I just gave you two ways to do it. If the "extra stuff" causes an issue, it will be heard, since it's associated with the DUT.
Why not do multi-generational loopback to really tease out the sonics? High quality ADC's are one of the best SNR devices we have and would really do well to tease out the sound of any amp.
Hi Stuart,
I don't see my terminology as being the inconsistent hangup, at least not compared to the hangups about 'accuracy'. An amount of accuracy is of course needed, when I
I put on a Loius Armstrong recording i dont want to have it come out being Vivaldi no matter how much more pleasant Vivaldi might be to tired ears.
But I don't agree that every bit of precision of some type is always preferable or more helpful to creating an impression or some type of human perception that we're going after. You can often see better (in tems of what is important to a human) through frosted glass if you wet it, even if what comes through is less accurate in terms of color spectra. In data modem development (in my former siignal processing profession) adding filtering and even nonlinear processing (limiting, squaring, thresholds) are commonly used to extract more information (accurate demodulation), even though the signal itself is being degraded were it examined in terms of how faithful it is to its carrier. Again, we cheerfully degrade he "accuracy" of our sound fields with self engineered ambience and add room liveness, which would totally fail a Diffmaker test far beyond what a percent or two of nonlinear distortion on peaks might do. It might even be argued that a lot of the convincing effect of our sound systems is provided by the room and the moods or expectation we bring in there, not by being carried in the recordings.
To make another (probably poor) analogy (to be torn apart😊), the attitude that waveform or spectral accuracy is some kind of superior moral high ground to defend is like arguing about the accuracy of the translation of a few pages of a foreign language novel - the concept of relevant accuracy has only vague meaning, since the desired product is to relate a human abstraction or emotion, not an integer value with a relevant absolute meaning. Treating accuracy as the highest concern is even less defensible when you notice that most of the pages are missing, as well!
I dont argue that the recordings aren't good or effective. I like them a lot! But they are certainly flawed if "accuracy" is the claimed standard, they have not captured an original soundfield only someone's attempt to communicate it.. They more capture the choices of microphones and placements, mixing, and yes, effects added by the recording people as they try to relate those human abstractions and emotions.
Accuracy is only a partial goal, we don't even WANT accuracy when playing the sound in the room (I became most convinced of this since I started playing with sound diffusers in my room. Dry sound with all stray sound absorbed might be more accurate to what is on those two channels. It is in no way more realistic nor perceptually true to the sound in the original 3D recording space.
Bill, I think we're agreeing, but I don't think that YOU think we're agreeing. 😀
My personal design choice is having the electronics be accurate, that is, to replicate the input signal as closely as possible (but scaled in amplitude, of course). If I want the signal altered, I have tone controls, software EQ, and about a zillion plug ins. I can adjust them or turn them on/off.
If I go the route of having the boxes of gain be effects boxes, I have the same effect regardless of whether the recording is a close-miked Louis Armstrong multi-miked mix or a distant crossed figure 8 recording of Vivaldi. Some may want that, and they're welcome to that, but it strikes me as a very constricted way to reproduce sound. It makes more sense to me to have the electronics be accurate, then dial in the effects as needed, varying by recording type.
My personal design choice is having the electronics be accurate, that is, to replicate the input signal as closely as possible (but scaled in amplitude, of course). If I want the signal altered, I have tone controls, software EQ, and about a zillion plug ins. I can adjust them or turn them on/off.
If I go the route of having the boxes of gain be effects boxes, I have the same effect regardless of whether the recording is a close-miked Louis Armstrong multi-miked mix or a distant crossed figure 8 recording of Vivaldi. Some may want that, and they're welcome to that, but it strikes me as a very constricted way to reproduce sound. It makes more sense to me to have the electronics be accurate, then dial in the effects as needed, varying by recording type.
1.) I see some arbitrary lines drawn: most seem to *want* the emotional experience, regardless the equipment. There is a large subset that wants their amps to be functionally neutral and deliberately add effects from tone controls (whichever source of that tone control). There is another subset that goes whole-hog on the emotional experience and run the spectrum of not caring to deliberately choosing the distortions they prefer. I see you and SY as largely agreeing.
2.) The quote above: are you stating you do or don't like the diffusers? I am not parsing this correctly.
I like diffusers as a tool. In my setup they made a lot more positive difference than choice of amp between precise Hypex or sound altering SET.
Sy, again arguing ideally and precisely, sure it would be better to have a box with knobs ahead of an waveform-precise amplifier. But practically, will you really want to adjust all those for each recording (more likely each track)? Doesn't that kind of screw the whole point of relaxing and concentrating on the music rather than the gear? I use an FIR eq, but I don't keep the controls open on the computer that sets them up. I set them for best overall compromise and don't think about them when I'm in music mode (as opposd to "audio" mode!). Now if we made a media center where tone control and other effects settings were stored with each track and automatially applied, and if anyone wanted to obsessively set them all at each listen, then we'd maybe have something. Maybe.
I'm actually in process of making an amp with class d out and triode input with a "niceness" knob to selectively voltage starve the triode. I'll report how that goes (though we all know that it wont be an accurate or likely repeatable experiment, people being abstraction obsessed creatures rather than metal micrometers). "Objectivists" (which I'm probably more aligned with) talk about how poorly people can judge how audio is affected, but it goes both ways - it also means that all that picky ppzillion stuff doesn't ave a lot to do with the goal of music listening and illusion portraying, which are better served even by the snake oil treatments that ARE frankly better targeted at how human perception operates. (Do I hear incoming flamethrowers? 🙂 ).
Sy, again arguing ideally and precisely, sure it would be better to have a box with knobs ahead of an waveform-precise amplifier. But practically, will you really want to adjust all those for each recording (more likely each track)? Doesn't that kind of screw the whole point of relaxing and concentrating on the music rather than the gear? I use an FIR eq, but I don't keep the controls open on the computer that sets them up. I set them for best overall compromise and don't think about them when I'm in music mode (as opposd to "audio" mode!). Now if we made a media center where tone control and other effects settings were stored with each track and automatially applied, and if anyone wanted to obsessively set them all at each listen, then we'd maybe have something. Maybe.
I'm actually in process of making an amp with class d out and triode input with a "niceness" knob to selectively voltage starve the triode. I'll report how that goes (though we all know that it wont be an accurate or likely repeatable experiment, people being abstraction obsessed creatures rather than metal micrometers). "Objectivists" (which I'm probably more aligned with) talk about how poorly people can judge how audio is affected, but it goes both ways - it also means that all that picky ppzillion stuff doesn't ave a lot to do with the goal of music listening and illusion portraying, which are better served even by the snake oil treatments that ARE frankly better targeted at how human perception operates. (Do I hear incoming flamethrowers? 🙂 ).
Last edited:
More likely than have one-setting-fits-all, and it can't be turned off. That seems... Procrustean. 😀
In reality, I think a tilt control would take care of 99% of anyone's needs.
Bill--thanks for clarifying. And looking forward to your experiment! Glowing tubes are super cool and therefore must make things better. 🙂
At least for me (saddled firmly in the objectivists camp), I don't see anything wrong with your coming experiment, as I'm confident you will be very qualified in your reporting. "I like this sound" is totally cool with me! The thing I get prickly about is when someone states something subjective is absolutely better and then tries to wrap some kind of pseudoscientific jargon around their (subjective) justification. It's especially trying when that flies in the face of much harder evidence.
I presently listen to music through an old Panasonic class D receiver that has TI's equibit amps in it, so ppm is lost on me. Presently happy, although I'm due to clean things up a lot (primarily redoing xovers), and want to build up some absurdly high GFB amps more as an engineering exercise than for any particular sonic reason. It's fun building stuff.
At least for me (saddled firmly in the objectivists camp), I don't see anything wrong with your coming experiment, as I'm confident you will be very qualified in your reporting. "I like this sound" is totally cool with me! The thing I get prickly about is when someone states something subjective is absolutely better and then tries to wrap some kind of pseudoscientific jargon around their (subjective) justification. It's especially trying when that flies in the face of much harder evidence.
I presently listen to music through an old Panasonic class D receiver that has TI's equibit amps in it, so ppm is lost on me. Presently happy, although I'm due to clean things up a lot (primarily redoing xovers), and want to build up some absurdly high GFB amps more as an engineering exercise than for any particular sonic reason. It's fun building stuff.
Yup, agreed building is fun. Its mostly why my most recent amp build is a SE zgnfb class A job....because I can. It has the added benefit that anything buiilt diy, unless obviously screwed up, will sound clearly better to the builder than all those inferior designs other people do!
I also get prickly about people insisting that something objectively meeting a standard of some kind will be better suited to the job than something with other characteristics. The job being, of course, to trigger certain desired brain synapses. 🙂. The most irrational things can be involved in that.
I also get prickly about people insisting that something objectively meeting a standard of some kind will be better suited to the job than something with other characteristics. The job being, of course, to trigger certain desired brain synapses. 🙂. The most irrational things can be involved in that.
Last edited:
numbers and the feedback debates
I had occasion to measure some simple supertriode circuits as a favor for a friend. He wants to use them, perhaps even combining outputs of more than one, as a stepup stage for MC cartridges. He's commented that transformers, except for perhaps the really expensive ones, sound "dead" to him.
Although batch variations are worrisome, the best devices were better than I expected by far, both in distortion and noise. At levels already high for typical MC, the distortion, predominantly second, was about 30ppm when the plate load was 100k. Aggregate noise in a 400Hz - 22kHz bandwidth worked out to as low as 1.8nV/sq rt Hz (and it didn't deteriorate that much going down to 22Hz). Microphonics will require some good vibration isolation, but the high Q of the main resonance, around 10kHz, means it will also be fairly difficult to excite it that readily.
But my remark was: wouldn't it be ironic if the performance were specified, and critics be skeptical that the piece could sound any good because the numbers were so low?
I had occasion to measure some simple supertriode circuits as a favor for a friend. He wants to use them, perhaps even combining outputs of more than one, as a stepup stage for MC cartridges. He's commented that transformers, except for perhaps the really expensive ones, sound "dead" to him.
Although batch variations are worrisome, the best devices were better than I expected by far, both in distortion and noise. At levels already high for typical MC, the distortion, predominantly second, was about 30ppm when the plate load was 100k. Aggregate noise in a 400Hz - 22kHz bandwidth worked out to as low as 1.8nV/sq rt Hz (and it didn't deteriorate that much going down to 22Hz). Microphonics will require some good vibration isolation, but the high Q of the main resonance, around 10kHz, means it will also be fairly difficult to excite it that readily.
But my remark was: wouldn't it be ironic if the performance were specified, and critics be skeptical that the piece could sound any good because the numbers were so low?
I 've considered building an SE amp a la NP. I note distortion at 1 W out is low at c. 60 ppm, and then rises to about 0.5% at just under full power (sim). But, I am getting lots of harmonics out to n> 10.
Second are down about 60 DB on the fundamental.
What is your experience in this regard? Should I just 'go for it'?
This is not a trick question BTW - I genuinely want to get the view around the table.
If you want low IMD in an open loop design simply design for low THD at
20kHz and low IMD will follow.
Open loop designs are dead easy to get linear at low frequencies however
as frequency rises the job gets tougher. Most people resort to forms of local
feedback such as bootstraps, baxandall pairs, CFP's etc.
If you have managed to avoid the above circuit 'tricks' and still managed to
get good zero FB linearity, the next challenge is OP stage loading.
And there is the OP stage itself which basically swamps everything,
especially into lower loads.
The one open loop amplifier that appears to tread a decent path through all
these obstacles is the Ayre MX series amps. True open loop design, I believe
Charlie H has a strong dislike of -all- forms of FB including CFP's etc etc.
Those Ayre amps measure extremely well considering this. In fact it would
be an interesting experiment to get one of those designs and make a closed
loop design from it, It would probably rival just about any amp in measurements.
And sound a whole lot different. 🙂
Terry
The irony is he's more than happy to employ it inside the DAC chips he chooses (ES9018). Perhaps 'out of sight, out of mind' ?
In 1994, at the height of the SET Fashion Victim Crisis, I built an amplifier with as linear a drive line as I could manage and with type 845 output tubes into 10KOhm loads, single-ended. The commercial stuff was fatally flawed (and out of my price constraints anyway, so Sour Grapes) and I wanted to see what the fuss was all about. I foolishly put the whole stereo amplifier and power supply on a single chassis - no real problem then, but I can't pick it up any more. Don't do the same error!
My long-term on sound quality is heretical; it actually works for me. This needs to be qualified that I only use the very best little 0dBW and down part of the machine. The way to make monotonicity-blameless amplifiers work is through wretched excess. Forget about thermal efficiency. Forget being green. Burn forests, whatever. No free lunch to have both monotonicity and intrinsic linearity.
Next project, with a (younger and stronger and smarter) friend doing the work is triple 211's in parallel and the option of single type 304TL's for experimental purposes. Will be about +14dBW max, but very linear. Multiple chassis. In progress!
Thanks, as always,
Chris
My long-term on sound quality is heretical; it actually works for me. This needs to be qualified that I only use the very best little 0dBW and down part of the machine. The way to make monotonicity-blameless amplifiers work is through wretched excess. Forget about thermal efficiency. Forget being green. Burn forests, whatever. No free lunch to have both monotonicity and intrinsic linearity.
Next project, with a (younger and stronger and smarter) friend doing the work is triple 211's in parallel and the option of single type 304TL's for experimental purposes. Will be about +14dBW max, but very linear. Multiple chassis. In progress!
Thanks, as always,
Chris
Last edited: