Coming up with a title for this thread that really summarized the question I want to ask without being inflammatory was more challenging than I thought. I guess we'll see if I was successful.

Up until two years ago I had zero experience with analog tape apart from my initial venture into the art and science of recording using a cassette 4-track. I was knee deep in digital gear at my studio and the DAW was the cornerstone of it all. On a whim I purchased a tape deck one day and then everything changed. A whole new world of sound revealed itself to me and my little digital world imploded. I have been slowly rebuilding my world and redefining what it means to record music in my studio. Despite all this I recognize that my output must interface with the rest of the world, which is largely digital. I often convert my tracks and mixes into the digital realm, however the results leave me dissatisfied as I never hear the same reproduction from the D/A post-conversion that I do directly from the tape deck. Hence my question:

Why, in the year 2010, are PCM A/D and D/A converters unable to capture and reproduce an analog source recording as a 1:1 copy?

I'm really hoping that instead of this turning into another analog versus digital debate, that we can generate some discussion (hopefully technical in nature) that really examines what is happening at the converter level that is changing the source signal. Maybe some converter designers will chime in and help me get to the bottom of this.

Some of my specific questions:

1. When I hear the sound change between source and dub (loss of depth and spatial info, flattening of "realness"), what artifacts am I technically hearing? Is it ringing? Aliasing? Are they a result of A/D or D/A?
2. What can designers do to eliminate these artifacts?
3. Are these artifacts a function of PCM technology? Or are they a function of design implementation? In other words if the converter is expensive enough, will it perform an unadulterated transfer?
4. If design implementation really is the root cause and doing it properly is a matter of cost, what needs to happen to bring the cost down while keeping the performance high?
5. If DSD is the answer, has anyone looked at doing real-time DSP conversion of 1-bit output to PCM so that this could all work with PCM-based DAWs?

My final thought is a hypothesis I would like to pose:

PCM conversion can make an accurate 1:1 transfer of audio when the original source is digital. When the source is analog, artifacts are generated during the conversion process.

Thanks in advance for all intelligent, civil, and thought provoking discussion.

Brad

[edit] For those that don't want to wade through all 20-something pages (there's a lot of tangential discussion and condescending misintepretations of the question along the way), here are two responses that I really liked:

https://gearspace.com/board/5575220-post443.html

https://gearspace.com/board/5586817-post624.html

brad, how is the radar working out for you?

hopefully this thread wont get ugly, because i would really like to know more about this topic.

my take on this is that one way or another, recorded music playing off speakers is an illusion. it is trying to appear to be something it is not in order to make the listener feel one thing or another.
to my ears, analog tape creates a more effective illusion with less fuss, but digital recording (GOOD sounding digital) can yield a great illusion as well, it just takes some extra work.

it makes sense to me that most people would prefer the sound of an analog recording because the process of converting things in and out of digital changes audio in ways that take it further away from what our ears are designed to hear and our brains are evolved to perceive.

The Radar is working out well and I'm hoping within in the coming weeks to be able to use it to create some sound clips that put my hypothesis to the test.

I'd really like to steer this conversation away from discussing people's preferences for one medium or the other or why they think one medium allows them to achieve bigger/better/fatter/faster/prettier/blacker/whiter results. I'd really like to stay focused on what is and isn't happening to a source analog recording when it passes through A/D and D/A conversion, and why the result of that process does not match the input source.

I'd like to respectfully ask those that do not have first hand experience participating in transfers of analog recordings through digital converters to not incite an analog/digital riot with regurgitations of 2nd hand internet techno-babble, or monologues about things they wouldn't feel comfortable basing their credibility on. heh

thanks,
Brad

I think that the problem isn't found or have to be searched around the AD/DA converter.
The "real" problem is how "digital language" is build.
Digital is or ON or OFF / 1 or 0. There isn't a thing between those 2. So even with hyper modern AD/DA 32bit/384 or DSD. You will have always that "artifact" steps of On/Off.

This is a great thread, and forgive me if this is totally dumb, but how scientific a compare are we talking? ie:

Analog Tape Out -> Poweramp -> Speakers -> Ears

compared to:

Same Analog tape out -> A/D -> D/A -> Same Poweramp -> Same speakers -> Same ears.

Like the only thing different in the chain being the A/D, D/A conversion...

again... sorry if this is dumb.

Brad, if you're actually after detailed technical explanations of converter design and implementation as opposed to subjective debate, then a forum isn't really the right place to do it...

Amazon.com: Principles of Digital Audio (9780071441568): Ken Pohlmann: Books

Why not read the chapter in this about sigma-delta conversion and actually understand what's happening under the hood? Should lead you to answers for most of your questions... it's not overly technical either.

Just like microphones, mic pre's, Eq's and compressors converters have their own sound, character, and or personality. So when you run a signal through converters going in and out, it will impart something to your signal. You should also get different results when you use 96 compared to 44. In addition it will be different from Lavry, Benchmark, Apogee or Lynx. Radar sounds a little more analog as does Apogee, while Mytek or Lavry will give you more detail in the top end and a little more enhanced as it's microscope power offers a little more detail that you would not hear with your original source since it brings more things to the surface.

To make matters simple, its like the difference between a film and real life. A film take slices of time, as where real life is continuous. Digital audio is a slice of time and analog is just like life, continuous. If you forget the problems with filters IE the low pass filters you are left with just that. An audio event is a continuous one and digital tries to represent that with slices. It's IMVHO as simple as that.

Thanks...

Do you also believe that the signal is still square when it leaves the D/A converter? Digital audio can have an infinite depth, although a finite noise floor thanks to dithering.

Ugh... really? We're on the 'digital audio is like a jagged jpeg' argument again?

Skip. If I took a pencil and I put two points on a white piece of paper, could you use a geometric compass, align on those two points and draw a continuous 'analog' circle?

If I put more points along the circle would that help you to draw it any more accurately?

This is crude analogy to how digital is reconstructed to continous analog audio. It's not like a flip book.

hmmm

interesting hypothesis...one interesting way to look at it is that there are no digital sources, which may actually be your whole point.

I take it you mean that if you start with a digital source, convert to analog and then back to digital (and back to analog to hear it) good converters will sound the same...

D>A>D>A

A=A sound wise

I take it you also mean that if you start with an analog source, convert to digital and then back to analog, it will sound different...

A>D>A

A does not equal A sound wise...

----

This is certainly an interesting claim. I can't say I've experienced the same thing in my studio, but I also can't say I've done the kind of careful testing you've done. This statement would assume that PCM is somehow inherently flawed... that the actual math behind it is flawed and that a different technology/set of maths would be required to acquire complete transparency. Or that maybe the execution of PCM hasn't quite achieved the 1:1 transfer it's theoretically capable of doing. But this begs the questions, why 1:1 with digital and not with analog in your experience?

Curious... when you bounce a desk mix down to 1/4" to you notice an equivalent change? Is the depth kept? When you bounce down to digital do you notice this depth change regardless of the converter you use? (Mytek, Radar, etc)?

A reversal question for Brad: When bouncing digital files to tape, do those resulting files on tape ALSO have this lack of depth you experience? Or have you tried this?

Also, it seems there is more to the overall PCM process than just math...I mean you are transforming one reality into another one...

I will also say this: listening to analog synths ALWAYS sounds better to me than listening to the Digital recording of them. The sound is more animated and moves around and conjures up mojo, voodoo, beauty and power all at once...once it hits the 'puter it is flattened out. I mention this because that is where I can really notice things.

This is a really cool thread idea and I really do hope that it stays civil.

I'm understanding Brad to say that, even if all pleasing tape artifacts were removed from the equation, tape is still deeper, more "real" and a better documentation device of spatial detail than digital. In other words, tape is a better capture of reality than digital is. Now to figure out why that is... Now let's keep the discussion going.

Mike at ATR explained it to me in one sentence. Tape holds way more information then digital can.Digital can't store anywhere near as much of the audio picture as tape does.Would that be called resolution? I don't remember the term he used at the moment?Thats why tape sounds so magical.Hard drive technology can't hold near as much info as tape.

I most likely agree. I'm pretty sold on the idea that it's not a problem with the math but I'm still not %100 sure. It's probably gotta come down to implementation but where that breakdown in the implementation occurs is the real question. The thing is though that top of the line converters use analog components every bit as good or better than the classic tape machines that everyone holds up as the best so I don't think implementation of the analog compenents themselves, in the analog realm, is the problem.

The problem's gotta lie where the analog is actually converted into digital, which would either point to the converter chips themselves or to the methodology (sample rate, clocking, jitter, filtering, aliasing, etc.) which is used to encode digital on these chips, DSD vs PCM being an example of this debate.

Now I'll say this though, I notice that things tend to sound better when I use higher sample rates. Why exactly is that? That would seem to point a finger in the direction of less than infinite resolution, which tape has and digital does not. This creates quite the quandary because my scientific side wants to say that the math should be perfectly fine, yet I still hear a difference.

this i believe may be at the core of the issue.

when sound gets converted from an analog waveform to digital 1's and 0's, then back from digital to the analog wav that we hear, some of the information that is in the original analog signal is lost in the process of truncating everything to 1's and 0's. it is certainly perceivable to most people and i am not sure digital will ever be able to do exactly what analog does because of a fundamental difference that cannot be overcome.

cool thread no less!

The almighty tape has it's flaws too. I had a song recorded on tape and I transferred it to another reel to reel and it lost a little in the top end. Depending on the quality of the reel to reel machine, the tape speed, age of tape, how many passes of tape went over the tape heads, when the last time the heads were demagnetized, wear on the tape heads, and even the tape you use are come into play. None if these issues you have to deal with in digital. Not to mention how many times you played the song on the original tape.

It's dumb at all. It's a great opportunity for me to clarify a little.

What you have described is exactly what I am referring to when I say that the analog source doesn't sound the same straight off the tape as it does after it has been converted. The only thing in the signal path that is different the digital converter.

On my car ride home I was thinking a bit. It would be fairly straightforward to prove whether it's the digital or analog filters affecting the audio quality. Hypothetically one could build any analog filters that are part of a DAC and insert them into the analog signal path to confirm or eliminate them as root cause. Has anyone ever done anything like that? Not that I'm volunteering...

Brad

In a 24 bit recording the noise floor is far lower than even really good analogue tape machines. You should have MORE resolution digitally.

Also the whole steps argument is a little ridiculous at this point. Nyquist's theorem explains with any number of samples per second you can PERFECTLY recreate a frequency of half that amount. Hence with a 48k recording you can PERFECTLY recreate any sound up to 24khz. It's not steps, it's not some close estimation.


I think there are two main things preventing us from getting anywhere close to an answer to this question. First off neither the digital nor analog machines we're using are anywhere close to perfect. They all color the signal in one way or another. There's only really three ways to do this test. One is playing an analogue recording and capturing it digitally, then comparing them. That's what we all default to since we just assume that digital should be perfect. However what we're testing there is not if they're potentially similar but if the digital itself is transparent. If we were to either play a digital recording and capture it on analogue tape or record to both mediums simultaniously we would never expect things to be identical because we KNOW that all tape machines color the sound. It's an unavoidable part of the process. Why don't we assume the same thing about digital converters?

The other problem is that the only thing that we're wondering about is our perception. It's all based on what we think we hear, not any measurable standards.


So, if the thing we're talking about doesn't fall into ANY of the measurable standards known to man then how can we even pretend to figure out the reasons for it? So far I've never heard anyone explain why these two things sound different other than the fact that each system has it's own unique analogue circuitry. There are a ton of people who design converters and software who have spent their entire lives studying all the details of this process and even they can't come up with a conclusive answer to what is different.

Personally, at this point, I don't believe the difference even exists. Do things sound slightly different when I record them digitally than they did before I started? Yes. Does the same thing happen with analogue? Yes. I think the real answer lies in designing the analogue circuitry around our converters to match the electronics in our favorite tape machines. Burl seems to be going in the right direction, and I'm very excited to hear his converters.

Whoo. Sorry for the rant. Just fell out of my hands I suppose.

Yes, this is precisely what I mean! And I believe this points to the crux of the argument when folks say that digital gives them back what they put in. They can make this claim and believe it to be true because they are putting in a source that has already been converted.

I have no qualms about taking a pre-converted track and looping it in and out of conversion another 4 or 5 times. To my ears I hear no difference. However, it really makes my face go crooked when I have something on the tape deck and I attempt to convert it to digital for the first time. The first conversion seems absolutely critical and therefore I believe it should be postponed as late in the production process as possible if one is working with analog tape.

Adam Miller--I'm going to check out that book you recommend. I would love to learn more about the details of this stuff. Who knows, maybe one day I'll design a product that solves the issue I hear...

Brad

That would be an interesting experiment for sure.

I agree with the infinite resolution theory. Unless overlapping or offset samples are recorded there will always be a gap in what you capture to digital. No matter how small the gap something is lost. Digital is ON/OFF where analog is always "ON".

Increasing digital resolution helps reduce the gaps, but it's never (because of design) going to MATCH analog perfectly.

Overlapping or offset samples.. It would take Two mics for a mono source.

imho

Why? Because for one. Analog vs. Digital is like Film vs. HD digital recording. Or computer art vs. Oil Paint, and on and on.

Another is different electronics and different method of capture technology. One toast the other grills. Don't make the toaster be a BBQ grill. Its an up hill battle. A useless battle. Take it for what it is and make music with it. Just a suggestion. AND....

Its a different medium. Cut and dry. Never will be the same IMHO. I quit trying to believe I can make it the same long ago. I use both where they are needed to make the best most artistic music I can. I use both for their "goods" not their bads. And they both have their bads as much as goods.

Resolution of 1/2" tape analog recording is approximately 4 times better than a CD is what John Dent mentions on the video link below.. But it doesn't mean you cannot make a nice recording or song. Its like saying again

Hey you can't make better art than Picasso by using only a pencil. NOT.

" Hey you can't make a Beat machine box sound like a real drum kit". You don't. You make music with it for what it is. I remember when sample drums first came out. People would put down and criticize the sound of a programmable drum machine. Especially, drummers and rock guys. Hip Hop guys and Pop 80 music fellas took it as a form of new sound and art and sold millions of records. Cymbals sound like paper, snares are too thin, sound robotic. Right! it does. But if you put it in the right place, twist the right knobs you might come up with an legendary Snare sound like PRINCE has done. Also Herbie Hancock "Rockit" made it work. And it never stopped thousands of Hip Hop artists, dance music artists, today's pop music and producers from making good selling hits! They looked at it as another medium to use to make music.

Digital will never sound exactly like Analog IMHO. 100%. NEVER. If it ever does by chance in the next 50 years, that will be the day when computers can make food and people. Its two different things. No matter if the resolution is at the same level. It will still sound different. Ones a Dog the other is a cat. They both make great pets.

Watch this video. John Dent has alot to say about this subject.

https://gearspace.com/board/video-va...john-dent.html

To sum it up. I look at it as this Brad. And it is only my opinion. As an analogy, You can't make computer graphic art work be exactly like, or smell, or feel, or look like a geniune oil painting. I appreciate both for what they are. And of course there is nothing stopping you from mixing them together to make something new and original. Same with music, digital , analog tape and vinyl.

Nice post BTWthumbsup And sorry if I could not be technical in nature. I am a creative type of person not scientific.

First, have you verified that at higher sample rates (eg. 96 kHz) you are getting the full 20 Hz -48 kHz range captured and reproduced? I've heard some cards just filter out >20 kHz regardless of rate.

Second, have you tried taking a complex mix (anything), sending it for a simple loopback, and analyzing the null? Audio DiffMaker is good and will sample align, clock drift correct, volume match, and extract the null for you, giving a correlation reading as well so you can see how deeply it matches.

It might be easiest to in this way start by first determining what is lost in a digital > analog > digital conversion.

Personally, I think if a converter A/D/A chain is to be truly transparent, it must be able to produce a loopback that nulls to at least -60 dB or more relative to the original (ideally >80).

I am not sure if any modern converter actually offers this level of performance with a complex signal. Would be interesting to know. Because if a converter can't produce a nullably transparent loopback, there's no reason to expect it would produce a transparent from analog capture either.

Just for shits and giggles I'd like to one day hear what a 384 sampling rate would sound like. Again, I'm not necessarily saying that the problem is actually assignable to sampling rate or that we don't already have sampling rates that are more than adequate. I'm just saying what if? If one were to be able to sample at the same resolution as tape, the number of atoms per square inch of tape, would the problem go away? I know that Nyquist is supposed to make this point moot but I'm just thinking out loud.

Well, the only alternative to PCM converters I know of is DSD. PCM requires digital decimation and reconstruction filters to operate in the manner it does. DSD avoids this altogether and records in a steady stream of 1's and 0's - as purely as digital can. I think DSD when reassembled in "PCM" comes out to a >200 sample rate (DSD 5.6448 MHz @ 1 bit/sample =~ PCM 235.2 kHz @ 24 bit).

I can imagine avoiding the filters this way removes some of the coloration, but the only cheap DSD recorders (Korg) use abysmal analog components so I don't see how that would help overall.

Also, to stay at DSD, you need special software that I think won't run any plugins as we know them so that's no use again ...

i'd love to see the computer that could record that sample rate in realtime..

Damon,

See what Once a roadie wrote at 4:57. That should clarify my hypothesis. When I said digital source I meant "analog source that has already been converted A to D".

To answer your question.. When I mix from the MCI 1" multitrack to the 1/4" Studer deck I get back what I put in (note: my Studer electronics are highly modified and it's aligned for high output ATR tape at 15 ips; take my word for it there is negligible sonic difference between input and repro on that deck, besides hiss...it's one of those things you'd have to hear in person). So the amount of depth is exactly as I hear it at the console outputs. When I take that same mix output of the console and feed it into A>D>A, the result is different when I compare the console main outs to the D/A output. I've experienced this with MOTU, RME, and Mytek conversion. I need to do some more listening tests with the Radar, but initial impressions are yes. And by the way, lower sample rates exhibit more "artifacts". I've tested 44.1, 48k, 88.2k, and 96k.

I think I understand what you are asking. I can say confidently that if I record a track simultaneously to tape and DAW at conservative levels, and then take the DAW version and bounce it to the tape deck so that the same gain staging is preserved, the two tracks when played back from the tape deck do not sound the same. The track that was converted sounds "altered". The sound of this is easy to hear when you have a group of tracks to compare, especially if there are stereo tracks with spatial cues. The next time you come by the studio we should do some of these tests so you can hear this stuff for yourself. It will make you scratch your head and say "what the heck is conversion doing to my audio and how do I get it to not do that??"

Brad