this from R.Cordell
post211
http://www.diyaudio.com/forums/equi...on-audio-range-oscillator-11.html#post3119369
The THD plots (and other measurements) are also shown on my website:
MOD86 Rev. 2.0: Modulus-86 Rev. 2.0: Composite amplifier achieving 0.000067 % THD.
PAR86 Rev. 1.0: Parallel-86: High-power composite amplifier achieving 0.00012 % THD.
MOD86 Rev. 1.0: Modulus-86: Composite amplifier achieving 0.00018 % THD.
I also do my best to update Post #1 in the Modulus-86 thread to include links to the relevant posts in the thread. http://www.diyaudio.com/forums/vend...mposite-amplifier-achieving-0-0004-thd-n.html
Tom
MOD86 Rev. 2.0: Modulus-86 Rev. 2.0: Composite amplifier achieving 0.000067 % THD.
PAR86 Rev. 1.0: Parallel-86: High-power composite amplifier achieving 0.00012 % THD.
MOD86 Rev. 1.0: Modulus-86: Composite amplifier achieving 0.00018 % THD.
I also do my best to update Post #1 in the Modulus-86 thread to include links to the relevant posts in the thread. http://www.diyaudio.com/forums/vend...mposite-amplifier-achieving-0-0004-thd-n.html
Tom
Last edited:
For sure 20khz fundamentals wouldn't be useful; however, that probably wasn't the point. Of much greater concern is that a lot of lower pitched notes may have 20khz harmonics, which are useful.
You know even the bass guitar needs room up to 11khz for its harmonics, so may I suppose that a violin could need some harmonic support at 20khz or maybe even slightly higher? Well, that's what I think of it. You ever bought violin strings to match a particular instrument? Same topic, and it is laborious as in the utterly tight specs.
I had a memory of having to correct a slightly baleful sounding audio card and had measured a very, very slight treble droop between 16khz to 20khz. Supposedly, I couldn't hear that. Well, I think that RMAA may have under-reported the severity considerably. Anyway, I managed to fix it so it reported a flat response, and certainly could hear the difference. What was reported is 0.03db, or maybe it was 0.3db--I just can't remember and what actually mattered is that when it reported 0.00db treble problems, that sounded pretty good.
The electronic basset hound simulation was evicted for a couple hours labor and parts cost of 12 cents. The newsworthy item probably isn't that the measuring means was off by at least a factor of 10x but the rather good news that measuring had pinpointed the area to fix on the very first try, which had succeeded post haste. That was great!
Apparently, the only thing that I know in this case is that you probably ought to measure the 20khz down to 0.00* decibels, just in case.
That probably wasn't the last word on the topic, but it is the one that I have.
P.S.
These observations were made with a "prescription" speaker (done similar to a hearing and/or eye exam). Otherwise, my age might have had the effect that you mention OR perhaps the sheer opposite effect that any tolerances got several orders of magnitude worse/tighter if/when a poorer matching speaker were used. So, I just had to mention that I had apparently become a little lost along the way and therefore my suggestion of 0.00* decibels tolerance for 20khz, may be inaccurate by an unannounced amount.
Due to a few things that you didn't mention, the Parallel86 seems to be remarkably more cost effective.
An excellent upgrade for discrete amplifiers is a more linear output device, and a really similar thing happens when you parallel the chip amplifiers. It is a huge upgrade when driving a vast variety of real speakers.
Last edited:
Could you elaborate more please? I am interested in why you think the parallel is more cost effective.
Yes, albeit not exactly. If it's too high to be directly audible and ABX (or similar) testing indicates a robust subjective difference then it follows audibility must be through some other mechanism. Measurements such as 18+19kHz IMD are much more direct than THD+N about looking at downmod products around the upper limit of audibility (13-17kHz for many individuals) and possibly extending into lower frequencies.
IMD also has the advantage of extending to the harmonic floor or noise floor, whichever comes first, meaning it's not subject to the bandwidth versus harmonic capture tradeoff of THD+N. For many amps that doesn't matter but in well executed composite designs such as the Mod typically what you get back from THD+N is the noise floor of the measurement equipment integrated over the measurement bandwidth. So the more spectrally compact the measurement is the greater its ability to resolve something related to the amp rather than the analyzer.
The limit of audibility for broadband level changes is usually taken to be 0.1dB. So direct audibility of 0.3dB is plausible.
If you were hearing 0.03dB in the highs and not a different issue which happened to be addressed by the changes for flat SPL I would be surprised. One place to look in particular is phase shift in the range of 3-4kHz; some care is needed in designing filters cornering in the low ultrasonic to avoid audible loss of phase coherence lower down. I've found 10 degrees at 3kHz to be just under the limit of audibility in ABX tests, so try to design the complete DAC+DAC buffer+line driver+line receiver+power stage chain to stay under 5 degrees. This is not trivial, particularly if LC output filters on class D amps are included.
if one applies a low pass single pole passive (input) filter and this becomes the dominant pass-band limitation, then what F-3dB frequency corresponds to your 5degrees of phase guidance?
Thus a frequency factor of ~11 gets the filter phase from 45degrees to ~5degrees.
Thanks.
This is a massively lower frequency for the RF attenuation than many claim is needed to have no audible effect on the audio band.
Some, myself included use around 150kHz to 200kHz and a few insist that it has to be around 1MHz
Thanks.
This is a massively lower frequency for the RF attenuation than many claim is needed to have no audible effect on the audio band.
Some, myself included use around 150kHz to 200kHz and a few insist that it has to be around 1MHz
Last edited:
I have no DBLT or ABX Data to proof, but I guess(!) the other effects that such a LPF will introduce are more audible. I personally would put the corner frequency somewhere near 200kHz as well. However this is purely hypothetical, a real world application would certainly use multi pole filters in DAC outputs and class-D amp outputs.
Ben
The ear/brain system does not decode phase for frequencies higher than 3500 Hz or so. About the wavelength of your ear to ear distance. You can phase shift all you want above this frequency, it just doesn't register.
But even below that, it is doubtfull if people can actually pick up on a 360 degree phase shift around 2KHz, as Billshurv argues. DBT also on this site was inconclusive.
But even below that, it is doubtfull if people can actually pick up on a 360 degree phase shift around 2KHz, as Billshurv argues. DBT also on this site was inconclusive.
I like something similar. . . very much like a "full range" (wideband) assisted by both a "helper woofer" and a "helper tweeter" like in these good examples:
For the Parallel86 Statements
For the Modulous86 Finalists
The general idea of full range with "helper woofer" and "helper tweeter" is also one of my goals in designing a new driver. The 3" I currently use seems a bit shy in the lower frequencies. Looking at 80~8Khz with a 6" metal cone. But there needs to be some compensation somewhere in the system to get a linear phase I am looking for as well as a whole set of other considerations.
Some weeks ago I read about a study that tried to define wether phase distortion was audible. They used an all-pass to turn the phase at different center frequencies. Test results indicated that phase distortion is audible only with headphones and only at low frequencies up to maybe 2kHz. But the results were barely statistically relevant. So imho it is sufficient to keep phase change below maybe 20 degree in the 100-2000Hz band. However, that only seems relevant in line level equipment and headphone amps, not in power amps.