yep isn't flat without rc please see better the pic you post , with 3300(why not more ?) and r10 is the better profile,living as Markw4 post on sound not so good in this case
Last edited:
So that’s slightly increased harmonic distortion up in the higher freq?
wonder if that colors the sound way down there
wonder if that colors the sound way down there
While there may be people who are able to hear a difference at these low levels I doubt the hump is a cause for the bad sound you reported especially since these are measured from evaluation boards. So the hump probably does not exist in every dac using AK4499 (or AK4493).
Isn't that just a measurement artefact? With a bandwidth of around 21 kHz (44.1 kHz sampling) the third harmonic will "disappear" once your fundamental gets above 7 kHz. There is also a small reduction just above 10 kHz. This could be where the second harmonic is outside the measuement bandwidth.
You are probably right. Although in 192k measurement the hump drops a bit too soon if caused by "disappearance" of 3rd hd and no reduction for 2nd hd disappearing.
Attachments
Whether it is "too soon" depends on the bandwidth of the analyzer input. The "disappearance" of the 3rd is just visible I think, but less clear. Both could indicate a bandwidth of around 80 - 90 kHz.
Starting to play around with an AK4499EX eval board. Noticed it has a 'Sound Quality' mode, as do some other AKM dacs. Interesting thing is that its controlled from the AK4499EX registers rather than in the AK4191 registers. AK4499EX only has a few registers, most settings for the dac are in AK4191.
Anyway, I tried turning on SC mode and sure enough it sounds more distorted. But it also sounds like transients at all frequencies are more punchy, there is more space between instruments, and that the depth or 'black level' between instruments is deeper. Started thinking about what AKM could be doing in AK4499EX to have such effects. It occurred to me that they might be changing the DEM algorithm in some way, either disabling it or maybe changing its statistical distribution from gaussian. Why? Because when the dac is in 'measurement mode' instead of 'SQ' mode, instruments are more blurred together, perceived dynamics are less, and depth between instruments (or, maybe call it low level detail) is less. Guessing that DEM, which is supposed to turn resistor and switch mismatch distortion into gaussian noise, is somehow audible as a correlated noise effect. Of course the above is only tentative speculation, nothing more.
Anyway, I tried turning on SC mode and sure enough it sounds more distorted. But it also sounds like transients at all frequencies are more punchy, there is more space between instruments, and that the depth or 'black level' between instruments is deeper. Started thinking about what AKM could be doing in AK4499EX to have such effects. It occurred to me that they might be changing the DEM algorithm in some way, either disabling it or maybe changing its statistical distribution from gaussian. Why? Because when the dac is in 'measurement mode' instead of 'SQ' mode, instruments are more blurred together, perceived dynamics are less, and depth between instruments (or, maybe call it low level detail) is less. Guessing that DEM, which is supposed to turn resistor and switch mismatch distortion into gaussian noise, is somehow audible as a correlated noise effect. Of course the above is only tentative speculation, nothing more.
@Markw4: Hi Mark ... Sounds interesting that you are beginning to delve into the AK4191 & 4499EXEQ combination ... I admittedly consider them to be "over my head" complexity-wise but the AK4499EXEQ could be interesting in itself IF it were capable of accepting signals not coming from the 4191.
Would you happen to know if this is possible? As far as I can see it is a 28 bit signal (7*4 bit) entering the 4499EXEQ, but sampling at 5/6 or 11/12 MHz. So likely it needs the 4191 to work, right? Or do you think it would be possible to e.g. enter a 1.536 MHz (i.e. the first 28 bits) signal into the 4499 split into four 7 bit "chunks" at e.g. 5/6 MHz bclk?
Just curious as to what are the options for omitting the 4191, if possible ... 🤔
Cheers, Jesper
Would you happen to know if this is possible? As far as I can see it is a 28 bit signal (7*4 bit) entering the 4499EXEQ, but sampling at 5/6 or 11/12 MHz. So likely it needs the 4191 to work, right? Or do you think it would be possible to e.g. enter a 1.536 MHz (i.e. the first 28 bits) signal into the 4499 split into four 7 bit "chunks" at e.g. 5/6 MHz bclk?
Just curious as to what are the options for omitting the 4191, if possible ... 🤔
Cheers, Jesper
"Would you happen to know if this is possible?"
You might be able to program an FPGA to drive AK4499EX using your own 7-bit sigma-delta modulator design, but it might end up being more work than figuring out how to use AK4191. Sorry.
You might be able to program an FPGA to drive AK4499EX using your own 7-bit sigma-delta modulator design, but it might end up being more work than figuring out how to use AK4191. Sorry.
Thanks for the feedback, Mark. Thought it would be so, but well, good to have it clarified.
Cheers, Jesper
Cheers, Jesper
As it turns out AK4191 doesn't seem to be that hard to use. The eval board comes with a script to program the registers to auto set the sample rate for PCM playback. I made another AK4191 script to play DSD256 in volume bypass mode. The scripts run in an AKM program that communicates through their proprietary USB to I2S/SPI interface. However using the same register settings with any microprocessor that can control I2C bus should work just as well. There is also a script for AK4499EX.
IMHO, as this SC is at the DAC part and not on the modulator part... so what they really do?
The documentation, as the Evaluation measurements shows, done as SC = 0, this means all AKM / AP graphs configured as for measurements.
Just wonder if one measures with SC = 1 what really alters in measurement terms. Or AKM provides some informations / glue about.
So if the reviewers post plots as with unknown SC state ... this will get soon an interpretation and compare mess.
Take a look at the PecanPi+ DAC, which uses the 4491 and 4499.
https://orchardaudio.com/product/pecanpi-dac-plus/
https://orchardaudio.com/product/pecanpi-dac-plus/
Holy Mozes that thing is expensive!!!
I don't get it honestly.... with that horrendous case and absolute crap of an SMPS (and I keep saying the PS is 99% of the final sound quality)... it costs $750. And, you can still feed S/PDIF only..? via unshielded wires..? to a screw terminal..?? Does it at least have a decent pulse transformer??? No network streaming, no USB audio...?
Seriously??
With a half-decent liner power supply and Raspberry Pi, you are looking at more than $1000...easily... but still, end up being confined to that horrendous case. I checked a few videos... using that long flexible ribbon cable to connect Raspery Pi and DAC PCB is anything but acceptable.
Psst... Gustard A26 is $1499
I don't get it honestly.... with that horrendous case and absolute crap of an SMPS (and I keep saying the PS is 99% of the final sound quality)... it costs $750. And, you can still feed S/PDIF only..? via unshielded wires..? to a screw terminal..?? Does it at least have a decent pulse transformer??? No network streaming, no USB audio...?
Seriously??
With a half-decent liner power supply and Raspberry Pi, you are looking at more than $1000...easily... but still, end up being confined to that horrendous case. I checked a few videos... using that long flexible ribbon cable to connect Raspery Pi and DAC PCB is anything but acceptable.
Psst... Gustard A26 is $1499