Best Home Cinema Processor - Page 18

I don't understand what you're saying here. Can you rephrase or explain?

--Andre

The quasi-logarithmic characteristic of our hearing does come with a cost penalty, in that at elevated sound pressure levels, like at 80 dBSPL and above, substantial amounts of nonlinear distortion are created in the ear. You literally can't hear as clearly at these elevated levels, due to the ear's distortion. The 3rd harmonic distortion (and attendant IMD) that is generated in the ear, tends to augment the apparent loudness of the sound source by spreading the bandwidth. So, if you the turn the volume down to more reasonable levels, like 65-70 dBSPL, not only does the level decrease, but the loudness augmenting distortion decreases, and you would never hear the loudness increase by decreasing the level. Or I should say, not if you are of human origin. I think that's what Roger was getting at.

We're discussing how to address the ill effects of nulls in the listening position. Sounds at certain frequencies are suppressed through acoustic cancellation. You said:
I am saying that since the ratio cannot be altered with EQ, both the direct and reflects sounds will track, and humans will thus perceive the attenuation imposed by the EQ as attenuation (if they hear it at all, being as it is in a null).

If that's the case, the equiloudness curves should change their spacing appreciably at lower levels, but they remain constant for most of the spectrum.



You can probably see the effects of distortion adding loudness for the curves that are crazy loud, but in the range you're talking about, they appear to be fairly constant with respect to SPL.


The reflected sound will have a sustaining reverb that persists after the direct sound is gone in the case of a peak. It's the perception of this ringing we're trying to reduce.

--Andre

I have no trouble reducing peaks with EQ. It's the nulls that elude.

That's why TTD correction is better

I see. Given a known seating location and a know subwoofer(s) location, what is the relationship at the seating position between time and phase?

I don't fully understand the question, but the phase in radians added by a distance L is 2*pi*f*L/V.
Simple correction in the frequency domain doesn't work - you need to counteract delayed signal that causes nulls: take into account signal delay/phase.

Okay,

Much noise and nuisance presented here because I had not posted distortion and noise plots for the AP20. As result I asked the Engineering staff at Datasat Digital Entertainment to send me a current set of plots and they did. The following plots were taken this morning from a production AP20 using a 'certified as accurate' Audio Precision 2700 test set.

Plots for the DDE AP20 Noise & Distortion for both the Analog & Digital outputs

________________
Best Regards,
Carl Huff

A quick point that has escaped some on this forum. It is not the cost of the raw DAC chips that is most important when designing a new product as they have become a commodity. It's how they are used that is most important in determining the fidelity of the finished good.

________________
Best Regards,
Carl Huff

Thanks Carl. Your status and that of Datasat went up good few notches in my book .

Using that data we can then repeat the earlier comparison we did.

Your measurements of THD+N:



TI's bench testing of the same DAC:




Great point and indeed, I made the same earlier that in-circuit performance is very different than bench performance by chip company where lab grade power supply and clock sources are used with no other high-speed signals whaling around.

Summarizing, the spurious distortion products went up from -110 db in TI testing to roughly -90db in AP-20. Assuming the distortion product as the floor of our performance, we are looking at 15 bits in your machine vs 18 bits TI's workbench.

For reference, this is the Benchmark DAC1 again:



Its spurs are at -120 db or a whopping 30 db better than measured AP-20! Man, these guys know how to grind down distortion!

BTW, the vertical scale is squashed so much in your graphs. Assuming you didn't do that to hide the variations, it would be great to have the graphs in proper aspect ratio. I had to really squint to read the -90db number .

For now, it seems that your noise figures are much better than Roger's testing showed. I do not expect a device with -100db noise figure to have the kind of hiss he heard. What was the test fixture there? Just the AP's input load? If so, what load was used? Either way, your potential customers have something to celebrate there .

Also, is the first THD+N Digital Out measurement correct? I don't see the signal that was used to drive the unit. Seems like the input was off in that test even though it says it was at full scale like the analog out above? [EDIT: never mind, I was looking at one graph, and reading the title of another ].

I got a kick out of digital out noise test. Looks like there is dither noise at 24-bit resolution which is what I had suggested it would do in that front . At any rate, it is an uninteresting graph otherwise as in digital domain, we don't worry about noise.

Again, thanks a bunch for the information.

Amirm I respect your posts on here so I'll give you a chance to check your graph reading skills before someone else spots your mistake.

*insert jeopardy music here*

I don't mind being corrected so by all means go ahead. Heaven knows that I have done enough squinting to try to read these graphs, especially the ones provided by Carl. It is someone else's turn now .

I did correct the issue regarding digital THD measurement but I assume you meant something else.

Thanks, Carl.

1) For the analog measurement of THD+N, what output voltage is full scale?

2) Did they provide traditional wideband THD+N measurement? I'm not able to convert a spectral plot to that. (i.e., THD+N less than 0.01%, +4 dBu, 20-20 kHz, unity gain, 20 kHz BW)

3) For analog output noise, the graph is in dBu. That's 0.775v for 0 dB?

4) The analog output noise floor varies between -110 and -130 dB from 100 Hz to 1 kHz with a combing or other harmonically related pattern. I do not see that characteristic in the digital output data. Any idea what's going on there?

5) Do you think that the unit I had was representative of this?

Regards,

Look at the graph you posted, the carrier isn't referenced to 0dB. You have to add the carrier power to the number you used to get the actual SNR. From the posted graph you cannot actually determine the SNR value as the peak is cut off.

Oh yes, the optimism of marketing collateral!

____________
Best Regards,
Carl Huff

After reading my post I have to ask something what do you audio guys call the signal ? I don't know if the term carrier applies outside of RF

I don't see any of them cut off. Which one is that way?

The standard test is to reference 0db. So none of the references I provided have anything chopped off even though it visually looks that way. The AP20 is also fully inclusive although the scale appears to go above 0db (poor work there in not setting the output level as 0db). I cannot tell with the blue graph over white whether it exceeds 0db. Assuming it is at +10db, that brings up their spurs to -100db or still 10 db worse than TI.

PS. Steve, did I get in before the bell? Was this the last round or the daily double?

Further clarifying, there is no "carrier" here. A 0db signal is fed to the machine and the spectrum plotted. What looks like noise probably is. The little spikes are harmonic distortion. You can tell this in how they are multiples of the 1KHz tone fed to the system.


Again, there is no carrier power. You are however right that we are talking about dbFS numbers so the top value matters in that regard. But not due to the notion of carrier.

Are you an RF engineer? Until now, I had not put two and two together relative to your alias . While I did not design RF equipment professionally, I had my share of it from repairing RF equipment while school and building transmitters and receivers as a hobby in 1970s....

The [whole] signal

It is me again . I see how those graphs could easily be mistaken for sidebands of an RF carrier. From a theory point of view, the jitter components actually works on the same principal, causing modulated sidebands. That said, as I clarified, this test and its graph have different meaning although they do encapsulate the effect of jitter.

Yes I am an RF engineer. Regardless of the terminology the signal power wasn't used as a reference (Which was an odd choice for the engineers at Datasat) so the numbers all have to be offset to get the SNR.

Bear with me here as I only know the basics of ADC/DAC design (again from a CMOS point) but the ENOB is related to the quantization noise, NOT the distortion. The even and odd order distortion products will in fact lower the SNR. but the ENOB should be taken from a SFDR.


P.S. Haha I actually started using the HF tag when I was a kid. I was once a audiophile/gearhead. I got into RF engineering during last year of my undergrad and continued with it into my phd. I am also big into R/C helis and help mod a site called Helifreak so again the HF works

Straight from the Engineer that made the plots ...

#1) 3.0v is full scale
#2) 22kHz filter on THD swept
#3) .775v ref on dBu
#4) Guessing random analog noise on the plot, digital is going to be less obviously.
#5) Nope. The AP20 that you had was a 2nd hand demo box with a very early serial number.

______________
Best Regards,
Carl Huff

Carl, is the analog noise measurement with balanced out or unbalanced?

Also, is #1 RMS or peak to peak?

I believe it's unbalanced as that is what I was most interested in measuring.

_____________
Best Regards,
Carl Huff

Clarifying, you only have that issue with AP-20 numbers, right?


Correct. I have noted this before and was careful in how I couched the my computation: "Assuming the distortion product as the floor of our performance."

My goal there wasn't absolutely correct but rather, try to translate obscure db numbers into *relative* bit fidelity that people can better associate with. I find that if I go to the level you are, I will lose everyone in the discussion . Our resident DAC expert elsewhere also makes a point of keeping me honest on this from time to time : http://www.whatsbestforum.com/showth...9-Sampling-101


I loved RF when I was growing up. If I had not found computers in 1970s, I would have probably gone there. Ironically, with our computers running in GHz frequencies now, some amount of RF knowledge is useful in that field also.

Peaks are also caused by time delayed reflections.


I would guess analog electronics adding noise and distortion and implementation differences. You can see mainly an odd order series, whereas the TI datasheet shows the presence of even order distortion as well. The measurements' resolution (ie. the FFT bin widths) are really too wide to say anything more about the distortion. I don't think you can say anything about jitter from these measurements.

--Andre

RMS

_____________
Best Regards,
Carl Huff