THD as a metric correlates poorly with human perception. The order of partials matters significantly. The old analog distortion tended to be low order. The more objectionable digital harmonic distortion is higher order.
In other words, yes, a few percent 2nd order HD might not be noticed, but
.01% higher HD order might be audible to virtually everyone.
The bom cost won't be high for Benchmark, though higher than for someone building in high volumes. We build stuff from very high volumes to very low - but because of that the low volume stuff gets the same pricing as the high vol.
The point is that some, probably a lot, of their price is recovering R&D - which is a decision based on many factors; and what the feel the price point needs to be. They will also, naturally, be trying to maximise profit.
Sound quality is not necessarily related in any predictable way to selling price.
Ah, yes. I wish any of it were transparent. Not sure we are there yet. Wouldn't be surprised if in a year or two somebody comes out with something better than we have today and it turns out some people can tell a difference blind.
You are right that THD has been shown to be a poor metric. But while your points re:digital may be true in principle, I remain skeptical. There have been performed quite a lot of blind tests of modern digital equipment where people could perceive exactly zero difference. This does not show that there is no difference, of course. People may be affected by the testing protocol, or they may not have received sufficient training to detect the differences. The only blind testers I know of who have consistently been able to detect differences between digital equipment is the Swedish Audio Society, who routinely blind test hifi electronics using a highly discriminating A/B-test. And even they have been unable to detect some fairly cheap gear, like the Yamaha WXC-50.
But what I think all this blind testing with negative results does show, at the minimum, is that any digital distortion (or whatever one may call) is far from being immediately objectionable to most people. It's a difference that is at best subtle, and it's therefore unlikely to account for any widespread perception of contemporary music as "distorted" by average music lovers. But there are other differences in audio and audio production which we can easily measure, and which also are detected in most blind tests: Compression, increased loudness, frequency response (much of today's music is too bass heavy and therefore perceived as lacking clarity), artificial stereo imaging (what happened to the good old recording techniques of live music?), etc.
All of these things are "big things" that are easily audible. Much more likely that it's these things that account for the perceived differences, IMO. And by using DSP, integrated active speakers and stuff like that, some of these big things can see very substantial improvements - like flatter frequency response, better bass in the room, etc.
Then there are lots of "small things" that may be audible to certain people, but that nevertheless often fail to show up in blind tests. What I find strange in audio, is that a lot of people are obsessing about the small things, and forget about the big things. Here are the things which matter most for perceived audio quality, as I see it, in descending order of importance:
1) Good recordings
2) Good speakers
3) Good rooms
4) Adequate amplifiers (in the case of passive speakers)
5) Adequate dacs
6) Format or source
But if you look at how much noise these topics generate, one would think that it's exactly the opposite! What do people talk the most about it audio? I would say that it's more or less like this, in descending order of generated noise:
1) Formats and sources (mqa, hi-rez, digital formats etc)
2) Dacs
3) Amplifiers
4) Speakers
5) Rooms
6) Recordings
It's almost the exact reverse.
This post is becoming a bit messy... What do I want to say? Perhaps that it's important not to lose sight of the big things in audio. That applies both to music production and sound reproduction in the home.
Personally I would be suprised. But I hope to be wrong. The big research at the moment is full soundfield processing on headphones for the gaming and movie markets. Studio created music is left in the dust and classical types seem in no rush to go surround sound.
I really had hoped that the Iphone effect would mean that immersive soundfields would get more research for music. Instead we now have phones with MQA baked in. Sigh.
I really had hoped that the Iphone effect would mean that immersive soundfields would get more research for music. Instead we now have phones with MQA baked in. Sigh.
I am perplexed in the opinions expressed here. I suspect that one of the big problems is the dependence on double-blind listening tests that just don't separate out the subtle differences between audio components. It appears to me that there has been a propaganda effort to discredit human listening from being the final judge as to audio quality. Yes, I know about subjective bias through knowing what you are listening to, but it just does not hold up realistically. You hear what is being played through the audio playback, no matter what the equipment price and finish, and that is what is the bottom line, not if you can't tell the difference between 'Live and Memorex', in most double-blind tests.
People have written books expressing that your subjective judgement can lie to you, but I don't believe it in the long run. In the same way, there were, (and probably still are) books written that spout Freudian Psychology for about 100 years now, giving people a strong sense that just about everybody was nuts! Where is the psychiatrist's couch today?
People have written books expressing that your subjective judgement can lie to you, but I don't believe it in the long run. In the same way, there were, (and probably still are) books written that spout Freudian Psychology for about 100 years now, giving people a strong sense that just about everybody was nuts! Where is the psychiatrist's couch today?
Last edited:
And yet you yourself state that subtle differences are impossible to discern when blind testing. Now, I'm sympathetic to the stress of testing argument, but all this another mumbo isn't going to fly for me.
That said, our experience of listening in our own homes isn't blind and should be chosen in a manner that makes each respective person happy. But it's also not going to translate from person to person in the same way identifying key characteristics that we all prefer (and or can actually hear rather than fit our ideology) will. The latter is where the benefits of blind testing become important.
It's the stating of preferences, especially ones that have no ultimate technical merit, as some sort of truth that gets me. "I like" is way way way different to "this is better". Stick to measurements and more robust evidence for the latter, with all the qualifications that it may not have an audible effect.
That said, our experience of listening in our own homes isn't blind and should be chosen in a manner that makes each respective person happy. But it's also not going to translate from person to person in the same way identifying key characteristics that we all prefer (and or can actually hear rather than fit our ideology) will. The latter is where the benefits of blind testing become important.
It's the stating of preferences, especially ones that have no ultimate technical merit, as some sort of truth that gets me. "I like" is way way way different to "this is better". Stick to measurements and more robust evidence for the latter, with all the qualifications that it may not have an audible effect.
Last edited:
Interesting to reference Freud given that his tests were less than scrupulous and most of his theories have not passed proper scrutiny. Like sighted listening...
Have you evidence that is not the case? My experience is most people are, it could be just that I've been unlucky
I think that it comes down to people fragmenting their decisions depending on the sonic faults. Everyone has various things they can deal with or excuse, and the perfect system, should it ever be created, would appeal to almost everyone except those who desire a specific colouration. But if you eliminate them, as the system gets better you'll find that more and more people are in agreement.
-Chris
-Chris
Spot on. It (propaganda) helps to sell a mediocre equipment, which also happen to dominate the market. So, in whose interest that propaganda is? Follow the money.. 😉
By that argument the movement to discredit double blind testing is to help to sell expensive equipment?
My view on listening tests: They are overrated in general. For most things in audio, I see no need for them. Just design for linearity in all aspects as much as you can - frequency, time/phase, dynamics, distortion - and be happy with the result. Listening tests have lots of confounding variables, and most of them don't get properly replicated anyway.
I'm also skeptical to listening tests for preference, the whole Toole paradigm. I can see an argument that preferences might be different short term and long term. My taste buds go crazy when offered Ben & Jerry's ice cream, but I certainly don't want to eat ice cream all the time.
The main goal of audio should be simple: Attempt to reach high fidelity. It ain't more complicated than that. Strive for the signal path of virtue.
So when is there a point in formal blind listening tests? Well, an obvious case if somebody tries to sell me something very expensive, and there doesn't seem to be an obvious scientific or rational reason reason why that device should be distinguishable from much cheaper devices. In that case, only a positive result in a formal listening test will convince me.
The other time a listening test may be called for, is if there's a trade-off between two aspects of linearity, particularly with speaker design. Say that we can either make a speaker very dynamic (big woofers etc), or we can make it approximate a point source. What to do? Which design goal is most important? In this case, listening tests can perhaps be useful. But I think sighted listening will have its use for determining this as well.
I'm also skeptical to listening tests for preference, the whole Toole paradigm. I can see an argument that preferences might be different short term and long term. My taste buds go crazy when offered Ben & Jerry's ice cream, but I certainly don't want to eat ice cream all the time.
The main goal of audio should be simple: Attempt to reach high fidelity. It ain't more complicated than that. Strive for the signal path of virtue.
So when is there a point in formal blind listening tests? Well, an obvious case if somebody tries to sell me something very expensive, and there doesn't seem to be an obvious scientific or rational reason reason why that device should be distinguishable from much cheaper devices. In that case, only a positive result in a formal listening test will convince me.
The other time a listening test may be called for, is if there's a trade-off between two aspects of linearity, particularly with speaker design. Say that we can either make a speaker very dynamic (big woofers etc), or we can make it approximate a point source. What to do? Which design goal is most important? In this case, listening tests can perhaps be useful. But I think sighted listening will have its use for determining this as well.
It's suggested that speakers positioned away from the front wall give a better sense of depth because we can see the space. I see no reason why this is not feasible
Not that I disagree with designing for linearity as a design goal but we have to accept that it's a philosophy of design, no? Some folks have different goals in mind, which is why there's plenty of room for the massive horns an 1W SET folks as well. We're here for enjoyment, in whatever form that takes.
On the n=1 thing for personal preference, I agree that playing with effects (including dsp x-overs) is one place where blind testing works well.
John, recently I gave some files that had been sample rate converted or not to some untrained listeners to evaluate. Sample rate conversion may have some possibly audible artifacts but it can also result in some peak level change. All of the untrained listeners responded in ways indicating they did not know what to listen for and they were fooled by the level differences. They reported nonexistent but to them very noticeable frequency response variations of midrange frequencies compared to other frequencies. Fletcher Munson at work.
Although the level differences originally went completely unnoticed by me (I was listening for something else other than frequency response), after I got the listener feedback and tried to check the files for what they reported then I heard it too. Of course, I was very puzzled and knew something wasn't right. It took me a few minutes to realize what it was and double checked by listening to one file at a time each at different volumes. Yep, that was it.
So, yes, people can be fooled and very easily so.
Or just opposite..It (propaganda against blind testing) helps to sell a mediocre but very expensive equipment. You can find a lot of most expensive "highend" equipment with very poor technical properties.. "So, in whose interest that (anti)propaganda is? Follow the money.. "
- Status
- Not open for further replies.