BTW, Bear, imho, it is not that bad. This thread is about what is wrong with opamp. Normally, people or OP wanted a justification for why many people seems to think that opamp is not "high end"...
I believe that in next 10 years, this will be questioned and questioned again. If previous answers couldn't satisfy the questioners, it means the data is not sufficient to support the (extra-ordinary) opinion that "opamp is wrong"...
That is why imho, the discussion about "is it true that people can hear opamp?" is actually needed, otherwise the question will definitely never be answered for the next 10 years.
I, for one, (may be many others too) think that I know the answer to the question. It is just not easy to move the message across.
That seems like a good idea in theory. In practice, there seem to be a huge number posts that go off topic for a post or a few, but not a whole lot. And some that veer off topic maybe a little, but are still related to the main topic to some extent. For example, it seems reasonable when talking about how opamps do or don't sound different, to talk about how hearing abilities affect determining how opamps do or don't sound different. And to dig deeper into determining how opamps may or may not sound different, we might need to dig deeper into hearing and testing or confirming hearing abilities. Since all this stuff is closely linked, it's not always obvious when to separate threads until it already is kind of too late, and any diversion seems to be nearing an end.
On the other hand, this is the lounge section, and the forum is for the mutual enjoyment of the bulk of the people who choose to participate in it. If most people are okay with diversion into talking about bicycles (for example) for a while, perhaps they should have some liberty to do that. In the end, keeping the forum a vibrant community is probably the most important thing, and if being too strict with law and order concerns interferes with that, then that could end with fewer good, knowledgeable, helpful people liking to spend time around here. It could go too far either way, too loose with rules or too tight. The members and moderators can judge what seems to be the best fit overall. But, your voice has been heard.
Why people cannot hear opamp in ABX
The above is a clear example/reason/explanation about my opinion that when people cannot hear things in ABX, it doesn't mean that the difference doesn't exist or not audible or it doesn't affect them... Human skill is so poor...
It took me a long time to find the right segment where audibility is possible to perceive. If I couldn't find the right segment, do you think I can do the ABX? No! I have checked and ABXed every notes along the sound clip and I was successful only with 2 segments...
In blind tests the case is similar. We have to check a lot of things (musicality, fatigue, second order distortion, etc) un-blind first (or blind but "unscored" or in "training mode"). And only after we know the difference between the DUTs we can do it blind.
The above is a clear example/reason/explanation about my opinion that when people cannot hear things in ABX, it doesn't mean that the difference doesn't exist or not audible or it doesn't affect them... Human skill is so poor...
It took me a long time to find the right segment where audibility is possible to perceive. If I couldn't find the right segment, do you think I can do the ABX? No! I have checked and ABXed every notes along the sound clip and I was successful only with 2 segments...
In blind tests the case is similar. We have to check a lot of things (musicality, fatigue, second order distortion, etc) un-blind first (or blind but "unscored" or in "training mode"). And only after we know the difference between the DUTs we can do it blind.
Oh. I just DLed and renamed them (to mp3, obviously) and right-clicked "open with Cool Edit." Cool Edit clearly pays attention to the header in the start of the file rather than the file extension.
But yeah, it shows these are the two files with 0.2dB difference.
This brings to mind the often subtleties (or even not-so-subtle things) that are sometimes not heard at all.
Circa 1997-1999 I downloaded an MP3 file and some software to play it onto my computer. I had the soundcard going through my stereo, and at first listen I thought it sounded just like a CD (though I didn't have a CD of that exact song - it was "Cows With Guns"). I had been reading online complaints about mp3's not sounding as good, but I didn't understand, and wondered what other people might be hearing that I didn't.
I downloaded an MP3 encoder and ripped a CD (a jazz thing with lots of snare drum), and listened to encodings at different bitrates. 128k sounded like what I was hearing from downloads, but lower bitrates had some smearing on the snare (and maybe other effects, the overall "underwater" type sound, but the snare was what I noticed the most), as if each strike were being faded in and faded out instead of the instant starting sound I was familiar with. But from doing that, I learned what to listen for, and I could then hear the smearing (though of course to a lesser extent) in the 128k encoding. The veil had been lifted, and I could then see the man behind the curtain. All I needed was a pointer to know which way to look.
And so it's my hypothesis that it's not that the "average person" doesn't or can't hear these differences, it's that they don't know what to listen for.
There are these Moulton ear training CDs (and I presume others have similar courses available) that I've known about for a really long time (15 years?), and I've considered buying them. I could go through these and it would surely help me hear some of the things others are discussing such as differences in op-amps (see, this post IS on topic!) and other rather low distortion circuits, but I've hesitated to want to hear THAT well, that it might reduce my enjoyment of music, especially from a mid-to-low quality source where I could "know" what's wrong with it after a few seconds of hearing it.
I already have problem listening to the pop music FM stations in the last decade or two. I find myself gritting my teeth, and it's not so much about singers and songs, but I wonder "is the radio station's signal too low? Why does it sound so bad and distorted?" Then I think "Oh, yeah, that's just the style of production, mastering, and radio station compression/limiting/actual clipping that goes on thesedays."
It has been my experience that op amps, either discrete or IC based, are not the very best kind of amplification stage for audio. IF we have to use an op amp design, then highest open loop linearity, and highest open loop bandwidth are to be sought after. The best designs are open loop and therefore not usually a typical op amp topology. The reason is that our test equipment only tests for AM types of distortion, and excludes FM types of distortion. If we had better test equipment, that measures the FM component as well as the AM component, then we would most probably be able to make our measurements match our subjective discrimination.
I REALLY want to know what EXACTLY is meant by "different treatment to work properly". It's well-known that all of the newer, faster opamps should always have power supply bypass caps located right at the IC power pins; paralleling the feedback resistor with a small value cap helps avoid oscillation; PCB traces should keep inputs and outputs separated; input pins should not have long traces attached. But these caveats are for ALL circuits.
The only other real difference I can see is that bipolar opamps will have lower noise with source impedances less than ~4K Ω (microphones, phono cartridges); FET opamps will be quieter when the source impedance rises ABOVE ~4K Ω (guitar pickups, tone controls).
But WHAT OTHER differences call for different opamps?
I think you have already know the answers...
Most of those treatments have trade-off, so the optimum one is rather hard to achieve: Adding extra cap when it is not needed will only degrade the performance.
Stability is only a qualitative term. Many fast (slew rate) opamps, no matter how you pay critical attention to the application, tend to show its instability in the form of "nervous" sound that soon becomes fatiguing. Sometimes it can also be found tho very slightly in OPA2134.
That's why many commercial products use JFET for the buffer and bipolar for the gain stages. I use LF353 for (input) buffer and JRC2068 for gain stages. Some opamps (especially when inverted) need buffer when they have to drive low impedances.
Gain. Some better as buffer, some only stable at certain minimum gain. For everything there is "sweet-spot".
Supply voltage of course. Current, etc. Basically it is just like discrete circuits so it can be seen as ordinary discrete amplifier.
That sounds great! The power limiting must then be a gradual thing. Good to hear.
Only one of my amplifiers has a relay in, but it doesn't need it. The turn on and off pop is very minimal with my designs. My current amp barely makes a sound (a faint click) when turned on, nothing when turned off, just a fade out when the rails drop too low to sustain the current sources.
If you really want to know more about that, and there is a fair amount that could be said, a good place to start: https://www.amazon.com/Art-Electronics-Paul-Horowitz/dp/0521809266 Not cheap, we know, but it's a pretty thick book printed on thin paper. Some of the information about opamps includes where manufacturer data sheets sometimes puff up some specifications, and then gloss over the not-so-good features of their products. The authors did some work to compile some opamp comparison tables for different sets of more or less similar opamps. I don't know if information organized for you like that is available elsewhere. Of course, by the time any book is published it is already less that fully up-to-date. Still, it's worth reading and a very good way to start.
If you raise the open loop gain, PIM reduces John. Cordell shows this in his book and has written a paper on it. PIM where it exists in a modern opamp is certainly down at lower than -100 dB and probably well below that. Maybe one of the experts can plug an exact number in for us using 140 dB OLG and a -3dB OLG BW of 3 Hz (itself well below human hearing range).
There is of course a place for discrete and I don't decry anyone's efforts in this regard. However, it is simply disingenuous to make sweeping generalizations that discrete is better than Opamps, and specifically IC variants whether you want to talk measured or subjective performance.
There is a reason IC opamps were used in the LIGO instrumentation, why they are used in space probes, medical and industrial instrumentation, and why for professional audio they are the building block of choice: they are just so damn good.
🙂
Last edited:
Here are a couple of examples of kinds of information summarized in The Art of Electronics, 3rd Edition. As you can see, engineers may find some pretty interesting info.
Attachments
Last edited:
I load the two tracks under test into Foobar, usually down in the list so not as the top two tracks.
With the mouse I can instantly jump between tracks, and also play particular sections of the two tracks and repeatedly.
By this method I easily discern fine differences between tracks and it doesn't take 'exceptional hearing powers unattainable by other humans'.....in fact I thought anybody could do this easily.
Dan.
Yada Yada Yada, and the answer after more than half a K posts still is that the best tire is a 210/55 on a 17 inch rim.
The topic of this thread was : what is wrong with opamps !
Even if a discrete amplifier circuit would/can sound better than one with an IC core, which would that be ?
A phono pre has discrete input stages, the best require dual monolithic parts as AD's MAT02 bipolars or Toshiba JFETs in plastic or metal can, the very best require selected parts. (let's skip SE with a bunch of Phili JFETs in parallel)
Only a handful of fools around the globe such as me have those in quantity, plus styroflex by the many thousands for RIAA correction. On top of that, there's hardly a DIYer left who's interested in top of the pops phono preamps.
For CD, I know of one guy only who went for discrete I/V.
Please do post a link to a schematic of a proper discrete output stage for CD.
Power amps ? How long ago has it been since I saw a really cool front-end schematic with an IC input stage.
Oh Yeah, Mr. Popa Ovidiu's baby. But I didn't notice a really large crowd standing in line to assemble that design, or the fully discrete one before that.
Which leaves line stages.
Yesterday I saw the measured specs of Jan Didden's Silentswitcher.
A couple of those added to a properly designed and laid-out IC opamp stage not only enable a preamplifier of low height and overall small dimensions, also one with a building price tag which will be hard to beat by fully discrete designs costing multiple times the amount.
Which also begs the question, if the very best discrete line stages outperform opamp preamp designs, they'd be preamplifiers costing well beyond 10k, who owns one ?
I know I do, but I seriously doubt any of the opamp rock&rollers are in the same position.
Frankly speaking, anyone ever seen an opamp roller mention the reference loudspeaker brand/model used for evaluating sound differences of opamps ?
What I've seen is attic living arrangements, cartons of two buck chuck, a live connection to the web, and rather poor soldering skills (if any).
Could be swapping out opamps saves them from jumping, but gather enough nutcases and the entire globe sings Hallelujah on the internet.
Last edited:
Correction/explanation to my post #572.
The OL BW on a modern opamp is typically well below 10 Hz, some as low as 3 Hz - talking VFA here of course. The open loop gain at LF 120 to 140 dB and GBWP 50 MHz or more.
When you close the loop, the loop BW will of course move up in frequency, and assuming typical line level gains will be in the MHz reagion. On older opamps like the 5532/34 with lower GBWP, it will still be 100's of KHz or low MHz. PIM in these situations, assuming of course it is even measurable in a modern opamp (high loop gains) will therefore lie orders of magnitude above the upper human hearing limit.
The OL BW on a modern opamp is typically well below 10 Hz, some as low as 3 Hz - talking VFA here of course. The open loop gain at LF 120 to 140 dB and GBWP 50 MHz or more.
When you close the loop, the loop BW will of course move up in frequency, and assuming typical line level gains will be in the MHz reagion. On older opamps like the 5532/34 with lower GBWP, it will still be 100's of KHz or low MHz. PIM in these situations, assuming of course it is even measurable in a modern opamp (high loop gains) will therefore lie orders of magnitude above the upper human hearing limit.
Nice info - pity some of the modern devices are not in there - I can't see 4562, 49710 etc.
As an aside, I'm just doing some work on a new MM phono stage.
I will use a BF862 JFET - what else 😉 - in front of an opamp. This is one application where you can get a clear benefit over a straight IC opamp by using discrete devices - about 8-10 dB noise wise. The noise voltage of the BF862's is very low at <1nV/rtHz and there is virtually no current noise term to have to worry about - important when dealing with a transducer with large inductance.
However, you have to pay for it with circuit complexity, and you get less benefit than worth it if you do not keep feedback network and gain setting resistor impedances low.
You can also use JFET's like this for an MC head amp and by paralleling easily get to 0.5nV/rt Hz. Benchmark MC stages claim 0.25nC/rtHz noise levels, but these require up to 8 devices in parallel and there are other issues to deal with.
DC input coupling is an added benefit if that's what floats your boat.
I will use a BF862 JFET - what else 😉 - in front of an opamp. This is one application where you can get a clear benefit over a straight IC opamp by using discrete devices - about 8-10 dB noise wise. The noise voltage of the BF862's is very low at <1nV/rtHz and there is virtually no current noise term to have to worry about - important when dealing with a transducer with large inductance.
However, you have to pay for it with circuit complexity, and you get less benefit than worth it if you do not keep feedback network and gain setting resistor impedances low.
You can also use JFET's like this for an MC head amp and by paralleling easily get to 0.5nV/rt Hz. Benchmark MC stages claim 0.25nC/rtHz noise levels, but these require up to 8 devices in parallel and there are other issues to deal with.
DC input coupling is an added benefit if that's what floats your boat.
I do not consider myself to be an "opamp roller".
However, I'm rather pizzedorf that rather consistently I hear noticeable differences whenever I do swap out opamps.
One reason I started to read this and other opamp "rollinig" threads is to see if anyone has come up with something that is reasonably definitive as to WHY.
I do not know "why" at this point in time.
Most recently I hauled out a 4 year old Chinese DAC board, ebay cheapo, that was the latest and greatest chipset back then (24/192). It sounded pretty bad actually. That's why it sat around doing nothing for a long time. Long story short, the reasons I got to this now being unimportant for now, swapping out opamps dramatically altered the subjective presentation. Numerous visitors have heard the series of opamps in the output buffer position of this DAC. All have reported their same "favorite".
Opamps used included the 5534, AD 797 and the LME49710, and one other.
The 5534 was stock in the unit.
The one that was clearly the best was the "other".
All sounded very much not the same - it is clear and obvious.
Now, I want to do some measurements.
To do this I've downloaded some digital files that are squarewaves, sinewaves to look at the waveforms. Also sweep files to check freq response too. In addition I have some high quality FFT gear.
Time and energy are my enemy here, it's tough to find the time to set up these things and run the requisite tests. 🙁
If an when I do, I'm more than likely to post the results and images...
As far as my system, right now this one being used is:
Quad ESL57
BEAR Labs SE Mosfet amps (35 watts DC coupled SE Class A - no output inductor)
Discrete stepped L metal film attenuator with large surface silver wiping contact switches (same for the source selector switch)
Whatever input/source device (DAC for example).
All BEAR Labs Silver Lightning interconnects w/silver plated RCAs (stranded silver 19ga wire, PTFE insulation, 3 wire weave)
Minimalist clean. Zen system.
Now this is not the first time I've been confounded by sonic differences.
Years ago now in a 4th order LP filter for subwoofers, changing the opamps altered the subjective perception of the bass rather substantially. Going from "soft" or "fatter" to "lean" and "tighter".
Don't think this can be handed off to "oscillations" or "bypass" caps, etc. The bandwidth is so limited that it ought not make any difference at all. That's another one that might yield to today's measurement gear. There ought to be nil "distortion" differential in such a bandwidth limited application. Ought to, but that remains to be determined by measurement.
Since apparently no one has tried to do measurements that follow the subjective sonic differences, which is what I was hoping to find, I guess that somehow I'll need to set up to do this investigation. Time is the problem. 🙁
I find the sonic results confounding, quite frankly.
_-_-
Last edited:
- Status
- Not open for further replies.