Well said, Mark! We both agree on this completely!
And I am looking forward to see anything but beliefs, that means some evidence supporting the subjective claims.
Maybe I should have taken more time to word that a little differently. What I was trying to get at is that Jacob2 seems to see the activity here as though it were some professional scientists trying to run an experiment for subsequent publication.
But, I think most of us know this is more of a fun social event. People with similar interests get to share opinions, swap stories, etc. Mixed in with it are some more serious attempts to do good science, but not in the highly rigorous way a research team at a university would do, and need to do.
Personally, I would like to see it be a little more in between different pictures I tried to express in the foregoing paragraphs, a little more towards what it might be if everybody were more formally trained. But, that's only my preference, and it only counts for one vote. Most people would leave if they had to do things the way I might prefer, or the way any single person might prefer. That's not what this is.
It's for fun.
Bill, Regarding pot and kettle, for preexisting beliefs actually I am paraphrasing Jonathan Haidt and Phil Tetlock. It's the opinion of two world class psychologists trying to summarize what has been learned from studies of human nature.
For better or worse, we are all human, and the reality is not always as noble as we might prefer to think. What can you do? We are stuck with what we are. Probably better to accept it and try to work with the hand we are dealt, rather than to go into denial.
For better or worse, we are all human, and the reality is not always as noble as we might prefer to think. What can you do? We are stuck with what we are. Probably better to accept it and try to work with the hand we are dealt, rather than to go into denial.
Last edited:
Well, the reason I was not surprised by the result is, it is fully expected. I would have been very surprised if uu were LM4562. I know it would't happen in my experience.
We performed the similar tests at the commercial studios a few times, and we already know that the fanciest spec opamp is not the best choice. 5534 is always one of the top choice, and TL072 is not as bad as we think.
These tests are simple and anyone can do it at home, so why don't you guys try it yourself to clean up your skepticism.
This test is enough scientific, and result is very reasonable.
We performed the similar tests at the commercial studios a few times, and we already know that the fanciest spec opamp is not the best choice. 5534 is always one of the top choice, and TL072 is not as bad as we think.
These tests are simple and anyone can do it at home, so why don't you guys try it yourself to clean up your skepticism.
This test is enough scientific, and result is very reasonable.
Actually I see a lot of products currently in production that do use the 1458 in the input stage along with a pair of transistors!
As to statistics, my favorite example of confusion is the "Let's Make a Deal" problem.
The contestant is given a choice of three doors, behind only one of which is a prize. After they choose a door, the host opens one of the unchosen ones to reveal that the prize is not there. The contestant is then asked if they want to change their choice. The correct answer is that they should always change! To understand why, one must actually understand the critical issue.
Same thing with surveying op-amps. If one participant can correctly correlate the amplifiers with their specifications and do this on multiple runs with the labels changed between runs, then it doesn't matter what anyone else finds. The only issue becomes what percentage of the time can they do this.
As to the issue of source material, there are folks who swear by records. Obviously inferior by measurement, but to verify that the same test could be repeated with a vinyl source material. (For those with way to much time on their hands.)
Now are the results likely to change, of course not. It would just be very interesting if it did.
As to my opinion of engineering, if I find no difference between op-amps at a gain of 1, but do find a difference at a gain of ten, just to be sure I would use the better op-amp. As to cost differences, by the time you calculate all the parts, PC card, case, labor etc. there usually is no significant difference.
Now I am not yet finished with my version of op-amp testing to allow the critics a feast.
I will not in any way criticize your version, you had quite specific goals and seem to have met them.
BTY my results are almost always perfekt after I know the answers...
Ed, I agree that there are good reasons to use better parts rather than the poor ones, I myself use only AD797, OPA627/637, OPA134/2134 and LM4562 for audio. And, the 4562 only with great care, as it is very sensitive to air-coupled EMI. I do not use neither1458, nor TL072. But, as you have already guessed, I have had certain goal which I demonstrated in this test. You know, experiment is over 1000 words, especially taking into account the endless and always repeated debates.
Yes, the 1458 would not be usable for gain 20dB. Too high noise then, increased distortion, reduced BW. But, the gain 20dB was not the case in this test. The test is valid just for the conditions used in the test.
Yes, the 1458 would not be usable for gain 20dB. Too high noise then, increased distortion, reduced BW. But, the gain 20dB was not the case in this test. The test is valid just for the conditions used in the test.
Last edited:
Actually I see a lot of products currently in production that do use the 1458 in the input stage along with a pair of transistors!
As to statistics, my favorite example of confusion is the "Let's Make a Deal" problem.
The contestant is given a choice of three doors, behind only one of which is a prize. After they choose a door, the host opens one of the unchosen ones to reveal that the prize is not there. The contestant is then asked if they want to change their choice. The correct answer is that they should always change! To understand why, one must actually understand the critical issue.
Same thing with surveying op-amps. If one participant can correctly correlate the amplifiers with their specifications and do this on multiple runs with the labels changed between runs, then it doesn't matter what anyone else finds. The only issue becomes what percentage of the time can they do this.
As to the issue of source material, there are folks who swear by records. Obviously inferior by measurement, but to verify that the same test could be repeated with a vinyl source material. (For those with way to much time on their hands.)
Now are the results likely to change, of course not. It would just be very interesting if it did.
As to my opinion of engineering, if I find no difference between op-amps at a gain of 1, but do find a difference at a gain of ten, just to be sure I would use the better op-amp. As to cost differences, by the time you calculate all the parts, PC card, case, labor etc. there usually is no significant difference.
Now I am not yet finished with my version of op-amp testing to allow the critics a feast.
I will not in any way criticize your version, you had quite specific goals and seem to have met them.
BTY my results are almost always perfekt after I know the answers...
Surrounding circuit and parts have more impact than opamp itself, and as far as I can see, professional designers seem to put money to the other things for better result at the same cost.
Pavel, From what you are saying now it sounds like you were trying to prove that you did not have to exclude using a particular part for a noninverting buffer application?
If so, you might have saved yourself some of the endless debate if you had not publicly stated you conclusions here, when you did not publicly state your true goal. An opmap part that might not be so good in a high quality professional application might be fine for a lower cost consumer product, for example.
Being sneaky, and using free volunteer labor to double check what you thought would probably be okay in the first place may have resulted in you saying things here that might have better been kept to yourself, at least for the purpose of reducing the amount of debate you would have to deal with personally.
Or, in a moment of honesty you could have just told us what you were trying to do and asked what you wanted to know more directly, if not in the beginning then at least sometime before now.
But, if you felt we had to be kept in the dark, in psychological research when it is necessary to mislead test subjects it is considered an ethical requirement to disclose the true intent of the study immediately following a test subject's participation.
If so, you might have saved yourself some of the endless debate if you had not publicly stated you conclusions here, when you did not publicly state your true goal. An opmap part that might not be so good in a high quality professional application might be fine for a lower cost consumer product, for example.
Being sneaky, and using free volunteer labor to double check what you thought would probably be okay in the first place may have resulted in you saying things here that might have better been kept to yourself, at least for the purpose of reducing the amount of debate you would have to deal with personally.
Or, in a moment of honesty you could have just told us what you were trying to do and asked what you wanted to know more directly, if not in the beginning then at least sometime before now.
But, if you felt we had to be kept in the dark, in psychological research when it is necessary to mislead test subjects it is considered an ethical requirement to disclose the true intent of the study immediately following a test subject's participation.
We must understand that the listening test is always biased somehow. Choosing String Qualtet as a material would draw completely different result than choosing vocal Jazz or electronic music. Analog / digital source, ADA converter, power supply should impact the result as well.
One can easily setup the test that proves his favorite A sounds the best, and we always see that kind of test on Gearslutz. Many are performed by the manufacture. Compare to those tests, this is much less biased than the similar tests, at least, PMA tried to do so, I believe.
One can easily setup the test that proves his favorite A sounds the best, and we always see that kind of test on Gearslutz. Many are performed by the manufacture. Compare to those tests, this is much less biased than the similar tests, at least, PMA tried to do so, I believe.
Thank you, I really did not have any intentions to make one of the parts a "favorite" and to make it more favorable than other parts. The main reason why I prepared the test is that I was unable to find sound differences between the parts in my AB box, though I was self-convinced, years ago, that opamps do make a difference even in the unity gain buffer circuit, when I swapped them in the sockets. I think I have already explained this. So I prepared the first test in 48/16, but was asked for the files in higher resolution. This resulted in this second test, in 96/24, with source file ripped from highly regarded Steely Dan DVD-A, Everything Must Go. At least, it really has quite enough energy above 20kHz and it is not only 44/16 resampled to hires. The source of the DVD-A data was a master tape. More on
Steely Dan 'Everything Must Go' | Mixonline
Not in a well designed-test like the one PMA did. The purpose of a well-designed controlled test is to make sure that the only clues are the sound, and as such delete the biases. Not easy, but can be done.
This could be the case if we assume that different types of music exercises different mechanisms in the equipment. I don't think you can say that as a generality.
Of course your test music must be dynamic, and wideband to make sure the equipment is properly exercised. But this can be done with any type of music and then the results do not depend on that type.
And sure, equipment must be first rate but that is a given.
Jan
Moreover, it has real music content above 20kHz! How to prove this?
1) use cool edit or a similar SW, load the music file and use scientific high pass filter option, steep enough (you have a choice) and cut the signal below 20kHz.
2) the result after filtering is of course inaudible. Now, change the sampling rate down, about 4x, so you get 24kS/s instead of 96kS/s, so the file is now 4x slower. Now, you will hear mostly transients of percussion, hi hats and other fast transients. This is a proof that the file contains useful info above 20kHz.
You can do the same procedure with another "hires" files and you may often find that there is in fact nothing above 20kHz, except for some studio digital artifacts (interferences).
Jan, sorry, but I can't disagree more. There is no best equipment for all the situation. Even lab test result is only valid for the same situation.
Why professional engineers have so many different mic preamps? Because they want to choose one of them BASED ON THE MUSIC.
Guys, really, it's Mark flapping his arms trying to distance himself from whatever this comparison was, because it is incongruous with his expectations (and or is upset with this disparity). Ask yourself a simple question: were the files prepared in a manner that they were equally able to acquit themselves on their own accord?
I'd argue yes. Nothing looks like deliberate or accidental tampering.
Was such a comparison of limited utility? Sure. And, at the same time, one really needs to be careful not to draw too many conclusions, but unlike Mark's post hoc complaint that this wasn't done by scientists (whatever that means, and ironic given he himself manipulated the files), I don't see anyone going off the deep end upon learning the poll results. No one was expecting to scale this up and try to publish the results, so you're all fine.
It's guaranteed that no test will result in universal satisfaction.
I'd argue yes. Nothing looks like deliberate or accidental tampering.
Was such a comparison of limited utility? Sure. And, at the same time, one really needs to be careful not to draw too many conclusions, but unlike Mark's post hoc complaint that this wasn't done by scientists (whatever that means, and ironic given he himself manipulated the files), I don't see anyone going off the deep end upon learning the poll results. No one was expecting to scale this up and try to publish the results, so you're all fine.
It's guaranteed that no test will result in universal satisfaction.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.