Just an update related to the iancanada HAT board.
He outlined yesterday that the board is specifically meant to isolate and reclock pretty much all major HAT DACs. The DAC+ Pro is also on his list of supported devices.
Bottom line. Thx Arbartels1965 for bringng the subject up. It very well fits this discussion.
He outlined yesterday that the board is specifically meant to isolate and reclock pretty much all major HAT DACs. The DAC+ Pro is also on his list of supported devices.
Bottom line. Thx Arbartels1965 for bringng the subject up. It very well fits this discussion.
@clive
Just had a look at your patches. Pretty impressive. No wonder though that upstream refuses
such a massive update for the time being.
Your own 4.4.y branch has it all in than. I might reanimate my ARCH installation to try it.
Three things come to mind.
1. I'm wondering if (I2S-)MMAP still works @ up to 384kHz
2. alsa-utils - such as aplay are not 384kHz compatible - these can't be used for testing unless they'll get patched.
3. Since you concentrate on the HAT fraction, there'll be no generic 384kHz support
for e.g. USB DACs. Correct?
PS:
If Vasili from MamboBerry is still around he might comment on 384kHz support - if the/an updated kernel supports it.
Just had a look at your patches. Pretty impressive. No wonder though that upstream refuses
such a massive update for the time being.
Your own 4.4.y branch has it all in than. I might reanimate my ARCH installation to try it.
Three things come to mind.
1. I'm wondering if (I2S-)MMAP still works @ up to 384kHz
2. alsa-utils - such as aplay are not 384kHz compatible - these can't be used for testing unless they'll get patched.
3. Since you concentrate on the HAT fraction, there'll be no generic 384kHz support
for e.g. USB DACs. Correct?
PS:
If Vasili from MamboBerry is still around he might comment on 384kHz support - if the/an updated kernel supports it.
Last edited:
Well, to be clear, this is not just me. I've bundled the Pi I2S specific stuff into a rpi-4.4.y tree to make it as easy as I can for you. (My kernels are built, combining 6 trees.) There is a heap of stuff on top of this, but the inter-dependencies start getting horrible. Some things have to be merged first before others can follow. And to give credit, the bulk of the patches in that rpi-4.4.y-simple tree are the clock/dma/i2s upstream backport, for which Martin Sperl gets the credit.
Yes, of course! MMAP is enabled by default now. No need to reference that i2s-mmap overlay in your config.txt. (Actually, one of my Revert patches removed the overlay.)
Yes, and the defines patch for pcm.h..... It seems that upstream ALSA aren't too keen on updating the defines, which is why everything I have done, is with KNOT/CONTINUOUS and constraints, so you don't need the "static" defines. But yes, some things outside kernel do need them. (But could be re-written not to.) I'm currently having a conversation with someone who may have a better plan on how to convince upstream that 352k8 and 384k are in common usage now..... Commonly enough used, that there should be "static" defines for them, and they should be in "core" rates array.
USB DAC's (and USB->I2S interfaces) have worked @ 384k for quite some time. (At least a couple of years.) If the USB device tells ALSA via Linux USB that is supports 384k, and if it actually does, it will work. All of the USB interfaces I use, like CM6631A, Amanero, JLLabs, WaveIO... work fine and have always done so with Linux at 352k8/384k, IIRC.
You will most likely see all of this in a future "official" RPi kernel, but it's going to take time. Martin Sperl is still waiting for upstream approval on some of his clock patches. Last time I looked, discussion was ongoing on rpi-kernel and kernel-arm mailing lists.
PS. And moving to using framework based things, like how those bclk ratios are specified from dt overlay, should also lend itself to other uses, such as choosing MASH1/2/3 noise shaping, for a specific clock, or sample rate/bit depth combination, with fractional divider. That's currently second on my list of things to look at. I lost my i2s-prefer-mash2/3 overlays with the move over to the new clock framework. I want to get that functionality back, but more fine-grained.
Last edited:
A couple of slow responses...
Two slow responses...
One, there's a question on what I use for linear supplies. For my low-end audio-level supplies such as the ones I used here, I use supplies based on the LT3080 or LT3083 chips… with the LT3080 good to up to 1.2a max and the LT3083 3a, both depending on the raw DC supply voltage and capability and heatsinking. I will buy and slightly modifiy K&K Audio Low Voltage Power Supply kits (here: K&K Audio | Other kits) with larger raw DC filter capacitors (KYB 12000UF 20% 16V RADIAL PN: EKYB160ELL123MM40S or 6800UF 20% 25V RADIAL PN: EKYB250ELL682ML40S depending on the raw DC voltage required) and the post-regulator filter cap with Nichicon Electrolytic Capacitors - Snap In 35volts 4700uF 20% PN: LKG1V472MESZBK. I also replace the mylar Wima bypass caps with polyprop equivalents.
Or I will buy Low-Current Dual Power Supply (LCDPS) boards from Twisted Pear Audio (here: Low Current Dual Power Supply (LCDPS) - A Dual Regulated CRC Supply) and build them up using the LT3080 or LT3083 chips depending on the current required as above. As these chips require a bit different output voltage setting network than the LM317-equivalent ones these boards are designed to use, that takes a little modification on the boards. I use either KYB 5600UF 20% 16V RADIAL (KYB PN EKYB160ELL562MK40S) or 3900UF 20% 25V RADIAL (KYB PN EKYB250ELL392MK40S) caps on the 6 positions per board, again depending on voltage required. The TP transformers work well here, but I generally use Hammond 229D-series transformers for these. I also use the same rectifier diodes as supplied by K&K in their kit (sorry, you'll have to buy one to find out what it is... I won't give away K&K's IP) and as recommended by TP for higher current applications, generally use a much smaller R in the C-R-C filter or jumper it out completely. There's another trick that K&K uses in their supply that I suspect is well worthwhile, again, you'll have to buy one to get it. And sorry, you are on your own on figuring out the mods needed to use the LT3080/LT3083 on the TP boards, it is simple and evident from reading the various datasheets, you should learn at least that level of understanding to build and use these anyway.
I use these units with LT3083's and setup for 5v to power Pi's, BBB's, and my Zotac ZBOX-PI320-W2 LMS machine.
The other had to do with questions about the MamboBerry and 384 playback... first, I've attached a closeup of the clock and there is no Mhz indication on the unit. Based on the requirements of the ES9018, I would expect a clock faster than 80Mhz, perhaps as high as 100Mhz needed for the unit to successfully play 384. OTOH, the datasheet states a max of 50Mhz.
Second, I've attached a picture of the bottom of the MamboBerry... draw what conclusions you want from what is printed there.
My own opinion is that it is likely limited to 192 based on my read of the ES9023 datasheet. OTOH, the ESS datasheets continue to be the sketchiest in the business AND I've seen people going beyond the recommended limits on ESS (and other) silicon successfully in the past, with ES9018's overclocked beyond the recommended 100Mhz ASRC clock with some success. AND since the ES9023 core is based on that of the ES9018 and other sister chips, it may be that it can go higher. AND I have no reason to distrust what the MamboBerry people say. Myself, I haven't tried them above 96 in my setup, I don't have any music above that available on my Pi Players AND don't have SOX upsampling running on the server at this time.
Still, it is a very sweet sounding unit for the $ and is what I consider a minimal-level unit for good sound (with two good power supplies (at least equivalent to the above... and even though I haven't tried them myself, my take on comments about them is that the iFi units are likely very worthwhile here) required for this result, one for the Pi and one for the DAC).
I will say my highly modified HFBD+P WITH Soundcheck's tweaks is better than the MamboBerry, but parts cost to DIY the mods puts you above the cost of a MamboBerry, it takes some VERY GOOD SMD assembly and modification skills to build up the supplies and perform the mods, AND requires a 3rd raw DC supply (more cost!) to achieve this quality. With all that, it SHOULD sound better (and that was why I was so puzzled about how it sounded before I got Soundcheck's tweaks to beta-test).
YMMV.
Greg in Mississippi
Two slow responses...
One, there's a question on what I use for linear supplies. For my low-end audio-level supplies such as the ones I used here, I use supplies based on the LT3080 or LT3083 chips… with the LT3080 good to up to 1.2a max and the LT3083 3a, both depending on the raw DC supply voltage and capability and heatsinking. I will buy and slightly modifiy K&K Audio Low Voltage Power Supply kits (here: K&K Audio | Other kits) with larger raw DC filter capacitors (KYB 12000UF 20% 16V RADIAL PN: EKYB160ELL123MM40S or 6800UF 20% 25V RADIAL PN: EKYB250ELL682ML40S depending on the raw DC voltage required) and the post-regulator filter cap with Nichicon Electrolytic Capacitors - Snap In 35volts 4700uF 20% PN: LKG1V472MESZBK. I also replace the mylar Wima bypass caps with polyprop equivalents.
Or I will buy Low-Current Dual Power Supply (LCDPS) boards from Twisted Pear Audio (here: Low Current Dual Power Supply (LCDPS) - A Dual Regulated CRC Supply) and build them up using the LT3080 or LT3083 chips depending on the current required as above. As these chips require a bit different output voltage setting network than the LM317-equivalent ones these boards are designed to use, that takes a little modification on the boards. I use either KYB 5600UF 20% 16V RADIAL (KYB PN EKYB160ELL562MK40S) or 3900UF 20% 25V RADIAL (KYB PN EKYB250ELL392MK40S) caps on the 6 positions per board, again depending on voltage required. The TP transformers work well here, but I generally use Hammond 229D-series transformers for these. I also use the same rectifier diodes as supplied by K&K in their kit (sorry, you'll have to buy one to find out what it is... I won't give away K&K's IP) and as recommended by TP for higher current applications, generally use a much smaller R in the C-R-C filter or jumper it out completely. There's another trick that K&K uses in their supply that I suspect is well worthwhile, again, you'll have to buy one to get it. And sorry, you are on your own on figuring out the mods needed to use the LT3080/LT3083 on the TP boards, it is simple and evident from reading the various datasheets, you should learn at least that level of understanding to build and use these anyway.
I use these units with LT3083's and setup for 5v to power Pi's, BBB's, and my Zotac ZBOX-PI320-W2 LMS machine.
The other had to do with questions about the MamboBerry and 384 playback... first, I've attached a closeup of the clock and there is no Mhz indication on the unit. Based on the requirements of the ES9018, I would expect a clock faster than 80Mhz, perhaps as high as 100Mhz needed for the unit to successfully play 384. OTOH, the datasheet states a max of 50Mhz.
Second, I've attached a picture of the bottom of the MamboBerry... draw what conclusions you want from what is printed there.
My own opinion is that it is likely limited to 192 based on my read of the ES9023 datasheet. OTOH, the ESS datasheets continue to be the sketchiest in the business AND I've seen people going beyond the recommended limits on ESS (and other) silicon successfully in the past, with ES9018's overclocked beyond the recommended 100Mhz ASRC clock with some success. AND since the ES9023 core is based on that of the ES9018 and other sister chips, it may be that it can go higher. AND I have no reason to distrust what the MamboBerry people say. Myself, I haven't tried them above 96 in my setup, I don't have any music above that available on my Pi Players AND don't have SOX upsampling running on the server at this time.
Still, it is a very sweet sounding unit for the $ and is what I consider a minimal-level unit for good sound (with two good power supplies (at least equivalent to the above... and even though I haven't tried them myself, my take on comments about them is that the iFi units are likely very worthwhile here) required for this result, one for the Pi and one for the DAC).
I will say my highly modified HFBD+P WITH Soundcheck's tweaks is better than the MamboBerry, but parts cost to DIY the mods puts you above the cost of a MamboBerry, it takes some VERY GOOD SMD assembly and modification skills to build up the supplies and perform the mods, AND requires a 3rd raw DC supply (more cost!) to achieve this quality. With all that, it SHOULD sound better (and that was why I was so puzzled about how it sounded before I got Soundcheck's tweaks to beta-test).
YMMV.
Greg in Mississippi
Attachments
Greg.
We're pretty much in line.
I would not recommend to go for the HifiBerry DAC+ Pro if you have good quality sound in mind.
In its current configuration it just needs too much effort to make the DAC sing properly.
And I think all of us agree that the 3.3V digital power rail design has been really a poor design choice. It's one of the major weaknesses.
They could have done much better - by implementing two or three LDOs fed by a single 5V supply rail - and that at almost no extra cost.
And then there are these intrusive filters of that PCM51xx family. Most people even run the
default setup because they don't know any better. That's really bad.
My conclusion here is to stay away from this DAC family in the future.
Bottom line.
The base device hooked up to a stock RPI SW image, IMO delivers MidFi at best.
It's possible to achieve real nice sound, but that requires quite some modification effort - on HW and SW - and then still limits my device to 48kHz.
The only reason it stays is that it sounds great after all modifications have been applied. Would I do it again? No.
Mambo:
There are reports saying that the latest Audiophonics DAC can compete with the MamboBerry. Nice developments.
I hope we'll see more upgrades or new and better DACs popping up later on this year.
@Vassili: Please confirm and let us know what you mean with 384kHz support!?!?
We're pretty much in line.
I would not recommend to go for the HifiBerry DAC+ Pro if you have good quality sound in mind.
In its current configuration it just needs too much effort to make the DAC sing properly.
And I think all of us agree that the 3.3V digital power rail design has been really a poor design choice. It's one of the major weaknesses.
They could have done much better - by implementing two or three LDOs fed by a single 5V supply rail - and that at almost no extra cost.
And then there are these intrusive filters of that PCM51xx family. Most people even run the
default setup because they don't know any better. That's really bad.
My conclusion here is to stay away from this DAC family in the future.
Bottom line.
The base device hooked up to a stock RPI SW image, IMO delivers MidFi at best.
It's possible to achieve real nice sound, but that requires quite some modification effort - on HW and SW - and then still limits my device to 48kHz.
The only reason it stays is that it sounds great after all modifications have been applied. Would I do it again? No.
Mambo:
There are reports saying that the latest Audiophonics DAC can compete with the MamboBerry. Nice developments.
I hope we'll see more upgrades or new and better DACs popping up later on this year.
@Vassili: Please confirm and let us know what you mean with 384kHz support!?!?
Last edited:
Perhaps you should try and put this point to Daniel? Maybe you can convince him to make a "DAC+ Audiophile" version, incorporating your design suggestions.
I have to agree. I have both IQaudIO and Mamboberry. If there are things that´s easy to implement I think there´s definitely a market for it. I really like the way Vassili has gone a step further an implemented several good thing to make a better solution from the start. I wouldn't have any problems paying even more to make it even better.
The ES9023 does indeed unofficially support 384KHz sampling rates. By that, I mean that sound actually comes out and it sounds fine..
However, I do not have measurements to confirm that it is indeed functioning properly (and not messing up sound-wise or something).
But be warned, that not every chip does. I have some very early samples of ES9023 that either do not output sound @384k or it is distorted.
That's the reason, with my es9023 codec patch set, I made 384k a dt config opt-in, rather than an opt-out and the default.
i had the audiophonics tcxo v2 and mamboberry is better. i would like to know if audiophonics v3 would be iqual or better than mambo.
i saw that reviews but for me they are marketing...To begin with:
There is pretty humble dutch guy who calls himself an "Audio Prophet" .
You'll find him and his HAT DAC reviews on Youtube.
I am thinking about feeding the digital part of the Hifiberry seperately, not with a linear power supply, but with a good quality lifepo4 battery. Is feeding the digital part with a lifepo4 battery a good plan of is it far better, one day, to buy a linear power supply? I hope someone can answer this question!
LiFePo4 should work. I wanted to try it myself, since I have plenty of A123s around. I left it alone because of the charging hazzle,
If you stay within the voltage limits give a try.
I also think also the analog part could very well be fed with a battery.
Most projects I came across did quite well with LiFePo4s.
If you stay within the voltage limits give a try.
I also think also the analog part could very well be fed with a battery.
Most projects I came across did quite well with LiFePo4s.
Hi guys,
Sorry to be off topic again but related to the question above, I can give you some comments here. I wrote earlier that I'm really happy with my Mamboberry. Nevertheless I build a new system for my son's birthday - complete setup from Audiophonics : I-Sabre V3 Dac, related housing, Tripath T-Amp and double linear power supply able to feed 12V to amp and 6V to DAC. I choose this time these components to save some time. I didn't want to build again power supply, find housings etc...
https://drive.google.com/file/d/0B3AonlTstB87bkRjdF9fVF9NcWc/view?usp=sharing
https://drive.google.com/open?id=0B3AonlTstB87bUxZd05PZktrakk
Before my son takes the system away I took the opportunity to compare it with my Mamboberry powered by the 6VAC torroidal transformer.
I spent the whole week turning back and forth the source selector and tried to evaluate and compare with a lot of 24/192 files (the one that gives me goosebumps).
At first I noticed a clear advantage to I-Sabre V3. But later it turns out I was misled by a mistake. Mambo had Moode 2.6 beta4 and V3 had the later Moode 2.6 stable. After upgrading the Mambo with Moode 2.6 stable, I was not able any longer to notice a difference. Both system, well fed with linear power supplies have an equivalent amazing sound. And for approx. same price (DAC, power suppy, housings). The 2 are really great. Audiophonics was quicker to set up with the ready to use double linear power supply.
And also I am a bit surprised how improvement in sound I noticed from one software release to the other (I doublechecked to have on both the same parameters).
Last edited:
I have to agree on the Mambo. I really love my Mambo but I would easily pay more to make every detail the best.
I saw that Audiophonics has a new project on Kickstarter:
https://www.kickstarter.com/project...io-player-raspberry-pi-3-sabre?token=acd1d854
The player is probably not for me but they seem to have a new DAC on it´s way:
ES9018K2M :First ESS ES9018K2M RPI DAC with differential I/V converter and output buffer.
Now available for the Raspberry Pi.
In addition to this DAC we will use a power controller including linear regulation for stable and quiet supply.
I saw that Audiophonics has a new project on Kickstarter:
https://www.kickstarter.com/project...io-player-raspberry-pi-3-sabre?token=acd1d854
The player is probably not for me but they seem to have a new DAC on it´s way:
ES9018K2M :First ESS ES9018K2M RPI DAC with differential I/V converter and output buffer.
Now available for the Raspberry Pi.
In addition to this DAC we will use a power controller including linear regulation for stable and quiet supply.
Last edited:
Remains to be seen whether that player (or a standalone DAC board) actually goes on sale, or at least the 9018 variant of it. I'm not sure Audiophonics has communicated the fact that they will need to release a driver to operate this DAC in software (I2C) mode, to ESS.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.