for a "multi-path" DAC - but will it really do 162 dB N when level is, say -44 dBfs?
I say: no.
I call bogus
https://audioxpress.com/news/imersi...analog-converter-completes-32-bit-audio-chain
You?
PS: Practically this is of course the same as just shutting off the output signal at say -110 dBfs... if you have control of the current noise in the output stage...
//
I say: no.
I call bogus
https://audioxpress.com/news/imersi...analog-converter-completes-32-bit-audio-chain
You?
PS: Practically this is of course the same as just shutting off the output signal at say -110 dBfs... if you have control of the current noise in the output stage...
//
Even if that did exist, who could possibly put all of that capability to use in a living room setting? Sounds like a drop forge was ordered when all that was needed was a tack hammer.for a "multi-path" DAC - but will it really do 162 dB N when level is, say -44 dBfs?
I say: no.
I call bogus
https://audioxpress.com/news/imersi...analog-converter-completes-32-bit-audio-chain
You?
PS: Practically this is of course the same as just shutting off the output signal at say -110 dBfs... if you have control of the current noise in the output stage...
//
I note that Jan missed a bit of proof reading there "There are already in the market microphone, preamplifiers and A/D converters that use a multi-path architecture to achieve a dynamic range of 1260dB".
Given there is a jump in the noise floor when it 'range switches' not sure how useful this could even be outside of intstrumentation, where if I read correctly this technique has been used for many decades and I notice no mention of linearity.
Given there is a jump in the noise floor when it 'range switches' not sure how useful this could even be outside of intstrumentation, where if I read correctly this technique has been used for many decades and I notice no mention of linearity.
Cheap and fast DSPs maybe. But CPUs generally have 64 bit registers now, and FLOAT64 - and most will do at least 128bit vectors with two 64 bit floats.
Even a slow passively cooled laptop CPU is freakin fast for audio processing unless you have a stupidly large number of channels.
Suppose you want to use digital volume control and that you have a two times 100 W amplifier, 90 dB at 1 W, 1 m loudspeakers, and a reverberation radius of 1 m. The highest level you can reproduce is then about 113 dB SPL. The threshold of hearing of an average young person with good hearing is about 0 dB SPL at 1 kHz. Assuming the same holds for A-weighted noise, and that you want the noise to be inaudible even in perfectly quiet surroundings, a DAC dynamic range of about 113 dB(A) would be needed, possibly a few dB more if your ears are better than average.
I think this can be achieved with a single DAC nowadays, without gain switching tricks. Gain switching multipath DACs could still be useful if you can't trim the analogue gain such that full scale drives the power amplifier to its maximum, or if someone has enormously sensitive horn loudspeakers or whatever.
It's all quite similar to https://www.diyaudio.com/community/threads/fixed-gain-field-recorder.373352/post-6685215 , but in the opposite direction.
Nevertheless, my 85 dB(A)...93 dB(A) dynamic range DIY DACs are good enough for me, but then again, I use an analogue volume control.
I think this can be achieved with a single DAC nowadays, without gain switching tricks. Gain switching multipath DACs could still be useful if you can't trim the analogue gain such that full scale drives the power amplifier to its maximum, or if someone has enormously sensitive horn loudspeakers or whatever.
It's all quite similar to https://www.diyaudio.com/community/threads/fixed-gain-field-recorder.373352/post-6685215 , but in the opposite direction.
Nevertheless, my 85 dB(A)...93 dB(A) dynamic range DIY DACs are good enough for me, but then again, I use an analogue volume control.
Not missed anything here Bill, this is a company press release and they can say what they like, we don't edit those. But an AX review is coming which will highlight 'the challenges' if I may say so. What I wonder is how they do all that amplifying and attenuating and switched combining while keeping the linearity to the LSB. What do you think about that issue?
Jan
I understand the direct correlation, but is there a case for 10x precision measurements to increase accuracy? (ignoring the noise etc)
In telescope positioning using optics, they will measure 10x the resolution or higher to ensure repeatability of positioning. Would it not be beneficial to measure higher audio resolution to ensure repeatability of the measurement?
That's an important point I thought Ed. In the press release (and the AES e-Brief) they gloss over that and just mention '-170dB dynamic range and linearity'. That's the only place they mention it.
Jan
To expand upon your points, a whisper can be 15dB and the rustling of leaves 30db. The detection assumes the absence of 113dB SPL of background. It seems likely that this was the argument Philips made to Sony when Sony nevertheless insisted upon the 16bit standard. The arguments relied upon the most advantageous "windowing" of signals being within the minimum and maximum un-digitizing capability in reproduction by CD players. This changed when computer based digital attenuation became commonplace as a cost effective means to controlling gain.
There seems a general argument to suggest that detection of artifacts is possible below the threshold of human hearing if in the presence of high levels of signals, signals that can't otherwise be heard in a dead silent anechoic environment. This seems too far fetched to even imagine as convincing, as pain inducing high level signals are generally understood to mask such detection, not to mention impacting on the pleasure of such detection even if true.
It is considered that there exists some ideal window that all DAC's operate at their best (however good or poorly that might be). The point being, why engage in dynamic broadband switching in some "multi-path digital to analog converter" when some perhaps fixed R2R ladder divider positioned after the converter can window a TDA1541a into its ideal performance band?
This DAC is possibly a great innovation on the technical side, but at the same time is a bit of a head-scratcher for me. 
Taking the aX marketing piece at face value, the next question I have is: please show me an amplifier that has a dynamic range anywhere close to 160 something dB... or a noise floor of -140dBu.
Also, what source (e.g. recording) will have anywhere near this sort of S/N ratio? A quick Google search shows the dynamic range of human hearing is only 120dB. So...
I get that the company is trying to cover the full 32-bit signal range, but what's the point of this sort of figure when the rest of the chain including our own ability to hear is lagging quite far behind?
Taking the aX marketing piece at face value, the next question I have is: please show me an amplifier that has a dynamic range anywhere close to 160 something dB... or a noise floor of -140dBu.
Also, what source (e.g. recording) will have anywhere near this sort of S/N ratio? A quick Google search shows the dynamic range of human hearing is only 120dB. So...
I get that the company is trying to cover the full 32-bit signal range, but what's the point of this sort of figure when the rest of the chain including our own ability to hear is lagging quite far behind?
My guess is that they are not trying to switch between the "High DAC" and "Low DAC" very rapidly because switching would produce monotonicity errors. They need to make that unnoticeable which amounts to "rarely occurs in practice".
Ed
I don't quite understand what you mean. According to Wikipedia, the hearing threshold for very short tone bursts is considerably higher than for somewhat longer tone bursts, so it seems that the human auditory system is already doing some sort of averaging to improve the sensitivity.
By the way, the noise of these high-performance DACs is often largely plain old analogue circuit noise. Reducing quantization noise to negligible levels is usually easier and cheaper in chip area and current than doing that with analogue noise.