My advice would be that you try a product before buying instead of reading an opinion of someone who hasn’t heard it but yet has a very strong opinion. Especially if someone else who actually used it and heard it may have said the opposite. A 30 day return is possible with almost every product. After all you would want to buy it to listen to it, not measure, right? Otherwise you may end up with ASR recommended papery and clinical sounding ear piercing playback system. Best Regards M
5 Likes
You get me interested in your products, if you show measurements that justify the price you are asking. It is as simple as that. Take the free advice @Graeme_Finlayson offered to you. No need to retype what he wrote, as I fully agree with his position. @Amir_Majidimehr never attacked you personally here. He just measured one of your products and the results showed flaws in your design.
11 Likes
In 1985 Hifi Choice already wrote about the Pioneer A-80 that measured extremely well on distortion, noise etc: “In a sense, this amplifier continues to show how weak the correlation is between conventional measurement and subjective sound quality. Lacking in depth, lige and ambience, the A80 was a sonic disappointed and cannot be recommended.”
Already in 1985! Have some people fallen asleep the last 40 years?
1 Like
I fail to understand what you want to say. Is it ok in your world that a bad engineered PS
produces distortions into the audio stream? Please explain.
3 Likes
I’m continuing to wrap my head around the fact that accuracy, preference and perception correlations are considered hard problems and that Harman landed on a preference curve that is not flat.
That curve results when measuring speakers from the listening position in room (optimized room acoustics) that are measuring flat with uniform off axis radiation under anechoic conditions on axis.
It’s originally been formulated as an optimal headphone target response.
2 Likes
The Harman curve does work in non treated room conditions with a lot of reflections primarliy.
The most important thing is a smooth decay time over the whole frequency spectrum. If this is granted a linear curve should be preferred in my opinion.
That is a good question. So the preference curve is a severe alteration of flat/linear. No surprise that just some extra 2nd harmonic could make a presentation sound more realistic.
There is a lot of discussion about bad engineered PS. Linear or switched or both… If it sounds good but measures not so good, one could try and find out why that is so to get a better idea of what matters and what does not, that is my opinion.
I think there is some misunderstanding here. As far as I know, the Harman curve is about the frequency response of headphones. As @Marin_Weigel said, headphones that are tuned to the Harman curve approximate the sound of linear speakers in an anechoic room, which reinforces the fact that people prefer a linear response. It has nothing to do with harmonic distortion.
5 Likes
I do have much more than anecdotal or personal experience on the field of loudspeaker quality evaluation and how to optimize the listening room. And there has been a lot of research done. But this has not necessarily been published. For example, every serious broadcasting corporation has its own method of evaluating monitors before they order huge quantities to equip all their control rooms. They assemble their most experienced recording engineers and have a listening panel with almost scientific methods.
If you have ever witnessed such a process, you would drop most of ideas about loudspeaker and room quality being fully measurable or determinable. These people are pretty serious, have a no-frills-approach, they know what there is to know about measurements and room acoustics. And they nevertheless talk a lot about transparency, detail resolution, bass impulses, ambience, localization, depth-of-field and audibility of reverb characteristics. All these things which can not be determined by measurements, not even closely.
You could walk into such a panel of experts and tell them a Sonos One for 250 bucks is fully meeting their demands and measures perfect and what they are doing is self-delusion and wastage of money. At best they would shake their heads. They know better.
Starting with a system without dramatic sonic flaws and good measurements is a good idea and I would always do the same as a starting point. I absolutely believe that it is possible to come to a good level of reproduction without major flaws, so I have no reason to doubt any research coming to the same conclusion.
But the more you try to understand what a ´neutral accurate system´ should ideally be based solely on very primitive and simplified measurements, you will come to the understanding that such a system cannot be real as you have to take the room acoustics and the interaction between room and loudspeaker into account.
To come even close to a sonically perfect reproduction from technical point of view, you have to have:
- perfectly flat RT60 of the room in the whole audible band
- perfectly diffuse reflexions in the room
- perfectly even directivity index and behavior of the speaker between 300 and 10,000Hz incl. all angles, narrow frequency bands (which prohibits any use of smoothening) and in the time domain
- which also results in having a perfectly small, perfectly behaving point source in the aforementioned frequency range
If you invest $200k in the room and $50k in the speakers, you might come pretty close to that and you can certainly EQ the rest of imperfections away. But I am sure that you still would be hearing some different characteristics of different speakers when it comes to the aforementioned parameters like transparency, bass character, ambience, depth-of-field/perceived stage distance and localization (and perceived dynamics, most probably). All these things you cannot influence by EQing and that’s why even serious recording engineers of big broadcasting cooperations go to the control room and have a shootout listening session.
Please note: I am talking about loudspeakers and rooms solely, neither DACs nor amplifiers.
That does not surprise me at all. On one hand, people are used to tilted frequency curves as most of rooms have decreasing RT60 towards higher frequencies, most speakers have (more or less evenly) increasing directivity index towards higher frequencies and everything from deep bass to low midrange is sounding ´fatter´ due to resonances and other in-room phenomenons.
On the other hand, a declining overall level towards higher frequencies in the room is sounding less annoying and smoother so a majority of people who are not used to the music and actual listening conditions are preferring this.
But it does not mean a neutral, accurate or anyhow superior reproduction quality. If you take measurements in the room and the idea of ´high fidelity´ seriously, a declining frequency response towards treble is a major flaw. A million times more dramatic flaw than anything that has been presented as ´evidence´ in this thread against certain DACs.
Nevertheless, people who think measurements should be necessarily perfect for perfect reproduction quality, take this lightly or do not ask themselves ´why?´. We have rock-solid evidence that a perfectly flat frequency response in the room is clearly inferior in most of preference tests.
So something is wrong with this equation in general. I have made the same experience with curves flattened by DSP or automatic room correction routines (DIRAC most of all). If the room is not perfect, almost certainly one thing is true: the flatter the curve, the more audible the flaws and the less balanced the sound.
Sorry to say, but headphone target curves are even less applicable here as headphones circumvent the head related transfer function of every listener so a flat curve is far from sounding balanced or neutral.
As mentioned, I have a pretty good idea about why a declining level is preferred by many listeners. The answer would not please anyone of the ´measurements-first´ people.
Reflections dominating the sound at the listener, even if it is limited to a narrow frequency band above 300Hz, is by itself an unexucasble flaw of the listening condition making it unsuitable for any controlled preference test. But it is also a hint that there are reflections and other phenomena which are annoying to listeners.
I have made listening tests under conditions which are much much closer to the aforementioned perfect ones, including even directivity index of the speakers, passively optimized room, optimized listening distance and opting for some subjectively transparent and ´silky´ speakers. Cannot confirm the Harman curve at all.
2 Likes
Seems, you’re barking up the wrong tree, since that’s not what I wrote - maybe reread my post to understand that we’re on the same page regarding at least this point.
It’s not a major flaw - It’s a perfectly valid choice and offered by various room correction softwares. REW offers a “house curve” option. A perfectly flat speaker will have a downward tilt in a real room due to absorption by soft furnishings. It’s the acoustic environment we’re acclimatised to. It should come as no surprise that it is the preferred choice.
This is very different to deliberately adding distortion to a DAC"s output. I went to great pains with ny latest speaker build to get the speaker positioning right and correct the room first (within the bounds of acceptable WAF), followed by equalisation in REW with additional correction in Rephase. I also applied a slight downward tilt as a “house curve”. The final result measures well and is very pleasing to my ears.
I still use one of the most revealing and transparent front-ends available (Benchmark DAC3HGC/HPA4/AHB2s). These terms aren’t defined by price or by hearing, but by measurement. Minimal noise and distortion means a device’s output is uncoloured and is as true as possible to the source material.
If you want to colour it to taste downstream, knock yourself out. But at least start with accurate raw material.
FWIW, I have 4 headphone amps, 2 based on ECL83s with transformer outputs, 2 are solid state. Can’t really say I prefer any one over any other. Headphones are equalised in Roon with convolution filters to closely approximate the Harman curve and tappings on the tube amps are chosen carefully to minimise any output impedance frequency dependance. With the solid state amps, output impedance is negligible.
Tubes don’t necessarily equate to “warmth”, “euphonics” or any other hyperbole popularised by the hi-fi press. Well designed tube amps can perform very, very well indeed.
7 Likes
@Arindal maybe we are talking about different things. The average person listening with Roon at home is very different from a large broadcaster putting together a control room!
The broadcaster wont be using Roon and the home enthusiast wont have a perfect accoutic space.
Still, lots of ways to get good sound at home.
Well, definitely not honest, the thread
… accurately describes the capricious manipulation that happens there with 5 examples, changing goal posts being the worse. It can be as much as 30dB+! depending on how Amir was feeling on a given day. A superficial and sloppy two sentence “review” where Amir gets his hand on what’s seems to be an early rendition of product, with early firmware, measures it, doesn’t call us with any QQ, doesn’t listen and trashes it again after doing the same with 90% of hifi products that costs over $1000. Then 50 minions rush to put their nails to the coffin using highly offensive language apparently allowed by Amir, the moderator. Same seems to happen with every other major brand product in this index when sorted by price:
You seem to represent both a fan of ASR and an unwilling victim of ASR misinformation and the way review is conducted, written and your resulting perception. This is what we are not happy about! About your skewed perception because of Amirs sloppy work. The product is definitely misrepresented - there is no clear word that this is a Streamer with a Roon Core computer built in, yet it gets compared to a $900 Topping DAC and then in the last post here to a $129 headphone amp! Why not a refrigerator? There is no mention about settings, the 2nd harmonic seems to be turned on, all measurements are completely skewed, no word on analog vs digital volume, all these things affect measurements. It’s a deeply unfair misrepresentation of a great product that every other review and customers value very highly, particularly the sound and sleek Roon user experience. If you considered it’s a Roon Core, just an I5 Nucleus alone with 4TB of storage costs $3000+. None of this is mentioned and the worse part is the lack of communication with and now hostility to the designers. It all results with your takeaway of " horrific power supply artifacts" etc etc. Now look at the FFT of the current shipping unit. There is a little of residue of 60 and 120 below -125dB! What about it is “horrific”?
Explain to me why is this bad to have some noise at -125dB when getting a beautiful rich musical playback vs -130dB with dry , clinically sounding Topping is the client preference? If you like max out Sinad, get the Topping , but don’t defame Mytek when there is nothing wrong with it!. We design for beautiful sound, and superior user experience (no word about it either in the review) and we don’t and will not design for max Sinad because this is not what we do.
Amir’s “reviews” are not real “reviews”. It’s the most sloppy unprofessional approach you can imagine to presenting any product from any review site I have seen. All products , even the ones that are recommended are capriciously misrepresented. In most cases if this was the only info you could read about a product , you wouldn’t be even able to tell what the product is intended to do. All this combined with hostility towards designers instead of typical courtesy and cooperation.
As for the 2nd harmonic. You could have turn it on or off in the menu, so if you don’t want it, turn it off and you get an FFT as above and a better Sinad.
With the mantra of “ASR is always right” you appear to think that ASR mission is indeed to teach everybody a lesson. Like someone said earlier “Presumptuous beyond belief”
Benchmark designs for max measurements and we respect that and are friendly w John. We don’t design for max Sinad and we won’t. We design to get Mytek Sound because this works for our clients. I have been designing electronics (ADCs and DACs) for the last 40 years. What did you design? How many circuits did Amir built? There is nothing in the superficial understanding of audio at ASR that’s valuable for us, I typically get inspired with fun idea things from other designers and products that many were trashed at ASR. ASR is useless as a source of deep hardware design clues (unlike for example diyaudio.com) and I’m really annoyed with its culture so this no place for me. ASR just happened to drag me involuntarily to their corner and here we are sorting this event out.
When I have more time I’ll try to better explain Mytek design process in more detail as I feel this may require a more convincing explanation to measurement only proponents audio hobbists. It’s far more nuances than what you guys think is possible.
BTW: Like I mentioned once, when we have this heated, but IMO still civilized exchange, a TROLL working for ASR in this thread, keeps flagging my posts and they got hidden/removed so for example Amir posted answers to my post that’s disappeared. Dear Troll, why don’t you stop this and just let ASR play a fair game?
Thnx M
7 Likes
How did you come up with this? If you are not happy with the frequency response of your speakers, adding disortion wont fix that??
The beauty of Roon is its DSP function that allows you to address flaws from speakers/room.
If you want to add some distortion, 2nd harmonic, there are much better ways than a fixed level in a dac. Do it in software, you can precisely control the level and turn it off if you want.
To buy flawed speakers and fix it with a flawed dac is… illogical…
If you prefer alternative frequency curves and added dustortion, no problem. You can tune it anyway you want - we have all the tools now. If your serious, and not just arguing for the sake of it, I can walk you through how do it.
4 Likes
Well, what is horrific is that you don’t even know how to read your own measurements!
The vertical axis is “dBrA” mean it is relative to something. You needed to set that to the peak level of the signal as I do. But you did not. Instead, the 1 KHz signal is at around -15 dB, not 0. That means we need to add 15 dB to amplitude of power supply spike, arriving at its level being 125-15=110 dB below reference signal.
Easy: your claim of sound fidelity is “nonsense” :). You are repeating typical audiophile myths that an audio device that measures superbly sounds “clinical.” This is an absurd claim that not backed by a single reliable controlled test. It is a convenient way to sell underperforming audio hardware by companies and owners defending purchases of the same.
Your own measurement shows a spray of harmonics going on to near 20 kHz, not just a single 2nd harmonic distortion anyway. Those later harmonics are not subject of perceptual masking and hence, if you want to make a claim of audibility, those would be it. As per your own reference earlier in the thread in that interview, higher order harmonics are not preferred.
You have no evidence of such “sound.” You keep claiming it without back up. Psychoacoustics research shows there is no way, no how what you showed as far as second harmonic is audible. You can still claim it with a controlled listening test but you lack it. This is unscientific and simply wrong.
Nope. I gave you credit for having Roon core in the review. I then compared you to Eversolo DMP-A6:
It is $850 and is a full blow streamer and can have local storage. Granted, it is not Roon core but is gorgeous with its massive display that can be customized to show things such as VU meters:
This is how it performed:
Notice total absence of power supply noise. This is again your streamer performance:
There is no comparison. You have that spray of harmonics mixed with spray of noise from the power supply making for messy amount of impairments. So much so that their combined energy sinks the difference between signal and those elements to just 98 dB. CD/16 bit audio has a rectangular noise floor of -97 dB so you are barely clearing that in a $5,000 device no less!
Someone can buy the Eversolo and put $500 toward a quiet NUC running Roon core and have a far more performant system for far less money than your solution.
I post other products to counter your claims that other designers hate ASR testing, not to compare yours to theirs. You have however produce DACs so in that context, they make sense as well.
We buy your products, not you coming alone with it to design DACs for us. So it doesn’t matter how long you have been doing this. That aside, it is interesting that you couch that as DAC & ADC. Clearly you haven’t been putting a computer inside a DAC for 40 years. Folks like Eversolo haven’t either but they have been building media servers for a long time. They understood, on their own, that they also need to build exemplary DACs even inside a noise platform that has a computer (they run Android). They got there. You have not. They charge a lot less. You charge a lot more.
While the above is shameful status of how badly the high-end audio market gone off the rails, there are very expensive products that are recommended as show in the above graph. Specifically, the Mola Mola DAC which retails for $11,500 and got my strong recommendation. Why? Because it produced superlative performance, the best I had seen at the time.
Look at how clean the FFT spectrum looks. Bruno Putzeys is a long time designer that cares about producing the most transparent product he can.
The high-end audio market needs to wake up and start measuring products and fixing issues they find. If they don’t, and product lands on my bench, then they get what they get. This is how TotalDAC tested for example in the above list:
The red graph is a $250 device. Green is the TotalDAC. Are you going to claim this the type of performance a DAC needs to have? That it has its own “sound?” Surely they have happy customers like you have. Are they up to something you are not? See the problem?
14 Likes
I am with you. A flat curve is preferable and the Harman target curve with its falling level towards the heights is only a (bad) choice for people who are not willing to treat their room for the best possible sound. Because with neutral speakers AND the strong reflections above the bass they get ear bleeding.
1 Like
Unfortunately this is the game of roulette many audiophiles play with high-priced gear purchased on the basis of sighted listening impressions, subjective reviews and the recommendations of other purchasers on internet fora.
One bad component after another is tried in order to fix something that’s “wrong” with the sound.
The benefit of having measurements allows poorly designed products to be avoided in the first place.
11 Likes
From acoustical and technical point of view and to anyone who has ever listened to the alternative, it keeps being a major flaw. Admittingly a popular and pretty wide-spread one, and there is an explanation for that as it usually sounds less annoying that alternative flaws (such as increasing direct sound level to higher frequencies and alike).
Nevertheless I wonder why people who insist on acoustically neutral and technically close-to-perfect reproduction take this as granted and do not try the neutral variant which they propagate extensively.
Having a vast experience with rooms and speakers, I cannot confirm that. Most of current rooms do not have enough of treble-absorbing areas to really make a difference in RT60 (if they were equipped with curtains, carpets and sofas in the 1970s, this might be a different thing).
A technically neutral speaker with even directivity index plus a room with balanced RT60 will have a more or less flat overall measurement as a result (maybe with the exception of everything >8K, but that is not really relevant).
I seriously doubt that. I go to live performances without reinforcement regularly, from chamber music to big opera houses, and I have never experienced such ´dull reverb´ and subjectively damped overtones in the treble region (even in concert halls paneled in wood sitting close to the last row).
The real reason why so many ´flat-on-axis´ speakers result in a tilted room response (with all its drawbacks reproduction-wise) is the tendency to have constantly and increasing directivity index. Why is that so? My guess would be that many manufacturers are not daring to implement measures for an even directivity index in the midrange on one hand, while an increasing index still sounds less annoying and ´warmer´ than an uneven one.
I do not doubt that it is pleasing your ears, as according to my experience this is a setting which is usually sounding smooth, non-annoying and sometimes even adding more depth-of-field. But would you rule out the possibility that a significantly better reproduction would be possible if the indirect sound field would be balanced as well?
According to my experience, the ´tilted-by-directivity-index´ variant in many cases lacks dynamics and resolution in the treble region, having a tendency to sound dull, lame and distant.
I do not see the difference between adding non-linear distortion to a DAC´s output and adding linear distortion to the indirect sound field. The main difference seems to be that the latter is clearly audible and changing the way we percept music reproduction dramatically, as the former seems to be more subtle given the level of the added components.
I have never personally compared it in this particular case, but it seems to me that the level of distortion added is regarded not to be audible as distortion as the level is way too low. I wonder how it is possible that seemingly the same people insisting that one’s ears cannot distinguish between the different settings are on the other hand taking it as a hint that the components are anyhow flawed or inferior.
But we can put that to the test:
Could you explain please why a constant tone of 150Hz, -115dB below reference signal, is audible under everyday listening conditions? The highest peak of harmonic distortion seems to be k2 peaking at -98dB or 0,0015% compared to reference signal level. Is there a reliable controlled test meeting your own standards existing showing that this level of THD is clearly audible and perceived as sonically inferior?
If we agree that it is not audible, what is the point in publishing the measurement, ranking devices based on such data and labelling them as ´underperforming´ just based on something that is not audible in your definition?
1 Like
What is “everyday listening conditions?” Why should you buy products that make you have such a condition for audibility/inaudibility when you can get absolutely silent response?
While this kind of noise may not be likely audible, it indicates poor design/layout which can translate to noises becoming audible. Evidence of this exists in the kind of reviews Michal likes as I have posted before:
“Downstairs, with my desktop system, I noticed some low-level noise and hash, the kind that can sometimes leak through a computer soundcard, and also some hum. The hash was not audible from the balanced or headphone outputs—only the unbalanced. At normal listening levels, with no music playing, the hash was audible but low in level.”
In other words, the measured noise is just the proverbial canary in the goldmine. If you allow power supply leakage, then other types of leakage can and does exist per above. Indeed, we have measurements to back this:
Notice massive amount of interference. Such interference is usually a function of what the device is doing and can get a lot worse depending on conditions. A good design has isolation so noise is not going to come out regardless. Here is the Eversolo again in the same test:
Clear audible? No. It is a bit like this: you are not going to get sick if the waiter spits in your soup before bringing it to your table. But do you want him to do that?
I commented on one specific thing, i.e. 2nd harmonic. Noise is another matter altogether and that is combined in SINAD to drag the Mytek way down the rankings. Noise is absolutely audible because it can exist in absence of signal so no masking is involved. You can see that in the reviewer case above where he heard it while nothing was playing.
If I didn’t measure, you wouldn’t know what product has a problem, and what doesn’t. My measurements allow you the analysis. Without it, you would be blind.
Note that people are running with marketing stories thinking they are getting a higher fidelity product than those costing 1/5th the cost. To the extent the situation is inverted and demonstrated using measurements, then folks may want to rethink why they would want to pay such a premium.
Let me repeat again that getting these noise sources down costs nothing. It simply requires skilled engineering and measurements. The Mytek could have been as good as Eversolo in which case, it too would have gotten my strong recommendation, assuming the filter bug was fixed.
Put your ear next to your tweeter. Do you hear a noise? If so, the case is made that noise can be audible. This audibility is much worse for headphone listeners so we need to do the best job possible here.
The power of measurements is that it has enabled people to save a ton of money, buying highly performant products that are transparent under all conditions, while costing so little. This is why I measure and reward companies that bring such products to us.
Of course, there is percentage of the market, myself included in that, who would pay a premium for products that look great, come with great support, etc. I am happy to even play thousands of dollars for them. But they need to not underperform budget products. It is irrational to get fancier case but worse electronics.
11 Likes