I think you'll agree the "impedance firewall" purpose has been covered adequately and quantifiably.Edmond Stuart said:
Your thesis seems to be that putting a cap across the speaker terminals is beneficial to amplifier performance by reducing RF ingress. Me and a couple of others have pointed out quantifiable drawbacks due to adding electrical mass, and potential for resonance.
Would you quantify the benefit to performance that your proposal delivers?
traderbam said:
Potential for resonance? Only if one is doing crazy things, like shorting the speaker cable. Besides, the other circuit is also prone to resonances Look here:
http://www.diyaudio.com/forums/showthread.php?postid=1558767#post1558767
and here:
http://www.diyaudio.com/forums/showthread.php?postid=1558771#post1558771
As for 'mass', that's BS. I've already explained why. Look here:
http://www.diyaudio.com/forums/showthread.php?postid=1560215#post1560215
I have done that already, but you seem totally deaf to any rational argument.
BTW, NOT only MY proposal, also from other 'idiotes' like E.M. Cherry, Bryston, Crown, NAD, etc.
traderbam said:
What you are preaching is the key to the whole "NFB bad" stupid debate. Just use a slow base amp and you are going to claim "NFB is BAD, it's increasing my distortions with significant amounts of Nth harmonic". In fact, increasing feedback should reduce distortion, provided the amplifier is operating below its slew rate limit. For signal conditions around and above the slew rate limit, more feedback will increase distortion and in particular TIM.
Only one caveat: Large SR should not be achieved at the expense of linearity. From this perspective, a local feedback loop like emitter degeneration is good, but other slew enhancement techniques (mostly used in early high speed opamps) are really poor. LF356 is a high speed (for the 70's) opamp that was/is notoriously bad for audio.
Among others, Walt Jung in the reference I cited above. Quote from that article:
"The circuit, including all possible loading conditions, should posses a slew rate of 0.5V/us (minimum) to 1V/us (conservative) per peak output volt."
You may want to go for the minimum (that makes 20V/uS for an 100W@8ohm amp) if you really have tin ears. That think is clearly audible according to Jung and others.
You don't have to treat audio as a religion, but you have to treat it as science, if you want reproducible results and community credibility. But then, as so many others, you may choose to live in the ivory tower of your own subjectivity.
This "T" network provides both a resistive RF load to the amplifier above 2Mhz regardless of whatever is connected to the output, and strong RF ingress supression without the risk of resonance with external capqacitances. 33nF is just a parasitic load for testing.
Why the usual amplifier geniuses are not using this approach? I don't know but I don't care either, I would do it that way (my way). Time is better employed for creative engineering rather than for pointless discussions.
Eva said:
That's why:
syn08 said:
They don't want to live in an ivory tower, so march together.
Hi Eva,
Your circuit is good, nothing wrong with that. But why would a sane person test an amp with a load of 33nF + 0.01Ohm + 0.1uH? As far as I know, not any speaker (+cable) exhibits such load.
Cheers,
Edmond.
Your circuit is good, nothing wrong with that. But why would a sane person test an amp with a load of 33nF + 0.01Ohm + 0.1uH? As far as I know, not any speaker (+cable) exhibits such load.
Cheers,
Edmond.
Sir, I have made no judgements at all regarding the efficacy of NFB.syn08 said:
I am going to explain this one last time and hope it sticks with some readers. Output slew rate in excess of the music signal is not a basic requirement for amplifier design.
Let me explain Jung's 1977 paper. He examines op-amps of the common architecture comprising transconductance subtractor feeding Miller style integrator and output buffer. He explains that if an op-amp slew rate is limited by subtractor stage bias current, and that if that subtractor stage is only linear over 25% of its current dynamic range, then the output slew rate for linear operation can only be 25% of the maximum possible output slew rate.
This is a simple design envelope issue. The effects of operation outside the design envelope have been variously termed Slew Induced Distortion (SID), Transient Intermodulation Distortion (TIM), Dynamic Intermodulation Distortion (DIM) and among ordinary folk, "input overload".
Jung's 25% ratio is an implemenation feature of this particular architecture. If the subtractor is linear over a wider range, the ratio changes. If the output slew rate is limited by a different stage than the subtractor, then the ratio changes. There are many ways to skin a cat. As a result, it is not correct to make a blanket proclamation that ALL amplifer designs need their slew rate to be 5x or 10x more than the signal they are amplifying.
Anyone who wishes to discuss this further, please email me or start a fresh thread. 🙂
traderbam said:
Thank you for explaining the Jung paper, I was desperately looking for somebody to help
Well, I've tried, and it's a 🤐 waste of time. Good luck and let us know how wonderful your 10V/uS amp sounds.
john curl said:
I prefer Rule of Occam 😀
Applied to electronics, it sounds: "Mix all together then discard everything that makes things worse"
Traderbam, in a narrow sense, you could be right, BUT not in a real-world sense. IF we had perfect amps, then they would not distort when pushed with more difficult signals. TIM signals are difficult signals, even if they might look OK on a spectrum analyzer. We make 'rules-of-thumb' from dozens to 100's of individual measurements to get a REASONABLE WORST CASE. That is where the multiplier comes in. However, I have measured extremely fast transients from real records, even after RIAA EQ. You should look at my IEEE paper from 1978, or the my Audio article that is very much the same thing. I MEASURED, rather than guessed and argued.
john curl said:
Yes, but today discarding one opamp I got a thermal run-away. 😀
How much power is needed to apply to 1 Ohm 5 W resistor so it unsolders and burns a chair? 😎
John, do I understand you correctly that the figure you mentioned, and now admit is a rule of thumb, is a kind of worst case requirement for all the amps you measured? That is, many or most of them got away with less slew rate without problems? Then it makes much more sense than just stating it as a hard figure for how much is needed. Probably one could even find amps where your figure is not enough to avoid TIM, even if they are rare. So after all, it seems to boil down to the usual and obvious answer "it depends", and analysis, measurements and simulations are necessary to really know how how much slew rate is necessary for a particular amp. Using your rule of thumb might then be a fairly safe way to play, but could also mean a lot of overkill for some designs, causing the designer to solve problems that really don't need to be solved.
Nothing wrong with a rule of thumb, but always tell when something is a rule of thum rather than some kind of hard fact. It avoids a lot of confusion.
Nothing wrong with a rule of thumb, but always tell when something is a rule of thum rather than some kind of hard fact. It avoids a lot of confusion.
In practice, you need MORE than 10V/us, even with expensive designs. John Meyer and I measured this for an audio magazine in 1977. We tested 4 power amps: Phase Linear 400, Audio Research D 100, Theshold (I think) and an Electroresearch power amp.
We found that 2 units passed TIM 30, with a nominal output that was below 100W, and 2 units did NOT pass. The Threshold and the Electroresearch passed, and the other 2 units failed. What the other 2 units had in common was a slew-rate that was 11V/us or less. When you see the distortion, even with 2 very different designs, it becomes obvious that 10V/us is really impractical for an IDEAL design. Now, I know that many IC power op amps are not faster than this because of cost of manufacture. OK, but don't expect perfection for every audio playback situation, with only 10V/us slew rate.
We found that 2 units passed TIM 30, with a nominal output that was below 100W, and 2 units did NOT pass. The Threshold and the Electroresearch passed, and the other 2 units failed. What the other 2 units had in common was a slew-rate that was 11V/us or less. When you see the distortion, even with 2 very different designs, it becomes obvious that 10V/us is really impractical for an IDEAL design. Now, I know that many IC power op amps are not faster than this because of cost of manufacture. OK, but don't expect perfection for every audio playback situation, with only 10V/us slew rate.
John;
were that cartridge tracking artifacts as high as the max signal level?
6 uS/V means full level of 20 KHz / half level of 40 KHz, so even 6 dB/Oct will work.
However, I have to admit that I did not face slew rate problems last 30 years since I left the camp of opamp designers, so speak purely theoretically...
were that cartridge tracking artifacts as high as the max signal level?
6 uS/V means full level of 20 KHz / half level of 40 KHz, so even 6 dB/Oct will work.
However, I have to admit that I did not face slew rate problems last 30 years since I left the camp of opamp designers, so speak purely theoretically...
Yes, moving coil phono cartridge mistracking is a full output pulse. I have captured it with a fast storage scope.