Interesting article, Marce. Much more involved with multimodal learning than just for voice recognition. We also use it to infer more detail about the intended meaning of particular speech uttered by a speaker. This is why people often communicate more effectively face-to-face then via voice only. And why communication accuracy suffers even more on an internet forum -- the reader may tend to infer a particular tone of voice when one cannot be heard. Attempts at humor may seem quite serious instead, for example. Hence the need for emoticons.
Multimodal learning is probably also implicated in recognition of distortion when peeking issues are taken into consideration.
When the silicon moved from one process to another smaller process... As to the rest sorry covered by NDA's and you are not a UK citizen, as to op-amps them selves I would have to delve through all my past stuff to find the numbers. The designs were 1% tol. resistors for standard and 0.1% for critical, also stuff I work on has to work correctly at a far wider range of temperatures. But these were well recorded and measured problems, so rolling op-amps can change things in my book, because the circuit has not been designed with the rolled op-amp in mind.
Only saying what happened in the real world when op-amps were changed inadvertently...
If what you say was correct then we would only need ONE op-amp as all other factors would be the same. I do know of many circuits where alternate op-amps can be used and are specified but there are circuits and there are circuits...
Only saying what happened in the real world when op-amps were changed inadvertently...
If what you say was correct then we would only need ONE op-amp as all other factors would be the same. I do know of many circuits where alternate op-amps can be used and are specified but there are circuits and there are circuits...
Did video conferencing stuff a few years ago, we got to see all the stuff regarding how effective it was compared to telephone only for getting meaning across due to facial expressions etc.
Regarding voice recognition, there is an evolutionary survival side to this, just cant find the notes at the moment, so voice recognition is in built, you could say its in our jeans.... (terrible pun!).
Exploiting the pun, old joke:
How do you sex a chromosome?
... pull down it's genes....
...argh....
How do you sex a chromosome?
... pull down it's genes....
...argh....
How many times do we have to explain this.
You say we can recognize different human voices therefore our long term memory must be good. We agree on both points.
We say you wont recognize the differences between TWO OPAMPS in an ABX test and especially so if there is a delay or gap between listening to each other (ergo claims like 'I heard the LM4562 last week and yesterday I heard the SA5532 and you know, the 5532 absolutely sucked' are not true). We say its because the differences are so slight, we cannot process the information and store it meaningfully in long term memory.
If I smack you over the head with a 1/4 pound mallet and then a 1 pound one you'll know there is a difference (analogy: human voices). If I drop two feathers on your head one after another you won't know which is the heaviest or which was dropped first in an ABX test (analogy: differences between opamps)
You are using the word memory as one simple concept, as though there were only one kind and as if all memory works the same way. It doesn't. If someone suggested that it is logically impossible for a computer memory to be fast or slow, or volatile and non-volatile, whatever, and insisted that's impossible any can't make and sense, you would might figure them stupid, or at least very ignorant.
I have a psychologist for a wife. I know there are different types of memory. Further, below certain levels of distortion/noise, the human apparatus just ain't good enough to reliably make sense of things so that you can have accurate LTM recall.
Sometimes people say things that were hard to resist when I was younger. I am holding my lips together with my free hand right now.
Yes, you can read that in a number of ways, I'll grant you that.
But its true - she used to lecture the subject.
🙂
But its true - she used to lecture the subject.
🙂
I had a wife for a psychologist, but she was married to someone else.
Perhaps the reason the two of us didn't get along (could've been three)
Perhaps the reason the two of us didn't get along (could've been three)
I'm not seeing explanations myself, I'm seeing assertions.
I'm not interested in discussing whether I would hear differences in an ABX test between opamps. You're quite welcome to your opinion.
The logic: we can have something as reference in our long term memory. Compare this "reference" to 4562 then compare this "reference" to 5532, then the delay between the two events doesn't matter much because the "reference" doesn't change...
Almost all people cannot recognize a note such as SOL in the absence of other note such as LA. So a strong reference must be built in the long term memory, such as the sound of DO-RE-MI-FA-SOL-LA-SI-DO...
That is how we can recognize a thing without having to compare with another thing (except that reference in the long term memory)...
So, it is possible to store our experience with either 4562 or 5532 in our long term memory such that we don't need direct comparison. We can even listen to only ONE and BLIND! And still possible to recognize the opamp the way we recognize voice in the absence of other voices...
From my ALL experience with 4562, it's "sound" is very unique and hence easy to remember. It has resolution and dynamics that seems to outperform all other opamps I have ever heard. At the same time it sounds nervous and fatiguing and can modify the rhythm or musicality of the music. These are both extreme performance! Such things can be easily stored in long term memory.
So whichever you remove (the resolution or the fatiguing character) the 4562 might still be outstanding, unless you can "balance" the opamp by increasing the stability and increasing the noise, e.g. by increasing the noise gain instead of signal gain...
I have tried swapping opamps in a few different circuits that tend to work the opamps pretty hard. In that case, when they cannot maintain maximum linearity through feedback, are almost starting to current limit, or with wide CM swings, it often seems that something about the same "sound" seems to follow each opamp from circuit to circuit. That's one type of thing people may be referring to when they talk about an opamp's sound. Fair or not, bad engineering or not, it is what it is. The result might sound good or bad, depending on one's taste and the purpose of the circuit.
On the other hand, if the goal is to optimize every circuit to keep the selected opamp in its most comfortable and most nicely speced zone, fine. Cater to the needs of the opamp if it suits your purposes. In that case, a single opamp stage probably doesn't have an identifiable sound.
I have a single channel 3-band parametric EQ that runs the signal through 11 opamps before it comes out the other end. Impedances, voltages, and currents are quite reasonable for many FET input opamps. Power is clean and decoupled at each IC. Stability, absence of ringing or frequency peaking are all good. Yet, it sounds different with different opamps, and they aren't working hard in this case. Some of them do operate with CM swings.
On the other hand, if the goal is to optimize every circuit to keep the selected opamp in its most comfortable and most nicely speced zone, fine. Cater to the needs of the opamp if it suits your purposes. In that case, a single opamp stage probably doesn't have an identifiable sound.
I have a single channel 3-band parametric EQ that runs the signal through 11 opamps before it comes out the other end. Impedances, voltages, and currents are quite reasonable for many FET input opamps. Power is clean and decoupled at each IC. Stability, absence of ringing or frequency peaking are all good. Yet, it sounds different with different opamps, and they aren't working hard in this case. Some of them do operate with CM swings.
Last edited:
The opamp will certainly have a "sound" of its own when you're getting it close and beyond its absolute maximum limits. 🙄
Now seriously. In my book an opamp that will allow for a larger variety of applications (Zin, Zout, Vin, Vout etc) is 'better' than a more restricted one.
When JC and Dave Wilson "rolled" op-amps they put the 741 right in the middle of the pack.
Of course one might view that as a mis-application. Nothing has been presented here better than the usual anecdotal op-amp rolling experiences, yawn.
I understand. This is similar to saying that tube, BJT and MOSFET has their own sound. Of course, as a distinct entity they don't have a "sound". The complete circuit, anyhow, tend to have its sound signature as a result of different properties and thus behavior of the gain device in amplifying signals.
The sound of an opamp is the result of their stability, THD+N, etc.
I have mentioned how an opamp may lose its sound signature:
But I have also mentioned that the context is in the preamp and:
Some opamps work better with lower gain, some work better with higher gain. Preamps are usually low-gain device, so those opamps that do not perform well at this gain will suffer. I use LM4562 only at high gain circuit.
- Status
- Not open for further replies.