Everything we perceive is contrastive, comparative. So all we have to do is create comparisons.
It's unfavourable if I missed the mark and have a fat output stage, e.g. 4 x 5 output transes or 4 x 5 power resistors, then I won't get around to swapping 20 components in order to compare;-) But what I can do is, as example, simply put different power resistors, for example 1 Ohm, at the outputs of the power amplifier. Or the psu, or main. The relative character will be the same in all functions.
But if I don't want to compare by listening, by using senses, I shouldn't talk about science or physics or technical basics - in the field of audio: to hear, to listen;-)
It's unfavourable if I missed the mark and have a fat output stage, e.g. 4 x 5 output transes or 4 x 5 power resistors, then I won't get around to swapping 20 components in order to compare;-) But what I can do is, as example, simply put different power resistors, for example 1 Ohm, at the outputs of the power amplifier. Or the psu, or main. The relative character will be the same in all functions.
But if I don't want to compare by listening, by using senses, I shouldn't talk about science or physics or technical basics - in the field of audio: to hear, to listen;-)
Hi ! thank you very much indeed again I am simulating mostly circuits with discretes If i had to try an opamp i would follow the schematics usually found in the datasheets as a start
Opamps are nice great performing some even very cheap and still they do not look fascinating They look like insects with small legs ...
However the idea that a single chip opamp could outperform something like this puzzels me a little
Thanks and this is sad Because what i like more are the THD plots
Schematics of old units can be found in the service manuals I have started from those of some units of whish i read great reviews
Those circuits should be tested even if i understand the layout cannot be loaded on the simulations
I am curious about simple buffers and preamps design I have realized now that to get very top performance complexity is needed An opamp is very complex inside I have seen some "simplified" schematics of opamps ... very impressive
I guess opamps are the exact opposite of a minimalist/purist approach The minimalist look for low open loop gain and low feedback
Opamps instead have usually huge open loop gain and high feedback Two completely different approaches
If we just look at scopes the winners are opamps for sure
It is already a challenge to get good performance at sim with discretes
If we add the fact that sim is quite far from reality and reality is almost always worse than sim the choice seems forced
"However the idea that a single chip opamp could outperform something like this puzzels me a little" I am puzzled that you think like that. Just think about all the parasitics in the image you linked and compare that to what is going on in an IC.
i admit i am very confused What could be the reason of overdesign an audio equipment if this choice will not provide some kind of improvement in performance
I have always thought that best performance is the goal A lower THD+noise better sw response higher slew rate etc.
I would love for instance t unnderstand what influences the slew rate in a preamp
i mean taken two opamps using the same +/- 15VDC power supply the slew rate can be very different
I have always thought that best performance is the goal A lower THD+noise better sw response higher slew rate etc.
I would love for instance t unnderstand what influences the slew rate in a preamp
i mean taken two opamps using the same +/- 15VDC power supply the slew rate can be very different
Now that's a bit fast, I think. Disagreeing with the content of the results of someone's investigations shouldn't really equate to discouragement of discussing the experiments.
My own foray into the "simpler is better" hypothesis had some setbacks, with resistor loaded BC550C buffers sounding markedly worse than NE5532 buffers. But it could be a different story with JFETs. The thing is, it's hard to overcome inertia and do seemingly pointless experiments like air-wiring with twisted pairs all over the place (doing all that work) just to test if stray picofarads in a feedback loop sound any different when the dielectric is polypropylene or air, vs a PCB made with "flame-retardant formula 4". Whereas most prototyping is done with breadboard or veroboard, with those tin stripes that look suspiciously like wifi-sized Yagi antennas. So, bad results achieved with breadboard, followed by a great improvement with FR4, could be another source of bias.
The to-do list also includes comparing how a CCS sounds compared to a resistor, in several configurations where it could be quite critical. If a spice simulator only suggests a 0.002% THD difference, but it doesn't include second breakdown, non-linear capacitance, line impedance, or diode junctions where dissimilar metals are soldered together, and everything runs at a fixed 20°C, real world results could easily be contrary.
The effects of layout parasitics can be included in the simulation. There are electromagnetic field solvers that can give you a parasitic model of the layout, which can be included in the simulation. But both EM field solvers and the simulators that support such shenanigans involve hefty licensing fees. I ran many such simulations when I worked at National/TI.
In the discrete world, I've included layout parasitics in the simulations in the form of a lumped circuit model of R, L, and C. You can easily estimate those if you have access to the layout. R can be calculated from the sheet resistance of copper along with the trace dimensions. L is typically 1 nH/mm, though it does depend on trace width. C can often be ignored as it's pretty small, unless you're talking about power planes.
As I think I mentioned earlier, in the IC world devices are free. Furthermore, IC design depends on the matching between components whereas discrete design depends on the absolute value of the components. That's because matching between devices can be excellent on an IC whereas it's basically a crap shoot in the discrete world unless you pay someone to match devices for you. That used to be a thing, but isn't done anymore for obvious reasons.
Because of the free devices, IC schematics will look more complicated. You'll nearly always find beta-helpers on current sources, for example. That makes a simple current source look more complicated at first glance.
If you bother to validate your simulation models and use them within their valid range the simulation should match reality quite closely.
I find it curious that you're trying to force this dichotomy. Some would consider an opamp to be just another component. Saying, "I don't like opamps" or "opamps aren't pure" shows a lack of understanding and also a tendency to want to shoot yourself in the foot. Why limit yourself? If a component gives you an advantage, use it.
Tom
https://en.wikipedia.org/wiki/Power_bandwidth
The slew rate is typically limited to the amount of current required to charge the capacitors in the circuit. Including the parasitic caps.
Tom
It might, or it might not.
The trouble with op-amps is that everything is thermally tied together, whether you like it or not. So, if the op-amp has a class-B output stage (as most seem to), even if the complementary NPNs and PNPs are perfectly interleaved, heavy grunge is injected into the power rails, which then have to be over-designed. And I imagine most of the "free" transistor real-estate is occupied by inordinate amounts of thermal compensation, so all the CCS' and current mirrors (etc) don't run wild due to the unwanted thermal cycling. Ditto weird noise performance "masked" by complex signals as everything runs hot-cold-hot-cold.
With that in mind, op-amps probably shine best with very light loads like MOSFET gates or as buffers for active filters. Not as heavy lifters.
A cascoded CCS could be a another case, where you can probably get great results with linear regulator for the CCS, but you don't want the cascode to be thermally coupled to it, so it's better off on a separate chip. That's not to say, there couldn't be some specialised IC that tries to brute-force the issue, with current dumping, temperature monitoring etc, etc. But again, the chance of weird, subtle, unfixable issues spirals together with the complexity.
And besides, there's a "peace of mind" when you go through the intellectual exercises, and you're happy with the result.
There are several terms floating around that are not clearly seperated and defined. Even here in this audio forum.
Performance, for example. If I don't know, and don't want to research, what the ear does, audio, then performance is not understood to be audio but something non-audio. We have to learn to analyze and clean up our "thoughts";-)
Most people here should start with loudspeakers: get a multi-way loudspeaker that is tuned in a highly linear way using a, e.g., highly complex crossover. Any Dynaudio, for example. And then simplify this crossover to experience how this elaborately developed and constructed, highly linear loudspeaker sounds terribly discolored and cloudy.
And then scratch your head: how can this demonstrably perfectly visual lines measured loudspeaker sound so terribly bad?
And scratch your head again: do the visual lines measurement methods correspond to the sound measurement methods, audio?
And a bit of "science": is what is claimed to have been tested, measured, actually tested, measured;-?
Hi ! i am back and thanks a lot indeed for your kind and precious advice
This sounds like space science to me Nevertheless i am sincerely admired of scientists The world has evolved thanks to them
No scientists among monkeys Science is what allows the humans to rule the world More or less
I understand and i agree completely with you But i see the challenges that comes from simple discretes circuits
One challenge is for instance using only the same bjt Usually they can be bought in a pack of ten
It is like climbing a mountain instead of using an helicopter that can take me there faster and safer
when i said that ICs are complex i was referring to their internal schematic
There are broadly two extremes ... the party of the high open loop gain and feedback (ICs) and the party of the low open loop gain and little or zero feedback (minimalism) They are two completely different ways to approach audio amp design
I have been always confused about which way to take
The minimalist designs have great difficulties in reaching the electrical performance of high feedback designs
I guess that audio analyzers are not minimalist in their design for sure
Very interesting and thank you I am more than satisfied about the checking ability of the sim software
I am quite ignorant I have simed designs that could not possibly work An educated eye would have dismissed them upfront
This is my main regret that translates in a waste of time
Sometimes it seems like the sim sw is telling to me ... the waves that come out have strange shapes Even offensive
You are right I am pretty sure that a NE5534 with some caps and resistors around could give me an almost perfect buffer or line stage And that would be the end of the story I feel like i am very close to the end of the story
Completely off topic i have looked at your excellent project The Modulus amp
Have you ever thought about adding a powerful output pair that could provide higher peak current ?
I understand that when loads (i.e. speakers) are difficult (low impedance and some capacitance) more current available is a good thing to have
The chip amp could be powered with regulated voltages having only to provide voltage gain and drive the output pair
Instead the output pair, always inside the feedback loop, could be powered with unregulated voltages
Thanks a lot So i guess that some opamps must have caps inside and other less or no caps
Hi thanks a lot but this i do not understand Some opamps work with just 10mA current
Moreover they exhibit great PSRR expecially at lower frequencies to be fair This of course will imply a better filtering of the supplies but nothing impossible Maybe a two stages regulations with monolithic regulators and discretes
Could you elaborate what you mean with over-designed ?
this is very difficult to me But my point is that some opamps based designs show electrical performance that challenge the measurement instruments The THD plus noise at the output can be the one of the testing device
With the amp adding nothing to it
If there were problems they should pop up in some way during lab testing Am i right ?
Very interesting and thank you for this explanation
Lately i have been more and more interested in high efficiency speakers for some reasons
A friend of mine showed me a very cheap amp working on 12VDC
Then he connected it to a difficult speaker The sound was very unsatisfactory
Then he connected the same amp to an high efficiency speaker with horns and high efficiency woofer
What a sound ! even exciting
My conclusion is that high efficiency is the way to go in any case
For this reason my interest about power amps stop at around 40-50W per channel maximum
Paired with some 90dB speakers with a easy impedance there is little need of more
About crossovers i cannot say much For sure they have an impact on overall performance of a speaker
Usually few or no caps. Caps on silicon are really expensive as they consume a large amount of area. And unless we're talking MIM caps (MIM = Metal Insulator Metal), which are even more expensive, the caps aren't that great.
MIM and finger caps (interlaced fingers of metal that rely on the fringe capacitance between the fingers) are often used in RF products. They're often below 1 pF.
Tom
You might also want to have a look at the Miller effect: https://en.wikipedia.org/wiki/Miller_effect
It's a neat little trick used to make a small capacitor act like a big one.
Tom
It's a neat little trick used to make a small capacitor act like a big one.
Tom
PSRR ~= the factor by which a dirty power supply is filtered for your own use. But as described above, op-amps can also generate a lot of additional noise on the rails that wasn't there in the first place.
This is a very interesting point that i have not hear of before Thanks a lot
As i said before i see two diametrically opposed approaches fighting each other
Minimalism usually results in designs with few active components (even just one) with low open loop gain and little or no feedback
Opamps usually have huge open-loop gain that is reduced by the massive application of feedback
In short, the monster seems to be feedback, at least for minimalists
As i said before i see two diametrically opposed approaches fighting each other
Minimalism usually results in designs with few active components (even just one) with low open loop gain and little or no feedback
Opamps usually have huge open-loop gain that is reduced by the massive application of feedback
In short, the monster seems to be feedback, at least for minimalists
Arguments are needed. And practical experience, experiment, trial, comes before arguments.
Once it has been proven that electronic parts modulate a signal, and not only in the taught line form but also, for example, via material resonance, then there is only one consequence: the avoidance of unnecessary parts and materials.
Unfortunately, modulation by material resonance, for example, is not taught in the standard electronics curriculum. It is problematic that commonly taught electricians assume that they have been comprehensively trained)-;
Once it has been proven that electronic parts modulate a signal, and not only in the taught line form but also, for example, via material resonance, then there is only one consequence: the avoidance of unnecessary parts and materials.
Unfortunately, modulation by material resonance, for example, is not taught in the standard electronics curriculum. It is problematic that commonly taught electricians assume that they have been comprehensively trained)-;
I once tried to purport that idea, here, and was summarily shot down by those having far more experience with such matters than I.
I've not read, to my recollection, of anyone building an electronic stethocscope for listening to power supply rails. Obviously with music playing through the device, be it preamp, power amp, player - whatever.
Basically a headphone, plus amp with gain and some input protection circuitry, for cases where your "rail" is 600V. If anything's "put" back on the rail by the devices carrying the music, you should be able to hear it.
Back in the 80's a fellow wrote me telling of an experiment which he did to evaluate op-amp PSRR. He said the op-amp rejected ALL but the most ridiculous levels - volts apparently - of injected noise. He also mentioned he didnt dare state his results in the "audio forum" we had at the time...but he wanted to share it someone.
At the time, the vogue "POOGE" (Progressive Optimization Of Generic Equipment) was to build shunt regulators with a bazillionth of an Ohm output impedance and pack CD transports with modeling clay, to reduce mechanical resonance artifacts. Dont want that laser focus servo working overtime, putting its compensation efforts back onto some power supply rail!
Back then, 40 years ago, it did seem a little exciting to go after the very lowest of the hanging fruit; power supply, coupling caps, connnecting wires, even the solder used to build something. These days - for me - it's kinda ho-hum. Still cant tell why an overbuilt behemoth would sound better than a single big hot FET with a resistor and a coupling cap to the speaker.