisher reciever, Advent spkrs) playback system. But it was night and day.
A/D converters have come a huge way from the SONY 1610, and recordings have improved tremendously since the early days of consumer digital.
But at least in the case of that Telarc Cleveland Sympony Winds recording, the magic was not in the A/D, which must have been primitive digital by today's standards, but in the analog delivery and reproduction stages.
Certainly you couldn't go wrong with the best A/D conversion both in format and unit, you can use. For archival if no other reason.
|
|
|
Post by Guitar on Apr 7, 2017 5:27:55 GMT -6
Last I checked I thought anti-aliasing filtering was related to digital sampling of data not strictly related to analog to digital conversion for audio, however my inquiry/pointed statement revolves around letting us decide how we want to handle aliasing in post. But it is important to note that there are filtered sampling tech for cameras, so why not for sound is my point? They are talking 50 megapixel sensors for crying out loud! That's beyond the resolution of the lenses itself. That's some promising specs in my book. Let me sample audio like that please and thank you LOL. Thanks -L. Without the reconstruction filter you would have a "stair step" waveform. The filter "smooths" the steps into a proper representation suitable for listening. From my understanding it is not "optional" as you seem to think, and rather an integral part of how a D to A converter works at the component level. It's not just something you can leave off, like the brakes in your car, or the flush lever on your toilet. It is part of the process of turning binary data on a hard drive into beautiful audio for the air. Which are two very dissimilar things so frankly I find the process incredible. Even though I struggle to understand it and still am learning. Also I'm referring to the DAC process, I am not precisely sure how the ADC process differs. Furthermore, at high sampling rates like dual and quad, the remaining bandwidth still reaches way above the human hearing range.
|
|
|
Post by Guitar on Apr 7, 2017 6:17:03 GMT -6
OK I just took a quick look at the Analog to Digital Wiki (LOL) and the low pass filter there seems intended to keep aliasing side band artifacts out of the audio. The "anti-aliasing filter." Also oversampling can be used within the ADC to push these upper limits even higher for better performance. You can approach the clock speed of the transistor circuitry, some up to 300 Megahertz currently, which is vastly higher than 44.1 kilohertz for example. You could express that sample rate as 300,000 KHz! Then a very high digital filter would be applied.
Specifically in the sigma-delta ADC, which I believe is the most common type, the "decimation filter" is integral to the process, just like the reconstruction filter in the DAC mentioned above. I am not sure if this is just a different term for the anti-aliasing filter, which I think it might be. It is the opposite in function of the reconstruction filter in the DAC.
I apologize for my enthusiasm, I am just really curious about the nuts and bolts of digital audio.
|
|