Hifiberry DAC+ Pro - HW mods anybody?

Did you guys see this in the newest moode?

I've not bothered to play with it because the volume is so screwed up using the Dac+ Pro, can't verify it's set to 0db, and out of the box it's set to a millionDB & is all distorted :D

Hi,

Help me understand what's not working correctly wrt volume and I'll fix :)

Background:

(1) OOB Moode settings for Analog and Analog Playback Boost are what Hifiberry recommended to me.

(2) Assuming MPD volume set to "Software", ALSA volume automatically set to 100 when setting up any DAC that has a hardware volume controller. In the case of Hifiberry DAC+ and Pro this setting shows +24dB in amixer!

To compensate for that that and get close to 0dB, Moode Customize popup settings should be:

Logarithmic curve = "Yes"
Maximum volume % = 81

This results in amixer showing 0dB when Moode volume is 100

Regards,
Tim
 
1. Not any optimization applied ( Or better: I couldn't find any)
2. Tiny/MicroCore Linux is as inflexible and complex as a Linux can get. That small size comes at a price and is not at all required.
3. Usually you won't find the most up2date SW. If you find certain SW at all.
4. Tweaking that OS is PITA

For those who're looking for a quick squeeze solution. Go for it.

Those who look for optimized SW that translates into better sound, look for a different solution. You might want to talk to Tim Curtis. Perhaps he's willing to provide a squeezelite option for Moode. Should be a No-Brainer for him.

Hi,

Not familiar with Squeezelite but after reading up on it, if I understand correctly:

(1) its a playback client thats controlled via a WebUI provided by LMS
(2) interfaces directly to ALSA like MPD
(3) probably would need a configuration UI in Moode

So, it would coexist with MPD on Moode similar to shairport-sync, but unlike upmpdci which integrates into MPD playlist.

??

-Tim
 
@Tim

1.
Yep. squeezelite is really micro-slim.
All it needs is a systemd startscript plus config file and a web config dialog - which would be your contribution.
I can send you my .service and squeezelite.conf if you like. Just PM me. Though making it up is a piece of cake.

I use these sources.
This way squeezelite recognizes samplerate and bitdepth from native PCM streams.
I really like that.

Of course you'd face a competing situation between MPD and squeezelite on the audio interface if they'd run side by side. It makes sense to have only one app running at a time.


Just to mention it. Triode, the brilliant man behind squeezelite and SOA and much more stuff, disappeared a year ago from the surface for unknown reasons.


2.
Use the current 4.4 kernel sources. This fixes the 81% mess. 81% becomes 100%. No more annoying workarounds or messups.

Cheers
 
Last edited:
@Tim

1.
Yep. squeezelite is really micro-slim.
All it needs is a systemd startscript plus config file and a web config dialog - which would be your contribution.
I can send you my .service and squeezelite.conf if you like. Just PM me. Though making it up is a piece of cake.

I use these sources.
This way squeezelite recognizes samplerate and bitdepth from native PCM streams.
I really like that.

Of course you'd face a competing situation between MPD and squeezelite on the audio interface if they'd run side by side. It makes sense to have only one app running at a time.


Just to mention it. Triode, the brilliant man behind squeezelite and SOA and much more stuff, disappeared a year ago from the surface for unknown reasons.


2.
Use the current 4.4 kernel sources. This fixes the 81% mess. 81% becomes 100%. No more annoying workarounds or messups.

Cheers


Hi,

#1 is do-able. Does Squeezelite support hooks for running script before play starts and after play ends (like hooks in shairport-sync)? This would make it easy to coordinate with MPD.

Triode and NwAvGuy (ODAC), two Internet mysteries...

#2 Well, what exactly is the problem in the kernel that causes volume issue with Hifiberry devices and not other PCM5122 based devices for example IQaudio Pi-DAC+?

When I look at hardware volume range and 0dB values reported by amixer for Hifiberry and IQaudIO devices I see the following:

Hifiberry DAC+:
- Limits: Playback 0 - 255
- Front Left: Playback 207 [81%] [0.00dB] [on]

IQaudIO Pi-DAC+
- Limits: Playback 0 - 207
Front Left: Playback 207 [100%] [0.00dB] [on]

If I bump ALSA volume to 100% for Hifiberry device, gain = +24dB.

- Front Left: Playback 255 [100%] [24.00dB] [on]

My interpretation of this difference in 0dB points for boards based on exact same PCM5122 chip is that Hifiberry driver is reporting wrong hardware volume range to ALSA. Something like that.

Regards,
Tim
 
Anybody tried RPI with DAC+ Pro (or others) as I2S source?

I am very eager to try this, but if there are negative user experiences, than it probably is
wasted time and money,,,,

Regards,
Alex

I bought it for exactly that purpose, I hooked it up to my Pedja Rogic AYA DAC and it works very well with Tim's excellent MoOde player. The only comparison I can make is with the CDPro CD mech I had been using with the DAC before (via I2S) and its certainly close in terms of fidelity. I have not split the supplies or made any mods at all but I run the Pi off a reasonable linear PSU. Well worth the modest outlay IMO.

Now a Raspberry Pi I2S cape board with isolation and reclocking, that would be something... I'm waiting and watching. Perhaps we should pester the twisted Pear boys :D
 
Last edited:
@Tim.

The DAC offers a digital VC range up2 +24db.

IQAudio runs a different driver.

With the new kernels/updated drivers the HifiBerry gets in line with the IQAudio and sets the limit to 0db by default.

Cheers

Hi,

Ok I'll take your word for it :) Sure will make things simpler for Hifiberry devices.

Btw, which mainline kernel ver will have these updates?

-Tim
 
Having listened to this Hifiberry dac+ Pro for the past week, I must say I am impressed with the soundquality it delivers for what it costs.

Upgrading the power supply on the analog side of the board is what I will do.
Looking at some extenders (risers) available for the GPIO connector got me thinking to use one of those, and supply power from there and leave the dac intact.

Should be easy to do for 5 V en 3,3 V inputs.
Other ideas for external power?

Ronnie
 
piZERO

Just some info on my piZERO:

I soldered only the needed pins for the DAC: power, grounds, i2s, EEPROM

Use piCorePlayer with large 420MB buffer (goal is no network activity most of the time, just the 2 sec burst at the start)-- I like that the RAM chip is stacked on the CPU. On the pi2,3 the RAM chip is across the board underneath (possible interference from the traces)

HDMI disabled.

Then of course dual linear power thru the DAC: separate analog, pi (thru gpio)
 

Attachments

  • IMG_0135.JPG
    IMG_0135.JPG
    368.1 KB · Views: 563
Thanks all for sharing your mods. Interested in picking one up myself

Btw: Did you see the picture of the Bryston version that is using a modified hifiberry board. Mods explained here at a high level

Interesting and smart choice by Bryston! They are using the Digi+

To me the Digi+ is near perfect after doing this ( no digititis/hash-- that's what I call it :D ) --> https://www.youtube.com/watch?v=T5C5oSs5ELk

And the Dac+ is surprising! Use it as an additional source to my Digi+/DL3
 
Last edited:
Member
Joined 2003
Paid Member
@Soundcheck,

Great timing! After some time breaking-in and warming up the last couple of weeks (not continuous, but over a week of time total), I listed to the HFBD+P Saturday and Sunday nights and the Mamboberry last night.

First, on the Mamboberry... yup, I still hear the sound of an ESS DAC running in ASRC mode. I don't consider it objectionable, but it does not provide the absolute best in fidelity... I find it hard to describe, but it is not quite as 'direct' as the synchronously-operated PCM1795 in my modified Sony HAP Z1-ES or even the PCM5122 in the HFBD+P. Even with that, it was a good listening experience last night and I plan to spend a little more time with it tonight (to try some additional recordings). Smooth, fairly dynamic, with a very rich and engaging midrange tone, very evident on Rickie Lee Jones' voice on her Pirates and on the woodwinds on Paul McCandless Premonition. My listening session was short, but enjoyable. While not beating that system's baseline reference of my modified Panasonic S47 DVD player overall, that midrange tone is to die for and shows up both the S47 and the HAPZ a touch. I rate it as very competent, as long as one is not terminally put off by the ESS DAC running ASRC sound. Also I rate it as a much better bargain at a bit more than 2x the price of the HFBD+P stock. AND I am running the MamboBerry mostly stock, with my only mods a connection to feed one of my raw 6.5v DC rails directly into the onboard power supply, bypassing the rectifying diodes that would be used with an AC feed from a transformer, and to feed 5v separately into the R-Pi, with a local 560uf reservoir cap at that input.

OTOH, I am very conflicted about the HFBD+P. Technically, my mods improved the dynamics, articulation, and detailing significantly. Dynamically, it is closer to the modified S47 than the MamboBerry (and the MamboBerry is not bad, but the S47 has a sense of impact on drums, other percussion, and bass that is beyond that of the MB while the HFBD+P is close). Highs are problematic... clean, but I still get a sense of 'digitalitis' after 10 minutes with it that I don't get with any other digital source I am currently using (an aside... I describe the sense of 'bad' digital as 'digitalitis', where I get a sense of un-ease and a bit of a sinus headache when I listen to a source that does this for me. I got it a lot with early CD playback and didn't really have a source that was mostly free of it until I did a mongo mod of a classic Magnavox CDB650 back in the early '90s based on the Audio Amateur Pooging article, but with massive separate power supplies and the DAC / output stages in a separate enclosure).

Also, the tonal balance comes off as a bit thin or recessed through the midrange. Very much the opposite of the MamboBerry, which might be a touch on the rich side of neutral.

The tonal balance does not seem wildly different than the stock unit to me, but with the improvements in dynamics and bass extension and articulation, I expect more from the unit and am disappointed. To paraphrase Snape in the Harry Potter movie, 'Mods aren't everything!'

I've been pondering a lot why this mod has not been more successful to-date. I'll share some thoughts in a later post along with my planned next steps.

In the meantime, I plan to enjoy the MamboBerry a bit more tonight, then put the Soekris back in and in parallel, cook the HFBD+P a bit more while I work on getting my next set of digital experiments running, two TP Buffalo-II DAC setups, one fed via an Acko S03 Synchronous Reclocker/Isolator from a BeagleBoneBlack and the other fed via an Ian FIFO Asynchronous Reclocker/Isolator from an R-Pi. Both of these setups will allow me to run the ESS DACs in synchronous mode so I can get a better sense of how these DACs sound without the ASRC running.

More later this week on the HFBD+P.

Greg in Mississippi

P.S. Although the MamboBerry is not up to the level of my modified S47 DVD player in overall performance, I am considering it my new baseline for that system... it is that good! And that DVD Player plus mods would cost roughly $700-$900 in current dollars, while the Mamboberry plus R-Pi is less than $200... pretty impressive.
 
Last edited:
Thx for the update.

Interesting that you run the MB with two different supplies.
I do think it's not such a bad idea if you run equally good supplies on both ends.

However. The MB folks do suggest to run one clean supply only.
I can understand that too.
1. Running just a standard supply on the PI as I2S source might not be the best idea.
Even if you'd feed the MB board with quality power.
2. You might introduce nasty groundloops by using a 2nd supply.

Anyhow.
I'm tempted to try a MamboBerry - once more. I had to send my testing device
back though. This time I'd need to order one myself.
I'm not at all happy with my steep filter setup on the HifiBerry which allows me to run 48khz max.
On the other side I can't really listen to any other filter. Catch22.

####

2 Qs.

1. Do you have the alsamixer settings of the HifiBerry properly under control ??
The default settings are all but OK. IMO.
2. Have you been using the integer mode driver we've been discussing earlier with the
MamboBerry ??

######

What happened over here.

I've been fiddling around with my SW setup - a never ending story.

I now introduced a realtime kernel on my PI2. I've been working with rt kernels before.

It took me awhile to get it stable. I ran into PI lockups all the time.
That's been a funny effect. Randomly sometimes after minutes, sometimes after 30 minutes... suddenly 2-3 seconds of a track were repeated continuously and everything else locked up. A hardreset was required.
My wife said "...sounds like a jumping needle on a turntable - good old times ..." "... please -- turn that mess off!"

Ok. You grow with your challenges. I made it work. Luckily I was not the only one who ran into rt-kernel lockups.

The whole project IMO was worth the effort though.
The rt-kernel is giving me an extraordinary clean low-end added to a very clean overall presentation.
 
Last edited:
Both of these setups will allow me to run the ESS DACs in synchronous mode so I can get a better sense of how these DACs sound without the ASRC running.

Greg, keep an eye out for the public launch of the Takazine SabreBerry32 Pi DAC. Not on public sale yet, but IMHO this one is worthy of your time if you want to tinker with hardware mods on a Pi DAC board. And with the 2x NDK's, (45/49MHz), you have SYNC with it as BCLK/FRAME master, which it will be by default unless you use the dtoverlay slave param.
 
Last edited:
The layout and chosen parts of that SabreBerry does not look very convincing to me.
Then I can't see an output stage. The Sabre is probably driven in voltage-out mode,
which has never been a good idea. The clocks are sitting on the opposite side of the board.
Asf.

A well implemented Sabre 9023 doesn't necessarily need to be worse then a rather compromised 9018 implementation I'd guess.

Let's see how things develop.