Ethan Winer

Couldn't it be that cheap equipment sounds better to me? Anything is possible, but the real answer is it doesn't really make much difference what converter (or mic pre) you use, as long as it's not audibly inferior such as the preamps in a 40 year old Radio Shack PA mixer etc. --Ethan

I'd really recommend you buy a cheap converter and spend the difference on $50. bottles of Cabernet! I'll drink to that! --Ethan

So what is the answer? Are we still waiting for more responses? I was hoping nobody would actually ask because then I could use the same files again the next time this comes up. Oh well. g1.wav is the Apogee g2.wav is the SB X-Fi --Ethan

If you don't mind uploading I'd be interested in hearing them. Tell you what, email me from my site and I'll send those clips to you that way with an explanation. Then, in a few more days if nobody else posts their guesses, I'll reveal which of the guitar clips are from which converter. --Ethan

I suggest you redo the tracks with either white noise or pink noise ... run it at 0, -3 and -10db Ah, it was you who suggested that. Apologies to Dave. All that's needed is a full scale signal because it already includes lower-level signals. The hardest thing for an A/D to capture is a 20 KHz sine wave at full scale, or a loud transient that has a similar spectrum. Indeed, a single impulse contains everything that is needed to analyze any device. Multi tracking is a different issue though. I don't see why. If a sound card captures the source with a response within 1 dB from 20 Hz to 20 KHz, with noise at least 96 dB below peak level (ie: the top 16 bits are clean), and distortion less than, say, 0.01 percent, then it is by definition audibly transparent. Whether you mix 16 microphones in a console and feed one channel of an A/D, or send each mic to its own A/D and mix them in a console later, the results should be identical. If they're not, then something is connected wrong or there's another error. Likewise, when people claim to hear differences in fullness, imaging, stereo width, ambience tails, and all the rest, they are either mistaken or the test was flawed. For example, the two signals were not level-matched, or acoustic comb filtering skewed the results while listening back through speakers. It doesnt go full circle though and show them how to identify the actual shortcomings of the cards I can think of no better test than recording the exact same signal into multiple converters and listening to the results. (Other than a real test using proper test gear.) If you have a better idea I'd love to hear it. As long as it's not along the lines of "years of experience will teach you to recognize the failings" because that doesn't explain anything, and specifically dodges any science-minded attempt to identify what's really going on. I know your test was to invoke some discussion Yes, that's one point, and I'm happy just to get people to start thinking about this stuff rather than blindly believe everything they read in magazines and on gear sites. But the main point is consumerism, and getting people to understand where their money is best spent (microphones and loudspeakers) and where it's not (preamps and converters). Those who purchase the high end stuff, unless they are complete idiots Many are complete idiots! Seriously, every day I see some forum newbie list all the high-end gear he just bought, and ask the group how to connect it all. --Ethan

Ok, I think the forums are being screwey today. One of my posts apparently never made it and I'm seeing a post from Ethan that I didn't see earlier. Here too. I could have sworn you said something about needing to record instruments at different levels to better exercise a converter, and when I went back just now to address that I can't find it. Or did you see the error and remove that part? I think a triangle is a much better instrument to use when we're comparing converters. I'd love to hear it. I see now that I can't do that. It's a long story, but if I post the triangle it's possible you'd be able to figure out which is which for reasons having nothing to do with the sound quality. I can't even explain why this is so, or you might be able to identify the earlier guitar file too. I'm not ducking anything! If you think it would help I could make a new recording of a triangle comparing my Delta 66 and SoundBlaster. Not sure what that would prove to you though. I can tell you that the two triangle files sound remarkably similar, just as the two acoustic guitar clips do. Room mics are the best way to go IMO. Let me hear how the converter picks up the sound bouncing around that room. Why do you think that matters? Firepod is FW and Aroura is PCI, but I've never tried installing both drivers at the same time. That should be fine. I have drivers for my Delta and SB, and I also installed drivers for a Presonus FireBOX which I needed for a project last year and never bothered to remove in case I need it again. I'm thinking it would probably be better for someone else to do this since my place picks up a lot of noise from the outside world. Now hold on there a minute pardner! You're obsessing over converter artifacts that are surely 80+ dB down, while recording in a room that picks up outside traffic? So you say my needs are "limited" compared to a full-time studio for hire, but your own situation masks all subtle detail with traffic noise and rumble unless it's a screaming amp stack or drum set? The cymbals (even their decay) survived the myspace rape process better than anything I've done so far. I'll gladly have a listen because I like your stuff. But hearing a mix made through a particular converter doesn't say much. Especially after it's converted to a relatively low bit rate lossy file. Tell you what, here's a much better test we can collaborate on: Email me from my site www.ethanwiner.com and send me a couple of short clips, maybe 30 seconds each, of your two tunes at 44/16. This is infinitely better than anything one could download from MySpace. Then I'll record both clips through my Delta 66 and my SoundBlaster, and post all six files. It's not quite as telling as recording a triangle at the cusp of overload, but it's still quite useful. If an entire complex mix can survive unscathed, that pretty well proves transparency. In fact, I could also take those clips over to my friend's house and run it through some of his high-end converters. I'm pretty sure he has at least a few other boutique brands beside the Apogee, and they're newer too. He's an hour away so it won't be tomorrow, but I could record your tune fragments through my two converters tomorrow. I also have the FireBOX I mentioned, and could do that too. It's not a Firepod, but it's the same family. You're also welcome to send me a short clip of a single room microphone track, since for some reason you think that is more taxing on a converter. --Ethan

Are you saying M-Audio may have redesigned it? It wasn't horrible-sounding, but its throughput was absolutely, positively inferior to other, more expensive boxes in the shootout. I have no idea if M-Audio redesigned my Delta 66 since I bought it. I assume that's unlikely. I've never even tested mine critically because it sounds fine. The closest I've come to testing my Delta 66 is recording an entire mix off a great sounding CD, then I played it back to see how it sounded. It sounded the same so I moved on to more important things. Is there a link to the shootout you mentioned where the Delta 66 sounded "positively inferior" to more expensive converters? If so, please post it because I'd love to hear the files and read how the test was performed. But if the shootout compares different performances, don't bother because the result are useless. But you also don't want to run a splitter cable unbalanced from a Mackie 1202 into a $6000 Apogee and then convert the results to MP3. Who said anything about MP3 files? And why do you think running a splitter cable unbalanced has even a tiny affect on audio quality? Certain guys (including myself) can create better and faster mixes with analog mix busses, because we're hearing what we need to hear all day, not just during the final bounce. I don't know what software you use, but when you click Play in SONAR you get exactly the same results as rendering a final mix to a Wave file. --Ethan

Admittedly, this was eight, nine years ago. Right, and a lot has changed in prosumer grade converters since then. Also, doing a proper blind test is not trivial. Levels must be matched to less than 1 dB, and preferably to 0.1 dB. If that is not done then all bets are off. Again, I urge others here to record and post clips. Dave, if you have a Firepod and an Aroura, you should be able to do this easily recording any instrument you think best shows off the difference. A $2 Y cable from Radio Shack is sufficient for unbalanced, which is fine. Or maybe you have a normalled patch-bay that can split one balanced signal into two paths. Or solder one yourself in 10 minutes. Recording the same signal into two converters at once is not difficult. Or maybe you can record one into your main rig and the other onto a laptop? Also, it's not difficult to include the D/A portion in a test. After you record through both converters at once, run a separate pair of passes where each converter's output is routed back to its input and record again. When you listen to that file you're hearing two A/D passes plus one D/A, plus the second D/A as you listen. The second file you post here will have only the first three, with the last conversion through the listener's D/A. So that captures both A/D and D/A in the second recorded file. --Ethan

What I'd love to see are correctly-performed experiments that show the opposite. Agreed. Hey, don't shoot the messenger! If my test could be improved on, then let's see Dave's better test using whatever cheap and expensive converters he wants, on whatever sound source he thinks will best show the flaws in the cheap gear. As I mentioned before, we also recorded a triangle and claves through both converters, and I could post those clips too. Though then Dave would probably find something else to criticize. It seems to me an electric bass is about as easy an instrument to capture, but if Dave believes otherwise, let's see his test using a bass. And not just Dave, but everyone here is welcome to post clips for the rest of us to try to identify. More examples can only increase the base of knowledge for all of us. An informed consumer is a smart consumer. --Ethan

I feel like this really is the weakest example you could give to argue this point ... I certainly can't speculate which converter Okay, you agree the SoundBlaster sounds very much like the Apogee, so now you attack the sound source as flawed? A close-mic'd acoustic guitar has lots of fine detail, many frequencies sounding at once, and plenty of transients. As I explained above, the source is mostly irrelevant as long as it has sufficient frequency and dynamic content. Maybe a trashy sounding guitar is exactly what the producer wants! So the converter needs to capture that trashy sound accurately. It seems that your main argument is the almighty dollar. And what's wrong with that? Most people have limited funds and need to get the most value from their money. Telling someone they must spend $500 per channel or whatever for "pro" results when $50 per channel will get an equally transparent conversion makes no sense to me. for the kind of music you record, it certainly wouldn't make sense to drop a lot of cash on converters. I can afford to buy anything I want. Whether I do this as a hobby or for hire is irrelevant. And I do record others for hire, mostly small classical ensembles. But even for my own stuff I want to capture my acoustic and electric guitars and cello with as high a quality as possible. I also record a lot of hand percussion which is very demanding. If you want to spend more than necessary I have no objection! I'm more interested in helping people who do not have unlimited budgets and want to know The Truth ™ about what they must realistically spend. We all see posts every day from newbies who are unhappy with the quality of their work, and wrongly believe the problem is their prosumer gear. The real problem is their lack of experience, and no amount of money invested in boutique gear will fix that. Yet they are told again and again that they must buy expensive converters. That's just wrong. --Ethan

Also, what was the microphone? What did the guitar sound like in the room? The microphone was my DPA 4090. The guitar in the room sounded like a guitar. Not trying to be cute, but there's no other way to answer a question like that. I could take a $5 mic into a Radio Shack DIY preamp to record a pawn shop guitar, run this comparison and decree that there's no difference between a Sound Blaster and an Apogee. The problem with that is that the converters aren't the limiting factor - there's so much fidelity lost before the signal gets to the converters that their differences go largely unnoticed. That might seem reasonable on the surface, but the logic is flawed. Whatever the source was as captured by the microphone, the job of the A/D is to convert it to a digital equivalent as accurately as possible. The source could have been an electric guitar cranked to 11 through a Marshall stack with an SM57 2 inches in front. Or it could be a $4 Million Stradivarius violin captured with a $10,000 vintage Neumann tube microphone placed optimally above the stage at Carnegie Hall. Either way, the "fidelity" of the source is defined simply as whatever the source "is." Good guitar or bad guitar, good snare drum or bad - it doesn't matter. The source is mostly irrelevant as long as it has sufficient frequency and dynamic content to tax the converter. --Ethan

The tonal quailites of a mellow acoustic guitar are a poor sources for an test as well. Niether frequency extreme is being challanged. Agreed, and we also recorded a triangle and claves for that reason. I could post those clips too if anyone cares. I suggest you redo the tracks with either white noise or pink noise ... run it at 0, -3 and -10db ... your approach to testing the frequency responce of different cards reveals nothing and has no basis in supporting any of your arguments. This was not meant as a comprehensive test! It's just a reality check to counter the conventional wisdom that $25 sound cards suck by definition, and you'll never get pro results until you invest in expensive converters. As you can hear from these short clips, that simply is not true. Whether some people like it or not. --Ethan

what is the piece being played in the samples. It's pretty... I like it. Would like to hear more of it. That's an original tune my friend Grekim Jennings was working on at the time. I haven't seen him in a few months, and for all I know he may even have a finished version on his web site for download. Grekim is a very nice guy and I'm sure he would not mind an email from you: http://www.acousticrefuge.com/ --Ethan

Do you have a source for that one? I'd like to run the calculations myself before I accept that. This is common knowledge. Jitter is measured in picoseconds, so it's easy to do the math. 1,000 ps = 1 nanosecond = 1 GHz. I have trouble believing your mp3 story. At 64 if someone can't tell the difference someone doesn't know what to listen for. 128 is harder, and beyond 128 it depends entirely on the source material Yes, 128 was the lowest bit rate I tested my friend with. And since it was solo piano, even at "only" 128 kbps it was tough for me to tell too. I can generally spot MP3 artifacts at 128 kbps with normal pop music having drums and fuzz guitars etc. Regardless, the point of my story is not at what bit-rate MP3 artifacts become audible. Rather, it's that people often have very strong opinions they are certain are correct, but in fact they are not correct. essentially what your saying is that every interface should sound nearly identical if they are all transparent. I don't think that's the case and I think the majority of the audio world would disagree with you. I don't care what the majority of people think. As the famous quote goes, "If 50 million people say a foolish thing, it's still a foolish thing." And so it is with many things in audio. I'm not saying every sound card sounds the same. I'm saying that every competent sound card that's not broken sounds the same. Aside from measuring, which is the very best way to assess A/D/A quality, second best is to record multiple conversion passes until they accumulate enough to be clearly audible. You can hear the degradation from a SoundBlaster after only a few passes, where a very good converter might still be transparent after ten generations or even more. --Ethan

So, you mean to tell me that Metric Halo, Apogee, Lavry, Prism, Lynx, et al. have been selling "snake oil" this whole time? Hell no! A lot goes into truly professional gear having nothing to do with the sound quality. Plus high-end gear does sound excellent. Besides raw fidelity, which is not difficult to achieve these days, there's also build quality, features, stability, reliability, quality of the drivers, and more. A real studio like, say, Universal records 80-piece orchestras that cost thousands of dollars per hour. They're not going to dick around which cheap crap that might break at the worst time. If a Behringer DI costs $30 and a Radial or whatever costs $200, they'll gladly pay the $200 every time without blinking. --Ethan

What was the preamp? What was used to mult the signal post-pre? The preamps in a Mackie 1202 were split using a normal Y cable. Was the DAW SONAR? What bit depth and sample rate were used? 24 bits, 44.1 KHz. What was the computer interface for the Apogee? The only $6000 Apogee I'm aware of is the 12-year-old AD-8000, which I own personally, and it doesn't have a computer interface, unless you buy an AM-BUS Firewire card. Darn, I can't recall how it was connected. It might have been Firewire because I don't think I had to open up my computer. The apogee is owned by a friend who uses it daily and swears by it. If your DAW was using the SoundBlaster as its audio interface, how was it seeing the Apogee at the same time? Surely you weren't running the Apogee digital out into the SoundBlaster's S/PDIF in. SONAR allows recording from multiple sound cards at the same time, even when they have their own separate clocks. What was generating the clock? Each card used its own clock. However, we also tried clocking the Apogee from my M-Audio, but neither of us heard any difference. I'm not sure if the SoundBlaster has balanced line-level inputs, but the Apogee certainly does. Did you compensate for this? We simply fed both inputs unbalanced. I hate these kinds of shootouts, because there are too many nebulous factors. Unfortunately, the only way to give it serious credence is to actually be there. At least it's not like most shootouts you see that are totally useless because they compare different performances. As in "I played my guitar through an xyz microphone, then played again using the abc microphone." Really, this is about as bullet-proof as it gets because the same signal was presented to both converters. Many, many inexpensive audio interfaces sound like ass (including a few in M-Audio's line), and it's not all in our heads. I can't disagree with that, and I have compared relatively few sound cards. It's not always in people's imagination, but it often is. Time after time I've had people swear that such and such is crap, only to find they can't tell after all when they can't see me switch. Last year someone I met at a hi-fi forum came out from Brooklyn to visit for a day. He's a hi-fi guy, not a studio geek, and he insisted he can always spot lossy MP3 compression, even at very high bit rates. So I took a track from a solo piano CD he brought, and made MP3s at three different rates: 128, 196, and 256, plus the original. The first time I tested him he got them all right! The second time he got them all wrong, totally backwards. He identified the 128 kbps MP3 as the original Wave file, and the original Wave as the 128 version. Before we did this test my friend was adamant that he could easily spot lossy compression. He was even a little po'd at me when I challenged him to prove he could hear it. --Ethan

But this is just one single track. You don't notice a lot of difference. But what if you're tracking a whole band? The whole notion of stacking is flawed, and is just another myth. We had a great discussion about Stacking Theory at Gearslutz a while ago. Yes, as expected there was a lot of insults and name calling, but I think all agreed in the end that stacking errors do not accumulate. I bet there's a greater difference when you consider clock stability Clock stability is mostly a myth too. Crystal oscillators are accurate to infinitesimal fractions of a second, and the errors correlate with frequencies in the GHz range. My Artifact Audibility Report lets you assess the audibility of artifacts such as jitter. I also have a PDF describing a study that concluded jitter needs to much worse than what you get from typical consumer grade gear before it's audible. I'll be glad to email that PDF to anyone who wants it. Just email me from my personal web site www.ethanwiner.com. Here's part of the conclusion: "The results indicate that the threshold for random jitter on program materials is several hundreds ns for well-trained listeners under their preferable listening conditions. The threshold values seem to be sufficiently larger than the jitter actually observed in various consumer products." --Ethan

In your A/B test you are only presenting the cards ability to record. Once its recorded you're posting it and relying on someone elses card to complete the full circle. This is only 50% of what a card must do for you albiet an important part. Excellent point. My intent here is not to create a scientific test, but rather just to get people to understand their own biases. A $25 SB card is about as ghetto as one can get, and my friend's Apogee is a good representation of high-end boutique gear. Also, A/D conversion is infinitely more complex and prone to distortion and response errors and ringing etc than D/A. A competent D/A can be made from a resistor ladder and one op-amp buffer with no gain to get a usable output impedance. The conversion back to analog may be enhanced, Flat... This is an excellent point. I have a theory that some expensive gear intentionally gooses the very high frequencies a dB or two to add some sheen and trick people into thinking the proverbial "veil" was lifted etc. Especially with hi-fi type products. As far as the sound blasters go, I used to use one for mastering my stuff years ago and I'm very familure with how well they work. I also have Yamaha cards, crystal, C-Media, Ess, M-Audio, Hercules and realtek. All of them sound different and have their pluses and minuses. Yes, though SB cards from the past few years are infinitely better than the old SB Live cards and other such models from 5+ years ago. When Creative came out with their Audigy eight years ago it seemed clearer sounding and was much quieter than the SB Live from that era. I assume they now use the same chip sets for SB cards too. At least it sounds that way. Then again, maybe I'm being tricked by a goosed high end. I am serious. I never measured either card so it could be. --Ethan

When I did tests to determine the flattest listening position in my studio, inches made a difference. No kidding, and the graph below from my Audiophile Beliefs article shows this in no uncertain terms. The two responses were measured only four inches apart in a 16 by 11 by 8 foot untreated room, yet they're so totally different you'd never guess this is even the same room and loudspeakers! --Ethan

Clearer highs, richer lows. Some more detail in the non-musical things, scraping on strings, strings rattling on frets is in the first one, too. I'm not ready to say which clip is which yet. I'll wait for Dave and WRGKMC to chime in. But here's a suggestion for you and everyone else too: After you listen for these sort of details, listen again an hour later to both clips. Sometimes we hear new details just because we missed them on the first playing, not because of a legitimate difference. Human auditory perception is very tricky and very fleeting! A few weeks ago, a guy in one of the hi-fi forums I visit was convinced that demagnetizing his LP records (!) brought out subtle details such as you describe. In his post he described more air from a flute part, and hearing details not present before he applied the demagnetizer. Clearly you can't affect an LP record with magnets, so all that's left is faulty perception. And of course placebo effect. --Ethan

Some of us can actually hear the differences between live amd memorex quite distinctly will continue to strive for perfection. Of course, and I strive for perfection too. I agree that where real differences exist, it's worth paying more for the quality. Sometimes you have to pay much more to get that last few percent. These days I own mostly modest gear, but I have a few pieces I consider high end because anything less was not good enough. Remember, the original question asked "At what point does the expense outweigh the returns?" My opinion is at around $100 per channel. And not just outweigh further small benefits either. $100 can buy total transparency. Not "almost as good" or "good enough for a project studio," but every bit as good as converters at any price. At least for one A/D/A generation. I won't say a $25 SoundBlaster card is as good as the best, but it's not the garbage some would have you believe. I'll be interested in hearing your pick for which clip is which too. I consider that Placebo thing you talk about to apply more to pot heads and amatures who really dont know what they're hearing or what they;re doing to begin with and does not apply to pros who have spent a lifetime doing this. Unfortunately that's just not true. Double blind tests are the gold standard in science for a good reason! Everyone is subject to bias, including seasoned pros. Even me and you. --Ethan

i must say i'm baffled by the similarity. Indeed. They don't sound exactly the same, but they are very close to my ears. Certainly close enough to be negligible compared to the quality of the room, which microphone you use, and of course where in the room the microphone is placed. Keep 'em coming guys. --Ethan

Ethans talking about double-blind tests, eliminating perceptions and psychoacoustics, variable control and unbaised interpretation of what you hear when changing only one thing. Exactly. Everyone understands and accepts that the placebo effect is real, but for some reason audiophiles don't think it ever happens to them. For all I know the Firepod really is a total POS. I already said I don't have one, but I have heard and used plenty of mid-level sound cards that are transparent and indistinguishable from high-end converters. Here's a blind challenge for everyone who is certain that all prosumer grade converters are crap. Below are two short files for you to identify. One was recorded through a $25 SoundBlaster X-Fi card, and the other through a $6,000 Apogee. This is one performance captured through one microphone that was split after the preamp, then sent to both converters and recorded at the same time. Anyone here care to identify them? Clip 1 Clip 2 After enough people have posted their guesses I'll reveal which is which. --Ethan

Preamps used were the Firepod's. Like I said, the converted signal was louder. To some this might make the converted signal sound better. It was clearly very bad sounding though by comparison. It was louder in a very "oh god please turn that down" kind of way. Then perhaps being louder was the culprit, driving something else into slight distortion. Again, the only way to do a fair comparison is level-matched and with someone else doing the switching. Either that or your Presonus really is defective or not up to what the specs claim. My last speakers were M-Audio BX8a's. They advertised a very flat response. On paper, they look like the perfect speaker. Speakers are a whole 'nother ballgame! As Joshua said, A/D/A conversion is a "pretty basic and simple process" that is easy to manufacture to the standards of total transparency. Speakers are never ever what the specs claims, for many reasons I'll be glad to elaborate on if you care. is it because when people skip the expensive interface, they'll have more money to spend on acoustic treatment? Dude, I'm trying to help you here. No need to be combative or disingenuous! If your FirePod really is not up to snuff, you'd be wise to at least consider the possibility that better and truly transparent converters are available for a reasonable price. The only bad thing I've heard about Presonus is apparently some of their Firewire drivers are buggy. Before reading your account today, I don't ever recall hearing someone claim the sound quality is flawed. It boggles my mind how you can fly in the face of logic, evidence and black and white personal accounts ... I guess you have a motive. That's why it's called anecdotal evidence. Again, I'm trying to help you here Dan. There's no need to cop a 'tude. If the signal changes from analog to digital, and it isn't the very low input impedance, and it isn't that it got louder and started clipping, it's in your best interest to try to track down the problem. And problem is the right word, because properly functioning prosumer grade A/D/A converters are definitely capable of full transparency. No matter what vendors of very expensive converters might tell you. --Ethan

If the converters were (for argument's sake) rolling off everything below 100 Hz, then it doesn't matter how many times he ran the signal through the converters, he wouldn't hear anything below 100 Hz, because the converters themselves were affecting the sound. I understand your point, and logically it might seem that's the case except for two important things: * Frequency response changes are cumulative, so if Dan's Presonus rolled off a little at 100 Hz for one pass, after five passes it would be rolled off five times more. * Specs don't lie, but people are often mistaken. Unless Presonus is not being truthful - which I suppose is possible! - errors such as you describe would be easily seen in the specs. Looking at the tech specs tab on THIS page, the FirePod claims to be very flat, with very low distortion and noise. So I have to assume something else is going on with Dan's setup or the way he compared. Two things come to mind: * I see on the page above that the FirePod has an input impedance of only 1,600 Ohms, which is awfully low. So depending on what device Dan has feeding the FirePod, it's very likely this is the source of the loss of fullness. Especially if the previous device is "prosumer" type gear. * Another factor for loss of fullness (or clarity at high freqencies) is level matching. When doing a proper A/B comparison, whether blind or not, it is absolutely critical that both A and B be level matched. A difference as small as 1 dB can make the louder version sound fuller and clearer. The standard for proper testing is to match levels to within 0.1 dB. --Ethan