One Q please:
Can someone comment the usage of
Riv connected to
Iout-s and virtual ground of +2,5V
?
Current outputs are in the constant DC potential (1/2 VrefH of +5V) of +2.5V
So it would be no DC current trough RIV if the Vref is also +2.5V
Only AC Iout (mAp-p)
.
(After RIV some Cout can be used for prevent this +2.5V DC...)
Can someone comment the usage of
Riv connected to
Iout-s and virtual ground of +2,5V
?
Current outputs are in the constant DC potential (1/2 VrefH of +5V) of +2.5V
So it would be no DC current trough RIV if the Vref is also +2.5V
Only AC Iout (mAp-p)
.
(After RIV some Cout can be used for prevent this +2.5V DC...)
Correlated noise is seen in FFT test tone skirts. Really what is seen is averaged correlated-noise over the acquisition time of the FFT data (and over the time of any averaged FFTs, which is sometimes used to reduce the apparent look of noise in the display by boxcar LP filtering it).
Also, in general noise crest factor is also averaged out in FFT appearance because noise frequencies are not always in the same FFT bin. The energy of noise is spread out and made to look small in many bins. In the time domain it didn't necessarily have all its energy spread out. It could have a very high crest factor caused by temporal alignment of the phase of multiple frequencies in the time domain.
Why is crosstalk of correlated-noise not crosstalk? Because its too hard to measaure?
Have you designed any PCBs with AK4499EX+AK4191?
IME not so. E.g. AK4493 has too much variance between DACs on chip so paralleling them does not lead to benefits.
That is a false assumption. There is no need for averaging and peak values can also be shown. So even non-correlated noise can be seen.
You should ask that from yourself as I have not seen you measure anything.
Its not an assumption. Its a fact as a Fourier transform in an integral transform over all time. An FFT is an approximation to a continuous time FT. It is effectively integrated over the acquisition time. And an integrator is a LP filter, which is an averaging function. What you get with an FFT is the average correlation with bin frequencies over the length of the dataset (which took time to acquire at some sample rate).
But that source signal for dual mono amps can come also from dual mono DAc unit?
Imagine that You have digital galvanic isolation from pure digital part and also galvanic isolation from pure Analog part. The DAC is actually not an analog or dital it is AD module.
.
If You take an integral over complicated PCB probably yes.
But if You emphasize the thermal issue maybe not. So the main heat generateor is under the chip where wea have even exposed heatsink metal area. Imagine the squared hole in PCB just under the chip and small heatsink directly bonded with main heat-emmited area. This coupling heatsink can lead to bigger one just under the PCB. (Top surface is not enough to maintain the generated heat - but helpful.)
Next, we have to place just few components close around the chip mostly decoupling capacitors.
That is all that is from vital importance... And we can extend durability, minimize the probability of failure and all other issues implicated directly by temperature. The great probability is that sound will change a lot, with better measurements too?
Not with these chips. But I am thinking of AK4499EX in hardware mode. Dual mono.
AK4499EX is interesting to me because it can be operational without programming registers additional hardware (and sofware). It is available and I am curious to try diy some AK dac too. 🙂
But have some unclear issues about current outputs. I would like to use passive Riv, coupling C and passive analog filter.
(Certainly I don't want to have a standard OP circuit, million times repeated... 🙂 )
.
For instance Can I use passive RIV conected to Iout and some Vref=+2.5V
Maybe You can help?
Last edited:
Non-correlated noise shows up as a noise floor. It can look like low level white noise in an FFT and still sound like popping and frying in the time domain. So what doesn't show up in an FFT is noise floor audibility, which is different for a little low level hiss, as opposed to higher crest factor and thus louder popping and or frying (those differences are encoded in noise frequency component relative phases). Understood in whichever case above total noise floor energy/power can be the same; its that louder events noise must occur less often than less loud events, at least on average. Whether or not loud noise events are likely to occur depends on the physical process that created the noise in the first place.
Last edited:
No, it depends on the uncorrelated noise. White noise would show up as raised but flat noise floor. But 1/f noise or popcorn noise have different shape in FFT.
See: https://www.iee.et.tu-dresden.de/iee/eb/forsch/AK-Bipo/2014/AKB2014_02_SisConsult_Sischka_v2.pdf
Not necessarily true. Please see: https://www.edge.org/response-detail/11715 Please see the graph and read the text.
All white noise in the pic below (notice the differences in scales):
"Whiteness just means that the noise spikes or hisses and pops are independent in time or that they do not correlate with one another."
Last edited:
What has that got to do with FFT?
Or with audibility?
Why don't you show us graphs of uncorrelated noise that is audible but does not show up in FFT.
Or with audibility?
Why don't you show us graphs of uncorrelated noise that is audible but does not show up in FFT.
I always have one galvanic isolation - @I2S between USB-interface and DAC and all other
Because the statement was about channel separation.
I encountered problems with heat only with the EC9038Pro, in stereo mode, where each channel has 4 paralleled DACs and a large total output current. But even there, the OPA1656 worked perfectly, with 250 Ohm in the feedback.
But I talked about the number of layers for another reason - on a 4-layer PCB is much easier than on a 2-layer to correctly and efficiently make power lines layout to all channels, even when there are not 2 but 8 of them.
Better measurement in not always mean better sound (unless you are from the Sect of Saint Amir 🙂), but w/o paralleling measurements are good enough.
Alex.
It shows that an FFT displaying white noise can have all sorts of different waveshapes in the time domain, which for example can sound like "hisses and pops." It can be popping and frying (a type of hissing) in the time domain. To be white just means the present state of the noise is independent from its prior state. OTOH, if the present state of the noise is at least in part a function of its prior state, then colored noise (non-flat frequency response noise), can be seen in an FFT.
The noise will show up in an FFT, what doesn't show up in the FFT is the audibility of the noise. Often people will assume noise floor in an FFT is thermal noise (like from resistors), but that's just an assumption (based on subjective opinion, which everybody has their own opinions).
Last edited:
The graphs in post #52 don't show anything related to FFT.
So once again lots of uncorrelated speculation but nothing concrete. This has become a common theme with your posts.
The graphs in #52 show various time domain waveforms that can only look like the very same white noise in an FFT (as it is a given that they are white). Hard for me to see how that wouldn't be obvious from reading the recent/present discussion and the linked material in #52, but maybe it isn't as obvious as I tend to think.
In principle its not much different from what is explained in the article at: https://purifi-audio.com/blog/tech-notes-1/doppler-distortion-vs-imd-7 The article shows that two very different time domain signals can look exactly the same in FFT spectral analysis.
That same principle also applies to how noise looks in FFT spectral analysis. If anyone here feels like they would like some tutoring to better understand the subject, please feel free to PM.
Last edited:
Yes, I had slight better results with added USB isolation (with Silanna chip) + one ISO followed with recklocking each digital bus line (and some level att to lo-hi margins) after pure digital modules, infront of DAC chip.
Clearly more "analog" sound, and measurements was I think with less noise in top HF region...
.
Thanks for the info about OPA1565. These new dac chips accumulating enormous thermal energy...
Yes. Probably more because in this specific case. I think, AK4499EX cant work without AK4191 chip?
I am certanly not in that sect 🙂
but I think that wisual representation of measurements can be used in context of electrical circuit simulations AND final sound outcome? Every other way, for my opinion, is or pure theorethical with measurements only, or pure subjective (listening only). These 2 distant points that not examine whole context, without clear connection to each other, neglecting final objective... Instead of excluding i rather choose complementary aproach 🙂
Some additional datas about summing outputs from AKM PDF document
ak4191eq-en-designtransition-myakm.pdf
considering pair of ICs AK4191 followed by AK4499EX
this PDF revealing that internal source resistance of (probalbly each Iout is 110 ohm.
And that input value (as DC voltage potential) at input VCOM of AK4499EX can be calculated.
.
Also there is ilustrated how can Ater I to V conversion can be set back to current value.
And more claer explanation of Voff and Vdcbias values.
.
In 7.4 some formula is given. That can basicaly be as
Voff = VCOM = 2.5V x [ Rfb / (Rfb + Rsource) ]
Vhere Rfb = feddbck resistor in transimpedance OP circuit
Rzource = 110 ohm
.
Putting the Rfb = Rsource = 110 ohms
We get round 4Vp-p (11.32Vrms)
and
Voff = VCOM = 1.25V
So maybe this is the virtual ground for potentional use of simple Riv at output?
But it is not clear... 🙁 I will check it out with spice...
.
Strange that in other document we have actualy more relevant information than in the DAC datasheet?
Somehow old story for AKM?
ak4191eq-en-designtransition-myakm.pdf
considering pair of ICs AK4191 followed by AK4499EX
this PDF revealing that internal source resistance of (probalbly each Iout is 110 ohm.
And that input value (as DC voltage potential) at input VCOM of AK4499EX can be calculated.
.
Also there is ilustrated how can Ater I to V conversion can be set back to current value.
And more claer explanation of Voff and Vdcbias values.
.
In 7.4 some formula is given. That can basicaly be as
Voff = VCOM = 2.5V x [ Rfb / (Rfb + Rsource) ]
Vhere Rfb = feddbck resistor in transimpedance OP circuit
Rzource = 110 ohm
.
Putting the Rfb = Rsource = 110 ohms
We get round 4Vp-p (11.32Vrms)
and
Voff = VCOM = 1.25V
So maybe this is the virtual ground for potentional use of simple Riv at output?
But it is not clear... 🙁 I will check it out with spice...
.
Strange that in other document we have actualy more relevant information than in the DAC datasheet?
Somehow old story for AKM?