Dither ONLY functions on the LSB of the OUTPUT - not internal inter-stage data truncation which should be correctly Rounded + dithered.
With 32Bit output mode, there is no dithering (on the output)...
This lack of correct internal truncation manifests itself as low level harmonics spread across the noise floor - especially apparent with low input signal levels...
Also, be VERY careful with the maxim upsampling ratio, eg. for PDM256 output you need a minimum of 176KHz input.
So you cannot upsample to PDM256 from 44.1KHz or 96KHz etc... The AK4137 will appear to work (you will hear sound) but the output will be distorted (visible via FFT).
The Max "upsampling" ratio's are described (in the admittedly VERY confusing Datasheet) Page 21 Input / Output examples.
External filters are possible with the AKM4137, eg x8 oversampling from 44.1KHz to 352.8KHz then input to AKM... The AKM can handle a max of x16 (768KHz input).
Last edited:
Not sure if adding extra digital filtering is a positive thing, IMO PCM sound quality is crippled by digital filtering - adding more is not the way to go!
Last edited:
Understood about all the filtering, but it can still sound better than baseband 16/44 reproduction. Do you have a better choice for us?
Sorry, not been following the thread, so maybe my reply is out of context, but as far as I'm aware, you have to use the ASRC mode if you use the AK4137 as a PCM to PDM (DSD) modulator.
A repeated design weakness with AKM designs is that they dont seem to Round and Dither when they internally truncate internal data paths - so you will see THD components at lower signal levels popping up from the noise floor
The AK4137 design is a conundrum IMV. A high performance chip by the specs however they were unable to give any information WRT jitter rejection
corner freq of the ASRC section. This is fundamental to the design so I'm
guessing the JR CF is fairly high.
I have not played with AK4137, I'm surprised with your results, thanks for sharing.
Have you compared either 4137 or HQP to any file converters for example
Saracon or Tascam etc?
TCD
Figure 59 of the data sheet shows the jitter tolerance to ILRCK. "The jitter quantity is defined by the jitter frequency and the jitter amplitude shown in Figure 59. When the jitter amplitude is 0.01UIpp or less, the AK4137 operates normally regardless of the jitter frequency."
True, the graph only indicates from 1Hz - but at least the attenuation at 1Hz is still "Flat".
We where REALLY surprised how bad HQplayer sounded... We could hear differences between every combination tried... but non sounded "Good" IMO... Maybe we have been spoiled by Native DSD, and to a lessor degree the AK4137 modulator.
Never heard the other file converters (can they convert from PCM to PDM)?, although I have heard the Saracon software ASRC, and was NOT impressed!!! I have not heard any PCM ASRC / SRC (hardware or software) that can compete with NativeDSD just reinforcing my distrust in any form of FIR filtering...
Last edited:
What do you guys recommend for the VCCA supply? As far as I understand this supply is only for the oscillator inverter/driver, is that correct? If so it still should be very clean as otherwise any noise on there might appear in the clock signal as AM phase noise, right?
Any advantages/disadvantages from driving the digital supply with 1.8 vs 3.3V?
Any advantages/disadvantages from driving the digital supply with 1.8 vs 3.3V?
A 3.3v good performance LDO would be suggested. Used LT1763 myself. Other ones are good too, some measure better.
One more (hopefully final) question about the schematic in #3003: it shows a 47u electrolytic or a 22u film cap on the output of the opamp. Is this the only filter cap on the AVCC line then? I read that paralleling caps with different values can lead to awful spikes in the impedance because the inductance of the bigger cap forms a parallel resonant circuit with the smaller cap. The value of 22/47u just seems a bit high to me as I'm more used to 0.1-1u directly at the chip pins, so I thought I'd ask to clarify.
And one more thing, I'm planning to use the 9038Q2M in mono mode, so i have to power two AVCC pins. having two opamps there to feed them seems a bit overkill, right? If so, do I put two caps with half the value each right next to the pins and feed them from a single supply, best with symmetric layout for the trace(s) from the output of the opamp to the two AVCC pins?
The thing about either a film cap or an electrolytic was an experiment. ESS recommended an electrolytic, and AKM uses standard electrolytic cap filtering for AK4499 in a partially similar way. After some more thinking about it, it may fine or possibly preferred to use a standard electrolytic, not even a higher performance one. Eventually, I want to do some experimenting with AK4499 to see if the particular type of electrolytic cap affects the sound audibly.
The ESS AVCC circuit and capacitor recommendation info is in a document available on the ESS website downloads page:
http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
ESS Technology :: Downloads
(I would not use DVCC as a reference voltage source, however.)
If mono mode, only one AVCC supply per dac should be enough. The purpose of having two AVCC supplies is to improve stereo imaging.
The ESS AVCC circuit and capacitor recommendation info is in a document available on the ESS website downloads page:
http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
ESS Technology :: Downloads
(I would not use DVCC as a reference voltage source, however.)
If mono mode, only one AVCC supply per dac should be enough. The purpose of having two AVCC supplies is to improve stereo imaging.
Last edited:
My experience was that a better CAP had better voice too.
(much earlier I tried several CAPs with the q2m version - OPA based ref.)
(Currently I'm playing with the PRO version of the board and uses LME49600 headphone amp for AVCC. (it is capable to feed any capacitive load))
- I tried 47uF Elna Silmic II + 22uF x7r
- and also tried 1000uF Elna Silmix II + 22uF x7r.
For me the 2nd version was better - stronger bass, somehow clearer voice in general.
I also wanted to remove the x7r - to see if there is any effect. (my former experience was that it coloured the treble on the OPAMP VCC/VEE line) - but not yet tried here.
(much earlier I tried several CAPs with the q2m version - OPA based ref.)
(Currently I'm playing with the PRO version of the board and uses LME49600 headphone amp for AVCC. (it is capable to feed any capacitive load))
- I tried 47uF Elna Silmic II + 22uF x7r
- and also tried 1000uF Elna Silmix II + 22uF x7r.
For me the 2nd version was better - stronger bass, somehow clearer voice in general.
I also wanted to remove the x7r - to see if there is any effect. (my former experience was that it coloured the treble on the OPAMP VCC/VEE line) - but not yet tried here.
Last edited:
Mark - question:
before Christmas my q2m gave up.
I was just listening music and suddenly it was muted.
Now the LED is continuously on (which shows that there is any input signal)
And strange, but the same happened at the same time with my SRC4392 board too. (same symptom, both LEDs are on continuously.)
Any guess what could be the problem?
The SRC4392 would be more important....
Thanks
before Christmas my q2m gave up.
I was just listening music and suddenly it was muted.
Now the LED is continuously on (which shows that there is any input signal)
And strange, but the same happened at the same time with my SRC4392 board too. (same symptom, both LEDs are on continuously.)
Any guess what could be the problem?
The SRC4392 would be more important....
Thanks
Which SRC4392 board do you have?
If both boards went out at the same time it could have been a power line transient that damaged something.
Have you measured all the DC voltages, input to the boards, and output of each of the 3.3v regulators on the boards?
If something damaged, could be the MCU, could be the dac chip, AK4137 chip or some mix of the foregoing.
Could be the AVCC regulator needed more caps on its own power supply. Could also be that it was not located close enough to the dac chip, not on the same ground plane, some other problem, or maybe a combination of problems.
Hi Guys again,
I realised meanwhile that the x7r was not in originally.
So I added, but quite quickly I removed again.
I would like to emphasize that listening test are always subjective.
For me it seems again that the x7r colored the sound. The trebble were more bright a little bit but on the other side the sound became how to say flickering / greasy / dirty (sorry I dont know the exact english word for that)
I recommended using Silmic II without MLCC a long time ago. There was a
good reason for this. If you really need to bypass, try a/ PML film or b/ Al
Polymer Both have very good HF performance.
TCD
Avcc advice for beginner
I am going to attempt the AVCC psu. I have a 49720 here to try aswell as opa1612
I have studied Slims schematic and also a very neat hand drawn on graph paper image that I saved from this thread.
I have some questions please.
I understand that pins 3 and 5 of the opamp are fed from a 3v3 source . I will probably use lt1963 for this for now
Pins 1+2 and 6+7 are shorted.
Pins 4+8 for 15v+-
Where does Vref actually connect to.? I understand that the 2 avcc feeds would go to the equivalent of the original AVCC cap + pads.
Thanks for your patience
I am going to attempt the AVCC psu. I have a 49720 here to try aswell as opa1612
I have studied Slims schematic and also a very neat hand drawn on graph paper image that I saved from this thread.
I have some questions please.
I understand that pins 3 and 5 of the opamp are fed from a 3v3 source . I will probably use lt1963 for this for now
Pins 1+2 and 6+7 are shorted.
Pins 4+8 for 15v+-
Where does Vref actually connect to.? I understand that the 2 avcc feeds would go to the equivalent of the original AVCC cap + pads.
Thanks for your patience