The best buffer we can try, and the help of passive filters to remove 10-20dB from the residual noise and distortion ?
We can imagine 3 fixed frequencies in the audio range (50, 1000, 20 000) could be enough for this very special feature ?
About this Mooly test, i tried-it blind with Foobar 2000. Believe-me or not, with my poor set-up (my hifi is 1500Km away) i had 1 at each first try of a serial of tests with a very short sample (may-be 10 serials of tests ?). And was worse and worse accurate with fatigue and boredom.
My best score is 9/10, my worse 4/10. average is around 6/10.
As i was never able to get 10/10, i consider the difference between the two tracks exists, but is of minor importance and hard to discriminate.
I'm sure i can have much more better results with better samples, adapted to discriminate transients (percussions), inter-modulation (many instruments together), little signals details (acoustic guitars, live recording).
So, my conclusion about this load issue (with this OPA sample) is: Load have a real influence, nothing to get nightmares.
My best score is 9/10, my worse 4/10. average is around 6/10.
As i was never able to get 10/10, i consider the difference between the two tracks exists, but is of minor importance and hard to discriminate.
I'm sure i can have much more better results with better samples, adapted to discriminate transients (percussions), inter-modulation (many instruments together), little signals details (acoustic guitars, live recording).
So, my conclusion about this load issue (with this OPA sample) is: Load have a real influence, nothing to get nightmares.
Last edited:
That part is clearly specified at 600 Ohm load drive, I find it of limited usefulness to do tests of devices used outside of their intended purpose.
What use would it be to judge a perfectly good 25W amplifier with a speaker that makes it clip at normal listening levels?
It would be so easy to put up a set of files of say two 20dB line amps with the ONLY difference that one uses Dale 1% resistors and the other uses naked Vishays. (I'll pay for the Vishays if you want). This is where the discussion wanders off to all too often. Same goes for coupling capacitors, but I'm not paying $400 for a bespoke teflon cap. 😉
BTW something like this was tried years ago as a pass it around black box that folks could use in any system they want, never got off the ground.
Last edited:
Scott, what do you want? Every audio designer to strictly follow guidelines? Guess what, many don't. I have seen real examples of the LM4562 loaded with 300 ohms or so, in today's audio products. 100 ohms just makes the problem more obvious.
However, when I bother with something like a listening test such as this, I use STAX electrostatic phones with a direct drive tube amp (by STAX) to separate the differences more easily. You have to have the BEST reproduction possible, if you are going to hear differences consistently and obviously. But this is just one IC with a marginal CD based source, and a DIFFERENCE is noted. What about a higher quality musical source as well?
However, when I bother with something like a listening test such as this, I use STAX electrostatic phones with a direct drive tube amp (by STAX) to separate the differences more easily. You have to have the BEST reproduction possible, if you are going to hear differences consistently and obviously. But this is just one IC with a marginal CD based source, and a DIFFERENCE is noted. What about a higher quality musical source as well?
They just want it to put 12V peak into 50ohm with -140dB distortions up to 200KHz.
Why? I have no idea. The latest AP 5xxx does 25V peak in 600ohm at -120dB
When someone asks me for a SOTA headphone amp, I'll tell them an x555 is the one to get. Much obliged.
Which problem is that? The fact that design guidelines are being ignored?
That's obvious.
What next? Complaining that output chips loaded way past SOA are blowing up?😀
Seriously, what about the supply feeding the chip. PSRR?
jn
And whose problem is that? My main point is don't get this mixed up with sound of precious components or wire directionality. In a way Mooly's test just muddies the water and gives a few folks an opportunity to say, "see, we hear these things".
I would rather suggest to wait holding his breath until somebody does "12V peak into 50ohm with -140dB distortions up to 200KHz" for free and then gets the approval of the golden ears.
I presume that the -140db into 50 ohms is for the oscillator buffer for a piece of quality test equipment. I had to design something similar once, but the requirement was only -100dB at 100KHz. I used all discrete fets with 2 pole feedback to extend the open loop bandwidth to about 100KHz.
We expect the most from our test equipment. It is frustrating to spend thousands of dollars on an audio analyzer and still have obvious (at least readable) harmonic distortion. It is still hard to separate the normal IC's from the exceptional IC's without taking this into account. For example, what has always mattered most to me is higher order odd harmonic distortion, as it is most dissident. I usually target 7th harmonic, but even my latest analyzer has some 7th and I have to add a passive filter to get it below measurement levels. A better buffer-oscillator combination would always be welcomed.
I doubt it. I have always thought that people's ability for blind test is WAY worse than sub conscious ability to perceive "difference" in long term, which is important for music listening enjoyment.
Here is example (some kind of proof to me):
1) People couldn't differentiate a tube amplifier from solid state (with R at the output) amplifier while in reality IMO they are way different.
2) Class-A amplifiers draw people into listening for long hours, which could be caused by less fatigue due to less crossover distortion. But this cannot be explained with numbers. If numbers are shown, then there will be question: "Is the difference audible???". Blind test will prove that it is inaudible. This left the first claim questionable.
I often mentioned that differentiating sound is not harder than deciding which one is preferable. That's why I have tried to practice listening to "fatigue" instead of the usual measurement because at the end you want enjoyment (which should be fatigue free).
I have no doubt that speakers are still the weakest link for most of us. If amplifier issue and speaker issue (such as distortion) is seen as additive in nature, of course we can ignore amplifier issue. The fact is, it is often more like multiplication, just like transistor amplifying stages. But what is important and what is not in an amplifier implementation could have been misjudged.
With the LM4562 loaded to 100 ohms --- dont care if it never gets used that way in applications... Not the point for listening test. I am curious as to what level we were all detecting. Was it 10 percent ? A tenth of a percent? because if it was more like 10 percent, we havent learned anything new. But if it was 0.1% or under 1%, then we can say we know that isnt the lowest yet and set about to detect even lower amounts.
-RNM
-RNM
Last edited:
yes. That is the reason for me/us to do it. [also the AP's -120dB is for thd +N.... I am referring to thd, only]
Waly gets an F for not paying attention.
Thx-Dick Marsh 🙂
If you mean that their equipment review is as lame as their blind test result, I think I agree. It is also hard to "read between the review lines" when money is the real motivation.
But I doubt that they can be considered as "experienced" listeners. They were not used to listening for slight differences.
I understand that if you use a transistor rated at 60V at 200V, the sound can be remarkably degraded. Narrow-minded engineers insist on that tiresome reading of datasheets, understanding component ratings, and specifying the proper components for a targeted requirement. That's why they'll never be successful in fashion audio, and have to be content with mundane things like selling a billion dollars of product into critical markets.
The test was only "can you identify R" between two audio samples. Yes, it's 50/50 you'll get it right. Knowing there's a difference, I have little doubt that "by golly, I'm gonna hear it" was a subconscious factor. So I intend to try blind identification of X & Y. Four runs of 25, then another 4. Use that data to proceed or walk away.
I did consider looking for Mooly's test, but I had already sent him my choice. All I did was listen (expressed and implied qualifiers apply).
I did consider looking for Mooly's test, but I had already sent him my choice. All I did was listen (expressed and implied qualifiers apply).
- Status
- Not open for further replies.