Here is the link referring to the clock and driver discussions.
One learning: If something is wrong with the kernel/driver/module setup the DAC will be driven with PI-I2S @ 2.82MHz as the DAC+.
It's not sufficient to just have a kernel/driver/modules supporting the "Pro" available.
It's a matter of what's being loaded and blacklisted. Certain modules even interfere if not blacklisted - and that causes issues!
I'm pretty sure though that the PiCore guy knows what he's doing. I'm a bit late to the party.
I did have a great working Arch image for USB DACs. It still took me almost a week to shape my Arch image for the DAC+ Pro.
Though I have to admit the main effort went into using the on-chip volume control with squeezelite.
The Hifiberry setup documentation for the Pro was pretty much useless on Arch Linux btw..
I'll fire up my scope later today to see if my clocks are working properly. Just to make sure.
One learning: If something is wrong with the kernel/driver/module setup the DAC will be driven with PI-I2S @ 2.82MHz as the DAC+.
It's not sufficient to just have a kernel/driver/modules supporting the "Pro" available.
It's a matter of what's being loaded and blacklisted. Certain modules even interfere if not blacklisted - and that causes issues!
I'm pretty sure though that the PiCore guy knows what he's doing. I'm a bit late to the party.
I did have a great working Arch image for USB DACs. It still took me almost a week to shape my Arch image for the DAC+ Pro.
Though I have to admit the main effort went into using the on-chip volume control with squeezelite.
The Hifiberry setup documentation for the Pro was pretty much useless on Arch Linux btw..
I'll fire up my scope later today to see if my clocks are working properly. Just to make sure.
Last edited:
Just a by the by... I'm using a DAC+ Pro with MoOde Player running the latest kernel for dual clock support. I have hacked the I2S from the on board pads see:
https://support.hifiberry.com/hc/en-us/articles/205711451-DAC-Pro-connect-external-I2S-DACs
Powered both Pi and DAC from one LPS. Connected to my AYA DAC it sounds just-listen-all-day beautiful with no hardware mods....yet!
https://support.hifiberry.com/hc/en-us/articles/205711451-DAC-Pro-connect-external-I2S-DACs
Powered both Pi and DAC from one LPS. Connected to my AYA DAC it sounds just-listen-all-day beautiful with no hardware mods....yet!
Great info. Thx.
I saw inmate DimDim mentioning in the Soekris DAC thread to use I2S from the DAC+ Pro instead of the PI-I2S.
The Soekris DAC even isolates and reclocks I2S. The isolator can then be powered with 3.3V from the PI.
That might be a future option/project I'm looking into.
First I want to get the D+ Pro under control. And the Soekris DAC, currently at rev. 3, - as good as it seems already now - still evolves steeply (due to great DIY-Audio/Soekris interaction).
However. Perhaps the DAC+ Pro is becoming good enough to stay - after the mods.
Price/Performance ratio of this project I consider outstanding already now.
I saw inmate DimDim mentioning in the Soekris DAC thread to use I2S from the DAC+ Pro instead of the PI-I2S.
The Soekris DAC even isolates and reclocks I2S. The isolator can then be powered with 3.3V from the PI.
That might be a future option/project I'm looking into.
First I want to get the D+ Pro under control. And the Soekris DAC, currently at rev. 3, - as good as it seems already now - still evolves steeply (due to great DIY-Audio/Soekris interaction).
However. Perhaps the DAC+ Pro is becoming good enough to stay - after the mods.
Price/Performance ratio of this project I consider outstanding already now.
Last edited:
Fair enough. I'll leave it there, then!
You might like to update your patch to the latest one I submitted in the RPi pull request, if you also want the option to get the gain back, via a dt param.
You were using a kernel built from the rpi-4.4.y tree, right?
0002-Limit-PCM512x-Digital-gain-to-0db-by-default-with-Hi.patch
Then the behaviour stays as it is with "dtoverlay=hifiberry-dacplus" in your config.txt, which you probably want if using hard-vol with squeezelite.
Boot with "dtoverlay=hifiberry-dacplus,24db_digital_gain" if you want the 24db gain available from the "Digital" volume control.
Thx for the heads up. I'll check it out.
I'm wondering though what the update of the patch does?!?
You basically just allow to disable the original 0db@100% patch, don't you?
Is there also a 4.1.y patch update?
###
I recently thought why using the on-DAC VC at all? I mean, it is digital anyhow.
The datasheet speaks of simple or primitive digital VC. Whatever that means.
I guess setting the DAC to 0db and using the LMS SW VC wouldn't be any worse
than just using the DAC VC alone. That also makes it easier to maintain those patches.
You basically can leave them all out.
I'm wondering though what the update of the patch does?!?
You basically just allow to disable the original 0db@100% patch, don't you?
Is there also a 4.1.y patch update?
###
I recently thought why using the on-DAC VC at all? I mean, it is digital anyhow.
The datasheet speaks of simple or primitive digital VC. Whatever that means.
I guess setting the DAC to 0db and using the LMS SW VC wouldn't be any worse
than just using the DAC VC alone. That also makes it easier to maintain those patches.
You basically can leave them all out.
Yes, that's it. That's all it does.
I wasn't going to bother with that. IIRC, 4.4.y is about to become RPi stable branch.
LOL. Yes, exactly. And I could tell you about the blind listening test again, where internal LMS vol (using 0db on DAC) was compared to using hw-vol on DAC..... No-one was reliably able to distinguish between that either.....
I now added a couple of OSCONs SEPC (330/16) (2 before and 2 after the 3.3V regulator and also 2 of them on the 3.3V rail for the DVDD.
Beside that I added some silver mica caps on the PI power rails. I also added another
2*1000uf right at the comb.
The left and right channel are now soldered to the bottom joints of the jacks. Before I was using the holes (R/GND/L) on the right side of the board.
What should I say!?!? I'm blown away - even after just one day break-in.
The earlier reported weaknesses, such as slight boominess or slight congestion on complex pieces disappeared. Beside that, everything else improved.
I can hear now very nicely articulated low-end, applause sounds more natural, reverberations are better articulated. I pretty much experience progress on all fronts.
Great. I really like what I hear.
Beside that I added some silver mica caps on the PI power rails. I also added another
2*1000uf right at the comb.
The left and right channel are now soldered to the bottom joints of the jacks. Before I was using the holes (R/GND/L) on the right side of the board.
What should I say!?!? I'm blown away - even after just one day break-in.
The earlier reported weaknesses, such as slight boominess or slight congestion on complex pieces disappeared. Beside that, everything else improved.
I can hear now very nicely articulated low-end, applause sounds more natural, reverberations are better articulated. I pretty much experience progress on all fronts.
Great. I really like what I hear.
Hi - it's me again.
Now with the more revealing DAC in place, I "quickly" tested
1. the variable ON-chip digital volume control controlled via squeezelite against
2. the variable Squeeze software volume control and the ON-chip volume control fixed @ 0db with alsamixer.
Running quickly three of my all time testtracks cleary show to me number 2. as my
preferred choice.
(Nope. I won't invite 10 guys for double-blind testing it. I could run a half-deaf test and would still hear it. )
Though. I'll keep an eye on that subject, since the DAC still needs some break-in. Things might change again.
However.The key learning for me was - there is a difference! I'll retest in a couple of days to avoid fooling myself.
Enjoy.
Now with the more revealing DAC in place, I "quickly" tested
1. the variable ON-chip digital volume control controlled via squeezelite against
2. the variable Squeeze software volume control and the ON-chip volume control fixed @ 0db with alsamixer.
Running quickly three of my all time testtracks cleary show to me number 2. as my
preferred choice.
(Nope. I won't invite 10 guys for double-blind testing it. I could run a half-deaf test and would still hear it.
)Though. I'll keep an eye on that subject, since the DAC still needs some break-in. Things might change again.
However.The key learning for me was - there is a difference! I'll retest in a couple of days to avoid fooling myself.
Enjoy.
Last edited:
Yes, no longer need patches. My pull requests were accepted for both rpi-4.1.y and rpi-4.4.y branches. (4.4.y, yesterday and 4.1.y, this morning.) So if building your own kernels from latest RPi git source, you no longer need to apply.
I also think that they have built and pushed pre-built kernel with those patches to RPi firmware git, so Raspian users will now have the functionality included in any binary kernel updates they receive.
Hi.
For those who consider to modify the DAC I'd like to post what type of cap has been discussed over in the Soekris thread and e.g. Hifiduino:
It's from Nichicon - PartNo: RR71C331MDN1
It's pretty low ESR - 7mR and costs around 50 cent each as 10-pack.
#####some thoughts############
I think the Hifiberry guys should have a look at improving the power/regulation/decoupling situation.
* clock power
* DVDD
should be supplied by separate high quality regulators.
It's IMO a NoGo to use the PI2 3.3V. ( See also e.g. Soekris DAC clock regulator)
Also chargepump and AVDD should be properly powered and separated.
The external 5V option on the board is pretty useless nowadays, since the 3.3V comes from the PI2
anyhow. That could change if they'd introduce a separate 3.3V regulator.
When looking at e.g the MamboBerry DAC, DurioSound, Audiophonics I-Sabre V2
they all seem to have put more efforts on the powersupply/decoupling. Of course they also charge more for their DAcs.
Pricewise - by looking at the market - Hifiberry could end up in the $60-70 range for a higher quality DAC.
That'll give them plenty of budget to improve the current "DAC+ Pro".
So HifiBerry - if you read this - I hope you'll be working on a "DAC+ Audiophile" soon.
Why am I saying that!?? The DAC+ Pro is quite a nice DAC. However. I think Hifiberry is not using the DACs full potential.
My basic tweaks already show benefits.
Adding a little more brain into the board design could make the DAC a real nice one - out of the box.
Cheers
For those who consider to modify the DAC I'd like to post what type of cap has been discussed over in the Soekris thread and e.g. Hifiduino:
It's from Nichicon - PartNo: RR71C331MDN1
It's pretty low ESR - 7mR and costs around 50 cent each as 10-pack.
#####some thoughts############
I think the Hifiberry guys should have a look at improving the power/regulation/decoupling situation.
* clock power
* DVDD
should be supplied by separate high quality regulators.
It's IMO a NoGo to use the PI2 3.3V. ( See also e.g. Soekris DAC clock regulator)
Also chargepump and AVDD should be properly powered and separated.
The external 5V option on the board is pretty useless nowadays, since the 3.3V comes from the PI2
anyhow. That could change if they'd introduce a separate 3.3V regulator.
When looking at e.g the MamboBerry DAC, DurioSound, Audiophonics I-Sabre V2
they all seem to have put more efforts on the powersupply/decoupling. Of course they also charge more for their DAcs.
Pricewise - by looking at the market - Hifiberry could end up in the $60-70 range for a higher quality DAC.
That'll give them plenty of budget to improve the current "DAC+ Pro".
So HifiBerry - if you read this - I hope you'll be working on a "DAC+ Audiophile" soon.
Why am I saying that!?? The DAC+ Pro is quite a nice DAC. However. I think Hifiberry is not using the DACs full potential.
My basic tweaks already show benefits.
Adding a little more brain into the board design could make the DAC a real nice one - out of the box.
Cheers
Last edited:
@Soundcheck,
Thanks for the information and the thoughts. While I haven't put mine into a system to assess yet (Some household repairs took more time and energy last weekend than I planned), based on your findings and thoughts, I have some comments too, both for mods to the existing unit and for an 'Audiophile Edition'...
On mods, I think you've pointed the way with your additional caps on the boards (both HFBD+P & the R-Pi). For mine, I'm thinking of a couple of rounds of mods:
1. Add additional low-ESR caps on the 5v rail to the DAC, the 5v input to the R-Pi, and the 3.3v rail back to the DAC. Along with this, I'll probably replace all of the ceramic caps on the AVDD and charge pump rails with Panasonic PPS film SMD caps. I first saw this on the Ayre USB DAC and the K&K RAKK DAC, later saw EUVL go to it on his ES9022 DAC cards, and more recently in the stock configuration in the Sony HAP Z1-ES I modified the heck out of last summer and also in a JLSounds 4490 DAC card I just got to play with... So it seems to be a hot ticket!
I'll start this round with a single power input to the DAC card, then later remove the zero-ohm jumper and input 5v or 3.3v into the Pi/DAC digital/clocks side.
2. Cut the trace on the 3.3v feed from the Pi to the DAC digital/clocks side and power the Pi and the DAC digital/clocks separately. If I get ambitious here (and if it is sounding really good), I may do some additional surgery and power the clocks with a separate 3.3v feed.
One alternative on the first stage is to consider the AVDD power as more of an analog power feed than a digital one. While the best solution would be to feed the charge pump and the AVDD separately as you suggest, I might just try Black Gates instead of low-ESR digital-style caps on the AVDD.
Then on what I'd like to see long term, first, I think what you proposed is a good start and at least a minimum for an 'audiophile' version. The two things I'd add for consideration are:
1. Adding isolators between the Pi & the DAC card with re-clocking on the DAC side. A prerequisite for this are completely separate power supplies & grounds. One MAY be able to do this while still using the master-slave mode of the PCM5122, but it may force a move to a hardware-generated bitclock that is fed to the Pi... I've seen one good configuration doing that for another application.
2. Moving to the John Swenson trick of using a specially-designed up-sampling filter to 352/384 on the R-Pi source device to replace/bypass the on-chip filtering of the PCM5122. He did this with the fairly-well received Bottlehead DAC (that as I remember used the PCM5142, a similar chip using the same D-A conversion process). I'm not sure the current R-Pi is capable of this and I suspect it will require some significant effort on the driver, but replacement filtering along with isolation and reclocking would put it in the same realm as the upper-middle tier of DACs... the ones above this would be ones with extreme power supplies and clocks, extreme paralleled DACs (DDAC), very exotic filtering (Schitt, PS Audio), or non-DAC chip D-A conversion (PS Audio, Chord, HQPlayer, Lampizator). Also, I understand the Sonore USB->I2S DIY interface has a special upsampling filter done in the FPGA to bypass the input filters on the ES9018 DAC chips commonly used with them.
Soundcheck, I know you've tested upsampling with mixed results, but since the DAC chip is already doing this, re-doing it in a way that we can better control the parameters could bring significant gains.
I could see a version as above selling in the $300-$500 USD range, depending on power options. AND it might be the best value for a player around if it were done.
My 2 cents!
Greg in Mississippi
P.S. @Bas Horneman, what is Durio Sound doing with the stacked DAC cards? I didn't see any info on the details or why on their website. Paralleling the DACs for lower noise and higher output drive? What & Why?
Thanks for the information and the thoughts. While I haven't put mine into a system to assess yet (Some household repairs took more time and energy last weekend than I planned), based on your findings and thoughts, I have some comments too, both for mods to the existing unit and for an 'Audiophile Edition'...
On mods, I think you've pointed the way with your additional caps on the boards (both HFBD+P & the R-Pi). For mine, I'm thinking of a couple of rounds of mods:
1. Add additional low-ESR caps on the 5v rail to the DAC, the 5v input to the R-Pi, and the 3.3v rail back to the DAC. Along with this, I'll probably replace all of the ceramic caps on the AVDD and charge pump rails with Panasonic PPS film SMD caps. I first saw this on the Ayre USB DAC and the K&K RAKK DAC, later saw EUVL go to it on his ES9022 DAC cards, and more recently in the stock configuration in the Sony HAP Z1-ES I modified the heck out of last summer and also in a JLSounds 4490 DAC card I just got to play with... So it seems to be a hot ticket!
I'll start this round with a single power input to the DAC card, then later remove the zero-ohm jumper and input 5v or 3.3v into the Pi/DAC digital/clocks side.
2. Cut the trace on the 3.3v feed from the Pi to the DAC digital/clocks side and power the Pi and the DAC digital/clocks separately. If I get ambitious here (and if it is sounding really good), I may do some additional surgery and power the clocks with a separate 3.3v feed.
One alternative on the first stage is to consider the AVDD power as more of an analog power feed than a digital one. While the best solution would be to feed the charge pump and the AVDD separately as you suggest, I might just try Black Gates instead of low-ESR digital-style caps on the AVDD.
Then on what I'd like to see long term, first, I think what you proposed is a good start and at least a minimum for an 'audiophile' version. The two things I'd add for consideration are:
1. Adding isolators between the Pi & the DAC card with re-clocking on the DAC side. A prerequisite for this are completely separate power supplies & grounds. One MAY be able to do this while still using the master-slave mode of the PCM5122, but it may force a move to a hardware-generated bitclock that is fed to the Pi... I've seen one good configuration doing that for another application.
2. Moving to the John Swenson trick of using a specially-designed up-sampling filter to 352/384 on the R-Pi source device to replace/bypass the on-chip filtering of the PCM5122. He did this with the fairly-well received Bottlehead DAC (that as I remember used the PCM5142, a similar chip using the same D-A conversion process). I'm not sure the current R-Pi is capable of this and I suspect it will require some significant effort on the driver, but replacement filtering along with isolation and reclocking would put it in the same realm as the upper-middle tier of DACs... the ones above this would be ones with extreme power supplies and clocks, extreme paralleled DACs (DDAC), very exotic filtering (Schitt, PS Audio), or non-DAC chip D-A conversion (PS Audio, Chord, HQPlayer, Lampizator). Also, I understand the Sonore USB->I2S DIY interface has a special upsampling filter done in the FPGA to bypass the input filters on the ES9018 DAC chips commonly used with them.
Soundcheck, I know you've tested upsampling with mixed results, but since the DAC chip is already doing this, re-doing it in a way that we can better control the parameters could bring significant gains.
I could see a version as above selling in the $300-$500 USD range, depending on power options. AND it might be the best value for a player around if it were done.
My 2 cents!
Greg in Mississippi
P.S. @Bas Horneman, what is Durio Sound doing with the stacked DAC cards? I didn't see any info on the details or why on their website. Paralleling the DACs for lower noise and higher output drive? What & Why?
LOL... Just saw this comment from EUVL:
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-37.html#post4611442
Greg in Mississippi
http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-37.html#post4611442
Greg in Mississippi
@Greg
I do not intend to make a new DAC from that DAC+ Pro.
First of all, the base needs to be right!
If some lightweight mods are not sufficient, I'll drop that device.
That's not the case right now. The device seems to have some obvious weaknesses.
However. I'm more then happy with the device as is - after my lightweight mods (HW and SW).
If I want to spent 200-300$, I buy a Soekris DAC V3. That DAC has everything you bring up as improvement proposals already onboard. And then it's an R2R DAC.
*******
I don't have anything against upsampling in general.
An external high quality upsampler might beat an on-dac upsampler.
It'll differ from DAC to DAC.
I wouldn't use the PI for extensive DSP work though.
And I wouldn't have an idea how to pass the internal 5122 DAC filters by.
I do not intend to make a new DAC from that DAC+ Pro.
First of all, the base needs to be right!
If some lightweight mods are not sufficient, I'll drop that device.
That's not the case right now. The device seems to have some obvious weaknesses.
However. I'm more then happy with the device as is - after my lightweight mods (HW and SW).
If I want to spent 200-300$, I buy a Soekris DAC V3. That DAC has everything you bring up as improvement proposals already onboard. And then it's an R2R DAC.
*******
I don't have anything against upsampling in general.
An external high quality upsampler might beat an on-dac upsampler.
It'll differ from DAC to DAC.
I wouldn't use the PI for extensive DSP work though.
And I wouldn't have an idea how to pass the internal 5122 DAC filters by.
One more hint. I did recommend the iFi iPower as power supply and I still do.
However.
To power the PI directly from the iPower it needs a 5.5mm to micro USB adapter.
iFi doesn't put one in the box -- yet.
I ordered some of these .
They work well and are more then affordable. Shipment took 10 days to Germany.
However.
To power the PI directly from the iPower it needs a 5.5mm to micro USB adapter.
iFi doesn't put one in the box -- yet.
I ordered some of these .
They work well and are more then affordable. Shipment took 10 days to Germany.
Not sure if higher output drive is as valid as for a current output DAC. But I've read "stuff" which claims improved dynamic range..etc. etc.
Just one quote I found on diyaudio:
"Accuphase states that paralelling two Dacs reduces distortion by square root of two (roughly 1.4, therefore a 40 % reduction) across the board, this is clearly stated in their product brochures."
Also found this:
"You can only parallell current output DAC's without problems. Voltage output units need to share load through resistors which is not optimal.
Parallel DAC's will have many advantages such as dynamic range, noise, lower bit distortion, output impedance etc."
So apparently it is not persé valid for the duriousound..being a voltage output dac. (But I did not research that when I took part in the kickstarter thingy)
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.