Hey !, I'm old too and may have cooked hearing...decades of SDRNR 😉 , but I note the differences
My hearing actually isn't too bad, but I do have listening experience and discrimination on my side....and a million or so hours of listening to a million or so songs on a million or so systems....and many thousands of items of audio gear over my benches.
IOW, my HF hearing may not be as good as it once was, but my discrimination and recognition is a million times better than when I was young, and increasing day by day.....my ten thousand hours is up several times over.
Regarding the cymbal build up/crescendo etc, a major difference is in the enveloping, a subject I mentioned recently, and the brush is strongly different between recordings also.
I don't understand that some report not hearing differences....are their systems lousy, do they have cloth ears or what ?.
We all agree that the differences are relatively fine.....and the importance/relevance of that difference depends on one's mode of listening and ones criterion, which can be strongly knowledge/experience dependent.
I agree a discussion of what to 'look out for' in each of the recordings would be very useful, and ought to help those who are not hearing differences.
I have struck these confounders in the past..ie the DAW waveform display is not properly representative of the actual audio tracks because of sample point differences.
The clock rate drift is not significant for subjective evaluation, but it does cause useless data problem for nulling comparisons.
The confounder for ABX is that A,B,A,B etc sounds different to AA,BB,AA,BB etc and it is not ones hearing that changes.
So far, some of us old guys are maintaining that we hear differences between the files, and yes we should explain to the 'young turks' what to listen out for.
Dan.
Hi Scott.
What kind of information do they reckon should be at the beginnings of those chirps ?.
Any HQ files available of what they have ?.
I'm finding some seemingly weird stuff like systems having dynamic memory of some sort, LF/VLF/ULF based I suspect.
Dan.
Max,
I think before we can get into the finer details of why certain opamps sound different in different application, rolling 8-legs, we should be looking at why the software for testing is doing what it is supposedly doing. By that I mean if you are seeing differences between A,B,A.B and A.A,B,B wtf is going on with the software to cause these types of anomalies? Before we can move forward to me at least those questions need to be addressed and a real solution that doesn't introduce those types of con-founders must be found. Until then all test results are suspect, if a simple order change is changing the measured results in software that makes any results of those types of tests results null to me at least. If you can't get duplicate results with just changes of order I wouldn't accept the test results and if that is the case with measurable, and I mean through electronic means how can we even start to do audible testing with that software?
Working in a medical lab situation I know I couldn't accept the test results for patients samples if I got different results just by changing the order of patient samples. And I will add in the very beginning of automated testing with the very first machines that used certain mechanical methods of passing the samples this did indeed happen, a high value from a previous patients sample would affect the next patients results, it was a real problem and multiple tests to confirm a result were necessary. Been there and done that.
ps. The first automated test machines used diastolic pumps and separation of samples by the introduction of bubbles between samples and the samples passed through semi-permeable membranes. This caused real problems with one result affecting the nest sample. i know I changed those membranes enough times and the tubing in the pumps to understand how that worked, it wasn't long before those machines changed and other methods were introduced.
I think before we can get into the finer details of why certain opamps sound different in different application, rolling 8-legs, we should be looking at why the software for testing is doing what it is supposedly doing. By that I mean if you are seeing differences between A,B,A.B and A.A,B,B wtf is going on with the software to cause these types of anomalies? Before we can move forward to me at least those questions need to be addressed and a real solution that doesn't introduce those types of con-founders must be found. Until then all test results are suspect, if a simple order change is changing the measured results in software that makes any results of those types of tests results null to me at least. If you can't get duplicate results with just changes of order I wouldn't accept the test results and if that is the case with measurable, and I mean through electronic means how can we even start to do audible testing with that software?
Working in a medical lab situation I know I couldn't accept the test results for patients samples if I got different results just by changing the order of patient samples. And I will add in the very beginning of automated testing with the very first machines that used certain mechanical methods of passing the samples this did indeed happen, a high value from a previous patients sample would affect the next patients results, it was a real problem and multiple tests to confirm a result were necessary. Been there and done that.
ps. The first automated test machines used diastolic pumps and separation of samples by the introduction of bubbles between samples and the samples passed through semi-permeable membranes. This caused real problems with one result affecting the nest sample. i know I changed those membranes enough times and the tubing in the pumps to understand how that worked, it wasn't long before those machines changed and other methods were introduced.
Last edited:
The ADA4841 part is still not too shabby noise-wise, especially on that low power budget.
But other than your AD797, the ADA4898 does seem the uber-fi part of choice in the AD lineup. 😀
To my own end, I had the ADA4897 in my head in terms of SiGe , which is also a 1 nV part. 5 V rails though, which is too restrictive without going to other measures for an opamp RIAA, but is fine for what I was thinking of using it for: DAC I/V.
(See, you guys have pretty low noise SiGe, too 😀)
I need to retry a number of Music Player softwares to compare...so far I find that Foobar is suspect, certainly I find that Foobar VST Wrapper and ABX modules are sub optimal.
Dan.
Last edited:
We had all TL072, all NE5532 and all 4558.
The five OP275 and five LM4562 were used in combination with two NE5532 in the bass section (the two opamps at the far right of the circuit).
The five OP275 and five LM4562 were used in combination with two NE5532 in the bass section (the two opamps at the far right of the circuit).
The LIGO site has full public access to some of the data sets/algorithms, just search around the site https://www.ligo.caltech.edu/WA/page/scientists. I personally don't understand what the theoretical physicists are saying but since the waves are much stronger as things spin faster they loose early data at low frequencies. There is also a desire to measure the background which I think they are far from.
Yes Andrew. As I read, the range of concern for them is from 1mHz to 100Hz
What do you mean by ‘composite’ amplifiers?
It doesn’t seem easy. Being able to help is great
https://arxiv.org/pdf/1602.03838v1.pdf
https://www.osapublishing.org/Direc...0-10-10617.pdf?da=1&id=232860&seq=0&mobile=no
http://dspace.mit.edu/openaccess-disseminate/1721.1/88409
https://arxiv.org/ftp/arxiv/papers/1502/1502.06300.pdf
George
Attachments
A second opamp (discrete or not) or buffer sitting inside the outer feedback loop of the input opamp. I'd imagine in this case allowing the feedback impedances to be VERY low for DC precision and eke every last drop of current noise performance out of the circuit. Plus whatever extra gain you get from the second amplifier (with all the caveats of compensation).
I noted a noise figure of 50 nV/rt Hz on the chopper stabilized amp. That's about 1 uV over a 300 Hz BW.
I think you can get to 100 pV/rt Hz using some (un)conventional tricks but much below this and it starts to get difficult.
I think Joachim has some stuff cooking in the kitchen that might be interesting
Fascinating reading - incredible technology - thanks for sharing George.
I think you can get to 100 pV/rt Hz using some (un)conventional tricks but much below this and it starts to get difficult.
I think Joachim has some stuff cooking in the kitchen that might be interesting
Fascinating reading - incredible technology - thanks for sharing George.
Last edited:
Already on 5V rails, which are easy to get a hold of on in a DAC (not that you couldn't use the 4898 on 5V!), and GBW for the I/V. Both chips are more than excellent. IIRC the 4897 is also less expensive--nope, just checked, just about the same cost.
In other words, arbitrary choice?
Okay, okay...
It's common to see confusion among IT juniors (see stackoverflow forum for example). They ask whether they need to learn this technology or that technology... They don't want the technology to be obsolete by the time they graduate from schools 😀
Computers are getting smaller, with Android seemingly the major OS and Google company seems to be the major contributor for advancement in so many related technologies.
My point, I knew it was hard to see the "importance", is that IT people are conditioned to become learning machines. They may be also forced to be more selective in their learning/reading...
For example, LTSpice is only one small software among so many software that I have learned or developped. It is admittedly the hardest because I need to also learn the processes, the theoretical electronics and amplifier design required to understand and use the software.
The main idea that you don't need to do the dirty job yourself, even if you can do it faster, is that your 1 second is more expensive than his 1 hour. But you know your mistake after it cost you the unpredictable overhead...
I made similar mistakes. When there is hardware/software failure, I think it is not my responsibility to fix it... And you know how the story goes...
About composite amplifiers in an opamp, I have never seen one I think (or they hide it in the "simplified schematic"). May be used in instrumentation because such require accuracy and often output drive at the same time...
Now about "output drive", I'm curious what has been done inside the LM4562 as it appears (perceived by ears) to have the best output drive among opamps that I have heard. It's hard to believe if it has "something" in the feedback loop to do the trick. If it does, the loop gain might be very high and the instability is understandable. I think this instability was the trade-off they took to have such a great output drive...
Jay, a composite op amp is usually made of a single IC that is attached to another IC, or a group of discrete parts, all within the same feedback network.
One of the best combinations that I have worked with, was a modification to the 5534 by bypassing the input stage and adding a quality jfet n-channel differential pair into the alternate pins on the chip. This REMOVES the bipolar input stage and replaces it with a discrete jfet input stage. It sounds very good, so good that Dave Wilson had us make them for his most expensive speaker system, more than 25 years ago. The drawback of this design is MORE noise due to the lower gain of the input pair over the bipolar pair that it replaced. For line level, this is OK. For lower level, not really the best.
One of the best combinations that I have worked with, was a modification to the 5534 by bypassing the input stage and adding a quality jfet n-channel differential pair into the alternate pins on the chip. This REMOVES the bipolar input stage and replaces it with a discrete jfet input stage. It sounds very good, so good that Dave Wilson had us make them for his most expensive speaker system, more than 25 years ago. The drawback of this design is MORE noise due to the lower gain of the input pair over the bipolar pair that it replaced. For line level, this is OK. For lower level, not really the best.
So it is not like 2 amps inside a single opamp IC then?
I think I have seen similar circuit in an old opamp PDF, in its implementation section. Also in the description of the internal operation of an opamp input stage circuit...
Quick reading on the OP275 pdf gave me an impression that the IC uses the same trick? The OP275 is the best in Mooly's test.
Yes, I'm aware of the strengths and weaknesses of JFET and BJT. But I still have no idea what's wrong technically with NE5532. It's boring and fatiguing character, when the slew rate is not that fast. If the fatigue is due to instability, what factors affect stability other than speed? Also, even if NE5532 is not fast compared to newer opamps, it was very fast long time ago, so may be it was not easy for them to achieve this bandwidth, may be they did it in not so a good way?
Many good amplifiers are JFET input. If it is good enough for discrete amplifier why not for input stage of opamp-based preamp. Besides, when I look into newer JFETs specifically from Vishay, there are many that have much lower noise and better linearity than the standard K170/J74 (with other issues in other area, but might be unrelated).
Do you have any link for similar circuit like yours? I would like to try it if the JFET is not complementary.
- Status
- Not open for further replies.