Shifting sands.
Violins and string sections could sound quite bland on a lot of early CDs up to about the mid 90s. Yes everything was perfectly hiss and pop free, but I later learned about the insidious jitter, and brick-wall filters needed to make most of the Nyquist rate actually usable.
After decades of battling ghosts, they simply increased sampling rates. Ironically, digital compression techniques might've had some low hanging fruit if the standard had been 196k/s instead of a much more tightly packed 44.1k.
Violins and string sections could sound quite bland on a lot of early CDs up to about the mid 90s. Yes everything was perfectly hiss and pop free, but I later learned about the insidious jitter, and brick-wall filters needed to make most of the Nyquist rate actually usable.
After decades of battling ghosts, they simply increased sampling rates. Ironically, digital compression techniques might've had some low hanging fruit if the standard had been 196k/s instead of a much more tightly packed 44.1k.
Thing one: flutter from tape was far worse than jitter from any D/A converter ever was.
Thing two: if you look at equal-loudness contours for the human ear (ISO 226:2003 curves attached), we cannot actually hear much above 15 kHz at ordinary loudness levels.
The second of the two attached images shows dotted lines above 15 kHz...the dotted lines are guesses, not actual measured data, because you can't really measure them, because people can't really hear them!
By the time you raise loudness to 110 - 120 dB SPL, 20 kHz might just barely be audible for young ears - but the same very loud 110-120 dB SPL will immediately destroy those ears (causing PSH, permanent shift of hearing), so they can no longer hear 20 kHz!
If you look at old Hi-Fi books, they often list 30 Hz - 15 kHz as the limits of human hearing. This is much closer to reality. Over time there was a kind of "inflation", or "specsmanship", and people began to talk about 20 Hz - 20 kHz hearing limits. But that is really too generous for our poor human sensory capabilities.
The motional-feedback woofer I developed was capable of flat near-field response down to 10 Hz, so I actually listened to a clean 20 Hz sine wave. It sounds like a moth fluttering inside your ear. Unpleasant, and not musical. You hear individual cycles, not a steady musical tone.
Raising the frequency to about 30 Hz changes that. Now I could hear a musical tone. It's probably not a coincidence that the lowest frequency from a 5-string bass guitar in standard tuning is 30 Hz.
All this means the gap between the highest frequencies we can actually hear (roughly 15 kHz) and the CD Nyquist frequency limit (22.05 kHz) was not nearly as tight as it appears at first sight.
Still, there were criticisms, valid or not, and engineers responded; CD players with oversampling arrived very quickly - it didn't take decades. In 1990 I bought a Sony CD player with 8-times oversampling, for only $100 USD - that was all I could afford, as a starving student living on a tiny scholarship. The first consumer CD player was released at the very end of 1982. That means cheap 8-times oversampling was available less than 8 years after the CD was first introduced to the public.
You don't actually need 8 times oversampling to get away from the brick-wall filter. Even 2-times oversampling gives the filter a lot more room to work with, as the aliasing frequency moves up to 44.1 kHz, while there isn't much audible music above 15 kHz. 4-times oversampling is more than adequate, with tons of room between the most generous estimate (20 kHz) and the 88.2 kHz aliasing frequency.
Battling ghosts is never-ending, because ghosts don't exist...they're just superstitious beliefs based on nothing!
A lot of ignorant people made comments about digital audio, starting in the early 1980s, which persist to this day. For example, you can find pictures of "staircase" waveforms all over the Internet, attached to claims that this is what digital audio signals look like.
But this is complete and utter nonsense. A crucial part of the digital audio chain is the reconstruction filter, which removes all frequencies above the aliasing frequency. In the process, it smoothly connects the dots. There are no "steps" left in the audio after it's been through the reconstruction filter.
Connect an oscilloscope to the audio output of any CD player and look for yourself. There are no steps, no staircases, just smooth waveforms. There never were steps. It was (and is) an ignorant myth.
196kb/S is mostly an enormous waste of storage space, producing no audible improvements.
There is a better case to be made for 18-bit or 20-bit A/D converters, which give more headroom and lower numerical roundoff errors during digital mixing.
But the simple fact is that CD-standard, 16-bit, 44.1 KHz audio exceeds the limits of human hearing in every way - and is vastly better than tape, vinyl, tinfoil, film, wire-recorders, and every other previous audio recording technology that existed before the CD came along.
But idiots like Neil Young weighed in with complete nonsense, and too many people listened to him, rather than with their ears.
C'est la vie.
-Gnobuddy
Attachments
Interesting.
MAYBE there is a lower frequency/minimum "clock" brain processing limit down there: we are happy (or at least not THAT bothered) by 50/60Hz light from Fluorescent tubes, which actually strobe at that frequency, same with CRT TVs, yet intensely annoyed if frequency becomes half that, where we perceive light as strobing or flickering.
There must be a biological reason behind that.
Do these have to be exclusive, at least at the low frequencies you've mentioned? Surely the steady state figures higher in what we hear.
I've been reading through the 1940s. I think H-P's oscillator with 20-200 on knob times 1, 10, or 200 on switch was the origin of "20cps to 20kcps".
I did some recording at a friend's studio a couple of years ago. Recording at 24 bit, 96 kHz sampling. When I listened to the live monitor feed of my electric guitar, versus the playback of the recording, I could literally not distinguish any difference between the two. For recording and mastering, the dynamic range and noise floor of 24 bit resolution is stellar. I've worked with ADAT, and earlier on with Studer reel to reel (with Dolby noise reduction), but the first time I tried 24 bit/96kHz it was an "ear opener". I think for playback of produced media at 18 or 20 bit would be more than enough resolution. But very good Audio Interface can be had for a couple of hundred dollars, and 24 bit 96 kHz is easily attainable.
But everyone knows that the human hearing obeys the law that the rang has to work out to major whole numbers (not sure if I am coining a term), 10k is too low so nature gave us 20 khz. And if you do not believe it why do you think 20 Hz is also built in to us? 10Hz is too low and if it was at 40 Hz there would be no scooped out metal. And where would we be then?
Got to leave now, was hoping for some SS amp content, I think I will look over there.
Got to leave now, was hoping for some SS amp content, I think I will look over there.
I had exactly the same thought. I think you are quite right.
Not only flashing light, not only very low frequency audio, there is a similar effect if you touch a vibrating object. If the object vibrate at under 20 cycles per second, you feel individual "bumps". If the object vibrates at 30 Hz or more, it starts to feel solid, as though it's not moving.
Our nerve endings and neurons can only fire so fast, and I think when sensory stimuli occur above that frequency, the brain shifts to sensing an average amplitude rather than individual events. So we see average light intensity once a light is flashing above 30 Hz, hear a steady musical tone once the frequency rises above 30 Hz, feel a steady force once the vibrations move above 30 Hz.
Of course 30 Hz is an approximate number, and varies not only from one person to another, but even with your level of mental alertness.
That's only a guess, but I think it's a plausible one.
Vocalist Tim Storms can sing notes far below the lowest audible frequency. If you use good headphones, you can feel individual pulses from the lower notes he sings in this video: https://twistedsifter.com/videos/worlds-deepest-voice-tim-storms-singing/
-Gnobuddy
I think (only a guess) we only hear a steady state when the individual stimuli occur too fast to be individually detected.
For me, 20 Hz was "moth fluttering in my ear", and 30 Hz was a clear musical tone. The 10 Hz range in between those two frequencies was hard to characterize either way. Musical no-mans-land for me.
-Gnobuddy
It seems like nearly all attempts to emulate vacuum tubes with solid state devices focuses on triode preamp stages. Does anyone know of an attempt to emulate tube power amps?
I've been tinkering with a solid state circuit for the past year or so that emulates the classic 4-input Marshall (or, rather, modded Fender 5F6-a 😉) circuit, including the phase inverter, power tubes, output transformer (to a limited extent), speaker load, and power supply. Since this style of amp has no master volume, the way the power amp behaves when overdriven is very important to the sound. I've found that the dynamic operating point shift (bias excursion) caused by grid conduction and screen grid voltage sag has a big impact on the sound and 'feel'. In my design, the power tubes are modeled using nonlinear unidirectional current sources coupled to a (scaled) loudspeaker-like load plus some additional components that emulate the output transfomer magnetizing inductance and HF characteristics.
I think it sounds very good, but unfortunately I don't have access to a real amp to compare it to... I've been using LTspice to investigate the behavior of the original tube circuit.
I've been tinkering with a solid state circuit for the past year or so that emulates the classic 4-input Marshall (or, rather, modded Fender 5F6-a 😉) circuit, including the phase inverter, power tubes, output transformer (to a limited extent), speaker load, and power supply. Since this style of amp has no master volume, the way the power amp behaves when overdriven is very important to the sound. I've found that the dynamic operating point shift (bias excursion) caused by grid conduction and screen grid voltage sag has a big impact on the sound and 'feel'. In my design, the power tubes are modeled using nonlinear unidirectional current sources coupled to a (scaled) loudspeaker-like load plus some additional components that emulate the output transfomer magnetizing inductance and HF characteristics.
I think it sounds very good, but unfortunately I don't have access to a real amp to compare it to... I've been using LTspice to investigate the behavior of the original tube circuit.
No, not really. that seems mostly unexplored territory.
Sounds very interesting, how did you do that? I would like to hear more!
I'm not going to provide precise details at the moment because I might attempt to make a salable product from my work at some point. I don't want someone else to get there first with my ideas 🙂.
But basically, if you look at any power pentode/tetrode datasheet, you'll see that it acts approximately like a current sink over a wide range of anode voltages. You can make a basic voltage-controlled current source/sink with just a BJT and a resistor, but it's much too linear compared to the tube. If you replace the resistor with an appropriate nonlinear element, you can approximate the tube's transfer curve (assuming constant anode voltage). There's obviously much more to do to get it to behave like the tube, but there's the basic starting point.
But basically, if you look at any power pentode/tetrode datasheet, you'll see that it acts approximately like a current sink over a wide range of anode voltages. You can make a basic voltage-controlled current source/sink with just a BJT and a resistor, but it's much too linear compared to the tube. If you replace the resistor with an appropriate nonlinear element, you can approximate the tube's transfer curve (assuming constant anode voltage). There's obviously much more to do to get it to behave like the tube, but there's the basic starting point.
You would've loved the persistent squeal of TV and CGA monitors, and the fourth-wall-breaking weirdness of 95% of other people being oblivious.
And maybe tinnitus is just the sound of the big TV screen in the sky? Cue X-Files music
I remember similar conversations with people who insisted that it was factually established that 24 frames per second was all that would ever be necessary to produce perfectly smooth video motion.... Until the dying days of CRT, with PC gamers enjoying 100 frames per second that spoke for itself, forcing TV manufacturers to play catch-up.
A similar approach spilled over into car tail-light design, with the early LED models producing very obvious stroboscopic effects, because there was "no way" that anybody could possibly detect 100-200Hz flicker.
It's the clocking on the mastering side that is critical. At 44.1k, that's only ~22us per sample, and if there's a tiny amount of noise in the clock timing down to the nano-second range, the wrong amplitudes will be recorded.
Try drawing a sine wave on a piece of graph paper, then sample it by drawing dots spaced out so that 5 cycles only get 13 dots. Then try to recover the original wave by using only the dots. That is the problem in a nutshell. The slightest imperfection in the timing of the dots, or the curve-fitting scheme, and not only does the "16 bit" best-case resolution go out the window, but the wave-form completely changes shape, introducing amplitude modulation.
OK, so (5/13) * 44100 = 16.9kHz. But it's amplitude modulated, so lots of other frequencies are introduced.
1 and a half 😉 very good ones :
Best, hands down, is Peavey Transtube.
They emulate the power stage including pentode bias shifting, the way they clip, PI clipping and PI interaction with power tubes.
"Cheating" (because they use a 12AX7) is VOX Valvetronix.
In my point of view, what they do after the dual triode is way more important and that´s their merit, just clipping a 12AX7 and feeding the output signal is easy , made by everybody (Valvestate) and not the real thing.
Mr Pritchard and Bob Quilter (QSC) have made very good sounding emulations, as well as Boss Roland Blues Cube amps.
Good, you are focusing on the important points.
Build one!!!!
Simplest which still sounds very good is Orange Tiny Terror, schematic is widely available, components are easy to find.
Simulations are a powerful tool but there is TONS of important data they can´t give you.
Such as feeding that overdriven amp into a 12" Eminence/Jensen/Celestion speaker inside an open back cabinet screaming at your face.
Jensen MOD speakers are very good, inexpensive, and "halfway between British and American" so great for shop/lab use.