The original QUAD 33 has a switchable low-pass filter, but no 20 kHz setting, only 5 kHz, 7 kHz, 10 kHz and cancel. I don't know about the new QUAD 33.
A full power pulse with rise time equivalent to a 20 KHz sine wave would likely destroy most conventional tweeters. The worst case for vinyl source is a scratch or dust click causing the cantilever to ring as a fast damped oscillation at around 15 kHz with most MMs that I have measured.
If you are into short 0.00X % settling times, try to avoid pole-zero pairs with relatively large time constants. The pole will never be exactly on top of the zero, so you will get a small term in the step response that damps out with a large time constant. Op-amps like the NE5532 and NE5534 are then to be avoided, as are vinyl records, tapes and FM radio.
Whether the treble unit would be damaged depends on how short the pulse was. A well designed 19mm treble unit should take 1000W peaks easily with low distortion if they were only 50 usec long eg 1 cycle of 20kHz
But you are right that clicks & scratches are the most demanding 'high slew' signals with vinyl but also the most likely to clip a well designed RIAA preamp. Whether they should be allowed to do so is unfinished business for me.
If an amp is adequately compensated (and that may involve attention to the front-end BW limiting filter), there should be no overshoot or ringing into a resistive load, no matter how fast the stimulus rise/fall times are. The only fly in the ointment here would be TPC compensation, which does show in-band peaking. But, in this latter case, you can mitigate it by further limiting the input bandwidth. However, TPC shows none of these problems, and neither does MC if the phase margins are adequate.
Actually I've used similar techniques, both analogue & digital, fixed & adaptive in some experimental versions of my Powered Integrated Super Sub technology but this was at LF rather than HF. The commercial implementations were all analogue as this when computing power was $$$ ... especially good ADCs & DACs.
I've conducted/been involved in separate DBLTs to investigate several of the aspects you list. In each case, it was important to conduct the test such that these other aspects didn't obscure the aspect being tested.
https://aes2.org/publications/elibrary-page/?id=10251
https://aes2.org/publications/elibrary-page/?id=3158
https://aes2.org/publications/elibrary-page/?id=3793
https://aes2.org/publications/elibrary-page/?id=2476 this one is about more than just Intermod
You can easily dream up artificial test signals where a significant part of the population, eg >20%, can detect a 20kHz brickwall filter. In fact, I think someone on this forum effectively did just this with a poll on the audibility of clicks IIRC. I was quite pleased that 20+ yrs as a beach bum in the bush hadn't killed my ears completely. 😊
But the percentage who can do this reliably on music, speech or other 'real life' siignal is MUCH smaller. I'm really ONLY interested in the opinions of this very select group.
Isn't anyone going to post a lossless file of music/speech or other real life signal with huge slew rate demands?
At AES Hamburg 1981, one of Prof Otala's students presented a very good paper detailing his most careful hunt for unicorns with da Professor's zillion V/us slews. Only to prove conclusively that they did not exist 😲
Surely we can do better more than 4 decades later? 😊
As a maker of commercial audio stuff, I want my stuff to be salable rather than accurate of even musical. It's just that IM not so LE, musical, accurate & salable usually go together .. apart from small glitches like 20kHz brickwall filters 😊 ... provided the buyer can listen to my stuff properly.
Today, the buyer has even less chance to hear stuff before buying than in da last Millenium. So "my stuff is hand carved from Unobtainium & solid BS by Virgins" is probably the most important factor 😲
Last edited:
To come back to this. Many years ago I participated in a DBT for audibility of FM multiplex filters, that limit program material to 15kHz.
They level-fall like 80dB in the first octave and have phase shifts of 1000's of degrees at 15kHz (no that's not a typo!).
We were 6 participants, one could'nd hear a difference, the other did, statistically significant, and they ALL preferred the filter in the chain, moi included!
Jan
Just so I don't have to start paying AES dues again, could you briefly summarize a typical number of test subjects, blind protocol (ABX?), any training provided to test subjects, and examples of controls? If that's too much info to keep it brief, would be most interested in protocols used and if any particular test subject training was provided. Thx.