Its a good question. The answer is very simple - transparency is subjectively determined not by measurements.
ABX tests are for determining if an audible difference exists between A & B - how is that relevant to determining transparency?
I've already mentioned the work of Bob Stuart in determining how many bits and what sample rate is necessary. So far you've not rebutted that so as far as I'm concerned it still stands as the design parameters for an ultimate DAC. You have though mentioned issues related to marketing - you don't think such a specification would be widely embraced. Care to explain how that matters when the aim is the ultimate DAC?
Its your thread after all and you're free to clarify that you only wanted a marketable 'ultimate' rather than an ultimate 'ultimate'.
ok... how about having your cake and eating it too...
the best of both worlds... we design our setup to handle all sampling rates upto the highest - but we provide an option to sample at a lower rate should the user desire and we will set the default as 96kHz as optimum and recommended setting and the user will be free to choose a higher or lower sampling capability...and let his ears decide which sounds best to him.
the best of both worlds... we design our setup to handle all sampling rates upto the highest - but we provide an option to sample at a lower rate should the user desire and we will set the default as 96kHz as optimum and recommended setting and the user will be free to choose a higher or lower sampling capability...and let his ears decide which sounds best to him.
Last edited:
By 'handle' do you mean 'accept'? That's no problem - does he have to know how we accept his higher sample rate?
The real issue here is whether when accepting higher sample rates, the DAC is being asked to reproduce signals beyond the audio band (i.e. > 20kHz). If so, then no, the DAC cannot be both 'ultimate' and also produce ultrasonic frequencies.
The real issue here is whether when accepting higher sample rates, the DAC is being asked to reproduce signals beyond the audio band (i.e. > 20kHz). If so, then no, the DAC cannot be both 'ultimate' and also produce ultrasonic frequencies.
Since absence of noise modulation was emphasized as an important objective by ESS guys, this graph looks a little strange to me... do you think it comes from the chip or from the implementation you measured ?
(also, you can't see all forms of noise modulation on a FFT)
I didn't make the measurement, a guy over on Head-Fi made it. It correlates well with the published measurements of the Weiss Medea+ DAC which uses the same chip so I'd say its a chip issue, not an implementation one. Over on that thread they are saying that ESS know about this and its not a fixable thing. You can go read for yourself - Ranking of 17 DACs and DAC Configurations - Page 10
I agree with your latter point.
I agree with your latter point.
Couldn't we just limit the frequency response to the sonic range (say 15Hz to 25kHz) using filters... even if the input had ultrasonics or the DAC generated it somehow?
I was suggesting that the sample rate at which the input is coded is the same rate at which the DAC will be asked to process it...
if input has been sampled at higher than 96kHz - by default it will be down-sampled to 96kHZ and then processed... but the user may select the option to use the higher sampling rate and the DAC and circuitry itself shall process it at the higher sampling rate of the input...
meaning down-sampling will be optional...and selectable by the user.
Would it be necessary to include some DSP circuity to upscale 44kHz to 96kHZ etc? needless complexity?...
I was suggesting that the sample rate at which the input is coded is the same rate at which the DAC will be asked to process it...
if input has been sampled at higher than 96kHz - by default it will be down-sampled to 96kHZ and then processed... but the user may select the option to use the higher sampling rate and the DAC and circuitry itself shall process it at the higher sampling rate of the input...
meaning down-sampling will be optional...and selectable by the user.
Would it be necessary to include some DSP circuity to upscale 44kHz to 96kHZ etc? needless complexity?...
As far as I'm concerned the audible range is 20-20kHz and yes in order to get ultimate performance there will need to be an analogue (passive) filter at the output. Hence it won't be physically possible for the DAC to produce frequencies much above 20kHz even if the chip behind the filter was running faster - they'd get rejected by the filter. The filter spec is one to be determined in the design process by listening - it needs to have decent phase response to 20kHz but be aggressive at curtailing out-of-band energy as OOB stuff is a primary reason for poor SQ.
I'd not try to upsample 44k1 to a higher rate, no. <edit> Having said that, the decision about this depends on the choice of DAC chip. It may well be that to achieve the necessary resolution, oversampling (and maybe even some noise shaping) will be required.
I'd not try to upsample 44k1 to a higher rate, no. <edit> Having said that, the decision about this depends on the choice of DAC chip. It may well be that to achieve the necessary resolution, oversampling (and maybe even some noise shaping) will be required.
Last edited:
Some sub-woofers claim to go all the way down to 10 or 14Hz - hence I set the lower bound to 15 instead of 20Hz and some tweeters like B&W Diamond.. claim to go well beyond 20kHz (and claim that it actually makes things sound richer and fuller) - that's why I gave the upper bound to 25kHz.
You haven't commented on making the optional ability to down-sample to 96kHz but if the user wants to process 192kHz or 384kHz at the native rate then he should be able to...
You haven't commented on making the optional ability to down-sample to 96kHz but if the user wants to process 192kHz or 384kHz at the native rate then he should be able to...
Last edited:
Yes of course - they're able to process 384kHz at that rate, just not with the ultimate DAC. Sometimes one has to be a bit of a Steve Jobs with customers and don't let them break a good thing by tying their hands somewhat. If they're a customer then they'll have already been educated that higher sample rates aren't an improvement beyond 96kHz. And if they disagree with that premise, ergo they're not a customer.
Thanks for the links, interesting read ! I have a simple 9018 test board in the works, so I'll try to measure this stuff too and see what I get. Probably within 2 months, I'll report back.
If it is true, well, bummer.
Could be bad sample. I would temper these results against others, such as Weiss Medea D/A processor Measurements | Stereophile.com.
No comparable measurements that I see there to temper the FFT against. However over here:
Weiss Medea+ Measurements
pay attention to the 2nd plot down which is THD+N shown against rising output level. The line is almost perfectly straight from zero output (left side of plot) up to a level of -40dBfs. After that, strange things begin to happen.
@peufeu - I shall be very interested indeed to see any results you care to show for your DAC with signal levels from -40dBfs and up.
Weiss Medea+ Measurements
pay attention to the 2nd plot down which is THD+N shown against rising output level. The line is almost perfectly straight from zero output (left side of plot) up to a level of -40dBfs. After that, strange things begin to happen.
@peufeu - I shall be very interested indeed to see any results you care to show for your DAC with signal levels from -40dBfs and up.
Last edited:
OK, it's visible... but the gremlins are around -115 to -125 dBFS, can someone hear that, I don't really know, unless the system gain structure is screwed up and the DAC is run ar a huge digital attenuation most of the time...
The average level is down around those figures (for THD+N there's low pass filtering of the residual to get to the graph point), but this kind of modulation noise is impulsive, rather like rub and buzz faults in drivers. Our hearing is very sensitive to such aberrations - they can have very low average (RMS) levels, but still be audible because of high crest factors.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.