The two boards were mounted on the same heat sink, probably 50 mm apart. The wiring harness to the power supply (Power-86) was about 15-20 cm long per channel. The wires zip tied together. This was a simple prototype setup and I did nothing to optimize any wiring routes.
Tom
Tom
cool. Was wondering if you had done anything to minimise interaction between the output inductors, as you are at the level where that might actually matter. Good to know its not a problem.
No, quite the contrary, actually. The inductors are lined up. They are probably a good 10 cm apart, if not a tad more... Two air core inductors that far apart don't couple that tightly. It also helps that any inductive coupling will be shorted to ground by either the output impedance of the amplifier or by the speaker impedance.
Tom
Tom
Anyone in the UK have one, I'm thinking maybe a bridged version could be the ticket. I've been considering an F5 turbo as well.
Interested in any customer feedback.
Interested in any customer feedback.
Bob Cordell interview (2005)
Allow me this digression, but I think that old interview (2005) will be much interested:
The Amplifier Guru speaks: Bob Cordell
-> Pjay Speaker gurus Bob Cordell
-> -> CordellAudio.com - Home
Allow me this digression, but I think that old interview (2005) will be much interested:
The Amplifier Guru speaks: Bob Cordell
-> Pjay Speaker gurus Bob Cordell
-> -> CordellAudio.com - Home
My question would be what difference you actually hear between 0.0004% THD+N, 0.04%, and 1%. Is it just piece of mind or a clearly describable difference?
I suspect it depends on the type of distortion. As far as I've read, the perception threshold for THD is somewhere in the 0.1-1 % range. A friend of mine can reliably hear THD in that range. I suspect that as long as the THD is below 0.1 %, the vast majority (refraining from saying "all" here) of humans will not be able to detect it.
However, the THD figure is useful as a figure of merit. If you want clean sound reproduction, the amplifier needs to have high loop gain at 20 kHz. I find that I want a loop gain of at least 45-50 dB for the highs to sound clean to me. That's also the figure Bruno Putzeys has arrived at. As the THD is attenuated by the loop gain, one can deduce that an amp with high loop gain will have low THD. So while THD <0.1 % may be imperceptible, building an amp with 0.000x % THD makes sense as it makes it more likely for that amp to have high loop gain at 20 kHz, hence sound good.
Tom
I would expect that different types of distortion will have different thresholds of audibility.
0.1% of 2nd harmonic maybe be inaudible to 90% of informed listeners.
But 0.1% of crossover distortion maybe audible to 90% of informed listeners.
The 0.1% "standard" came from the period before Solid State became readily available.
Much of the Tube/Valve equipment from that era was ClassA, or high bias ClasssAB and crossover distortion was much less of a problem.
0.1% of 2nd harmonic maybe be inaudible to 90% of informed listeners.
But 0.1% of crossover distortion maybe audible to 90% of informed listeners.
The 0.1% "standard" came from the period before Solid State became readily available.
Much of the Tube/Valve equipment from that era was ClassA, or high bias ClasssAB and crossover distortion was much less of a problem.
Distortion metrics are not just important because distortion is audible at varying levels by varying individuals. Below a certain point, nobody will be able to detect
distortion. This point in all likelihood is something like 0.1% THD, of course, as Andrew points out, depending on the kind of distortion. Crossover, or the 7th, are really annoying at levels that measure even lower than this.
So why then should we want ever lower distortion in amplifiers, and what is the purpose of going from 0.015 to 0.0015%? In other words, what is wrong with a naked LM3886, and why spend money on nesting it within the loop of a high gain opamp?
The reason why striving for the lowest distortion figures is worthwhile is because this distortion figure stands for other performance as well, eg damping factor. Lots of feedback is what is needed for a low damping factor, and vanishingly low distortion figures show that gobs of it are available. But even more important is PSRR. Low distortion for the same reason goes together with high immunity to power supply fluctuations.
distortion. This point in all likelihood is something like 0.1% THD, of course, as Andrew points out, depending on the kind of distortion. Crossover, or the 7th, are really annoying at levels that measure even lower than this.
So why then should we want ever lower distortion in amplifiers, and what is the purpose of going from 0.015 to 0.0015%? In other words, what is wrong with a naked LM3886, and why spend money on nesting it within the loop of a high gain opamp?
The reason why striving for the lowest distortion figures is worthwhile is because this distortion figure stands for other performance as well, eg damping factor. Lots of feedback is what is needed for a low damping factor, and vanishingly low distortion figures show that gobs of it are available. But even more important is PSRR. Low distortion for the same reason goes together with high immunity to power supply fluctuations.
Member
Joined 2006
Not sure if having this merit of very low distortion is responsible for the good sounding of mod86... its more likely due to its topology....wondered if Tom could hear at all a difference from a badly tuned mod86（say with a 100 times more distortion) to one of his nicely built ones...
BTW, speakers likely to have a double digit distortion figure right? So listening to a system as a whole...
I am not quite sensitive to distortion figures these days...maybe childhood horror TIM stories still lingering...😀
BTW, speakers likely to have a double digit distortion figure right? So listening to a system as a whole...
I am not quite sensitive to distortion figures these days...maybe childhood horror TIM stories still lingering...😀
Last edited:
An oddity of the audio world is distortion figures for electronics are typically stated under best case circumstances whilst those quoted for drivers or speakers are typically for something resembling worst case. For the most part, I've been unable to measure the THD of the various drivers I've worked with over the years for home audio until the system's operating at levels someplace in the deafening to painful range. From a naïve perspective the limiting factor is more the nonlinearity of air itself. If memory serves that's something like 0.07%. The human ear's also pretty nonlinear---around 3% if I'm remembering correctly, which I may not be---which seems to be compensated for, or at least masked, by perceptual processes.
Another limitation of acoustic measurements is the ambient noise floor is most always 30+dB SPL, even if one's got appliances shut down and is measuring in the wee hours of a windless night. So if the system's operating at a just coming up loud 60dB SPL at the listening point then THD+N is necessarily 0.1%. In practice THD+N is often 10-20dB higher what with traffic, refrigerators, forced air heating, lower SPLs common in shared living arrangements like apartments or flats, and whatnot. This works, in part, because of the brain's ability to hear things 15-20dB under the ambient noise floor.
The Mod has 47dB loop gain at 20kHz (post 734). Reduce that by 40dB and there'll be a subjectively audible difference in driver control. It's an open question how much loop gain is required for subjective inaudibility. Bruno Putzey's proposed 60dB as a design target, which is pretty reasonable if potentially somewhat conservative. I've found no discrimination in ABX trials up around 150dB loop gain at 20kHz, which is thoroughly unsurprising, and have a build going that'll come in around 120dB, which is likely to be equally uninformative.
Another limitation of acoustic measurements is the ambient noise floor is most always 30+dB SPL, even if one's got appliances shut down and is measuring in the wee hours of a windless night. So if the system's operating at a just coming up loud 60dB SPL at the listening point then THD+N is necessarily 0.1%. In practice THD+N is often 10-20dB higher what with traffic, refrigerators, forced air heating, lower SPLs common in shared living arrangements like apartments or flats, and whatnot. This works, in part, because of the brain's ability to hear things 15-20dB under the ambient noise floor.
The Mod has 47dB loop gain at 20kHz (post 734). Reduce that by 40dB and there'll be a subjectively audible difference in driver control. It's an open question how much loop gain is required for subjective inaudibility. Bruno Putzey's proposed 60dB as a design target, which is pretty reasonable if potentially somewhat conservative. I've found no discrimination in ABX trials up around 150dB loop gain at 20kHz, which is thoroughly unsurprising, and have a build going that'll come in around 120dB, which is likely to be equally uninformative.
I have a lot of admiration for the work Zaph put into measuring distortion in speakers and I know some have argued about them but they appear to be good for DIY use even if not up to JBL R&D standards.
From my perspective I am not worried if the Mod-86 is not the last word in performance and other amps sound better, neither am I worried if I couldn't ABX it from a standard ebay gainclone in my system. But I will know it is the best my budget could stretch to at the moment. Brains are funny things, but that will matter when I sit down to listen 🙂
From my perspective I am not worried if the Mod-86 is not the last word in performance and other amps sound better, neither am I worried if I couldn't ABX it from a standard ebay gainclone in my system. But I will know it is the best my budget could stretch to at the moment. Brains are funny things, but that will matter when I sit down to listen 🙂
Why would an amp with 0.00014% have a sound?
So, if it does not, then how could another amp "sound better"? The other amp would have to be distorting would it not?
So, if it does not, then how could another amp "sound better"? The other amp would have to be distorting would it not?
Well my numbers look ok, but I am borderline on current capability with my 3ohm ribbons. If you run the numbers a la cordell I am suboptimal on current. It's possible at least in a larger room than I am currently in that I could run out of headroom on peaks. Unlikely in current listening room though. And yes serves me right for not waiting for the parallel86.
But this is an edge case on an edge case 🙂.
But this is an edge case on an edge case 🙂.
Is higher damping in the higher audible range really that important for reduced audible distortion? Once you get to a reasonably high damping factor, the main limiting factor is in cabling and voice coil resistance.
This can be tested easily enough with a small resistance in series with the tweeter. I've seen schematics of "good" speaker crossovers with just a series resistor to lower tweeter sensitivity to match other drivers (not even an L pad), so I'm doubtful high damping factor in that frequency range is critical.
But maybe there's more to higher PSRR. Regardless, these things can certainly be measured, er, listened to independently, by powering a "mediocre" amp with a "good" bench power supply, and lowering damping factor of a "good" amp by adding series resistance.
Yeas, but as just mentioned in recent posts, despite higher THD numbers speaker distortion is almost all from the lower and less-audible harmonics. Class AB solid-state amps generally have the "full" spectrum of all harmonics of audible frequencies, with lower levels at higher harmonic number, but still higher than speakers at higher harmonics. Thus you can easily hear "amp distortion" of 1 percent crossover distortion (which is "wide-band") through a speaker that gives 10 percent second harmonic distortion.
And it's only because common solid-state amps give such similar distortion patterns that THD has any validity at all in comparing different amps.
This thread describes an attempt to replace THD with a distortion metric that better relates to perception:
http://www.diyaudio.com/forums/multi-way/121253-geddes-distortion-perception.html
An amp's sound is determined with music playing (crest factor of at least 12dB for the most compressed grunge, 20dB or more for classical) not with a sinewave (3dB crest factor). The statistical properties of the waveform being reproduced do matter because a classAB amp's most crucial characteristic for its sound (its PSRR) varies with the instantaneous output voltage level.
