@Souncheck,
LOL, I wasn't expecting you to, modifying a unit beyond any ability to recognize it is my stock in trade!
I'm in the mode of trying to learn how good I can make a piece of gear. Once you pointed the way on the HFBD+P with your mods, I was just looking at how far I wanted to take it... and musing on what I'd like to see from HFB (or anyone else) for a future, VG R-Pi DAC hat.
BTW, I poked around on mine last night and was able to confirm that two circuit board trace/land cuts will separate the DAC Digital/Clock 3.3v from the R-Pi 3.3v.
One question here, feeding the R-Pi with 3.3v from the appropriate pins on the header works ok and won't mess up the on-board 5v->3.3v regulator?
My impression of the Soekris DAM is that it still requires some work to make it sound its best, even in V3 form. I consider the on-board power supply (AC In->diode bridge->filter caps) to be inadequate along with the output stages (although not everyone will need them). Also the shift register supplies, while improved from V1 & V2 to a now adequate level, still could use some additional help. Of course, these are my personal views... AND I intend to take mine quite a bit farther.
Still, it is pretty good in stock form based on listening to one of mine in a system AND based on my quick listens using my cell-phone headset to check operation of my other one. Comparing it to the HFBD+P running at the same time, it is likely a significant step up, but I need to put them both in a system to confirm. What I'd personally recommend as a minimum implementation on the DAM puts it up close to the $500 USD DIY cost level.
Reading some of the current thinking about digital filtering (which you can find on the Logitech, DIYAudio and CA forums along with a little in the Bottlehead forum) strongly supports the hypothesis that the filters imbedded in most modern DAC chips are definitely not the sonically best available. I'm closely watching some of the commercial units with alternative filtering (PS Audio, Bottlehead, Schitt, among others) and some of the DIY efforts (HD-Player to various NAA's and to direct-to-DSD units and even the DAM DAC, where most preferred with some of the alternative filters available by the user community).
One thing I found illuminating is to understand that the DAC chips use different filters depending on the sample rate of the input digital stream... higher rate inputs required less 'invasive' filtering. On the PCM51xx series of chips, AFAIK, they do not use their internal filtering if fed a 352/384 sample rate data stream, which makes it pretty simple to bypass them (with source upsampling). One of the things I have on my list to check with the R-Pi sources is whether I can get 352/384 SOX-upsampling to work ok. Not sure if the HW or drivers will handle, though.
Thx. I will have to get one of these to play around with.
@Bas Horneman,
Thanks for this. I was hoping you also had some information from Durio that might have been available during the Kickstarter, but is not on their website today. Alas.
As abraxalito mentioned above, "Since the DAC chip has CMOS opamps for output stages I'd suggest a high input impedance buffer (darlington transistors work fine) biassed into classA with CCS loads." Durio-styled paralleling should improve the ability of the unit to drive low-ish impedence loads, which I'd expect to be audible mainly as increased dynamics. My gut feeling is similar to what you stated, that the lower distortion from paralleling (as done on the DDDACs) will not manifest as much on voltage-output DACs, but I could be wrong here.
Greg in Mississippi
They were BSing. Paralleling DACs with on-chip opamps certainly can improve the output drive and hence the dynamics provided output levels are fairly closely matched. But 3dB improvement in SNR for each DAC added in parallel? Complete nonsense.
Yesterday I attached L/R channels right to the chip output, passing the output filter by.
I had that tweak in mind from the very beginning and couldn't let it go, even though I was more then happy with the sound a night before.
I attached wires above R3 + R6 first. I thought it's worth it and a couple of hours later removed the filters completely. (C7,C8,R3,R6)
While I was at it I also removed the RCA jacks. It makes work easier and I don't have any antennas on the board. I also cut most parts of the R/L traces. The layout looked like a RF antenna to me.
While analyzing the R/L routing, I figured how weird the routing is. I think it's done really bad.
In the early phase of poking and testing I didn't realize how weird it really was and without realizing shorted R/L by attaching on-the-fly wires the wrong way...
...and connected that pretty casual to my beloved speakers.
Oh man. What a nasty effect. I thought I'd killed my speakers.
I figured the problem out short after. Neither the DAC nor the amp nor speakers were affected by that short. Lucky me.
######
So far the good. I could leave it at this point.
The last thing that comes to mind (as lightweight mod) is adding some caps in front of the clocks to somewhat stabilize the situation over there. Any suggestions, beside hooking them up to a quality supply?
######MamboBerry###################
I also had a chance now to checkout the Mamboberry. The first thing I noticed is that it uses the old 16bit HifiBerry driver. To me a 100% NoGo.
I could have stopped at this point. But I thought give it a chance.
To figure out how to get the thing going took me while.
The 16 bit limitation caused this or that issue in my setup.
I think it is not a good idea to buy products that 100% rely on other peoples proprietary software.
It should have been very simple to introduce 24 bit support otherwise.
The HW looks really good though. Well made. Many of the mods we're discussing here are already implemented.
I made it work and hooked it up to my desktop Adam A5x. As PS I was using the iPower straight ( including my flashy brandnew adapter ) into the PI2.
I wasn't motivated to hook up an external supply, even though that's highly recommend by the MamboBerry folks.
Bottom line. That DAC sounds not bad at all. That was kind of expected.
I think MamboBerry claims to have the best sounding PI-DAC. (stock Pi-DACs ).
Still -- I wouldn't keep or buy it.
First there's the driver issue and 2nd the clocking scheme.
I do prefer the DAC+ Pro master-clock setup and slaving the PI. Beside that the DAC+ Pro comes with a quality I2S output which makes the DAC more flexible.
The clocking scheme weakness is not exclusive to MamboBerry though, it basically excludes all other PI-DACs from my list as well.
Enjoy.
I had that tweak in mind from the very beginning and couldn't let it go, even though I was more then happy with the sound a night before.
I attached wires above R3 + R6 first. I thought it's worth it and a couple of hours later removed the filters completely. (C7,C8,R3,R6)
While I was at it I also removed the RCA jacks. It makes work easier and I don't have any antennas on the board. I also cut most parts of the R/L traces. The layout looked like a RF antenna to me.
While analyzing the R/L routing, I figured how weird the routing is. I think it's done really bad.
In the early phase of poking and testing I didn't realize how weird it really was and without realizing shorted R/L by attaching on-the-fly wires the wrong way...
...and connected that pretty casual to my beloved speakers.
Oh man. What a nasty effect. I thought I'd killed my speakers.
I figured the problem out short after. Neither the DAC nor the amp nor speakers were affected by that short. Lucky me.
######
So far the good. I could leave it at this point.
The last thing that comes to mind (as lightweight mod) is adding some caps in front of the clocks to somewhat stabilize the situation over there. Any suggestions, beside hooking them up to a quality supply?
######MamboBerry###################
I also had a chance now to checkout the Mamboberry. The first thing I noticed is that it uses the old 16bit HifiBerry driver. To me a 100% NoGo.
I could have stopped at this point. But I thought give it a chance.
To figure out how to get the thing going took me while.
The 16 bit limitation caused this or that issue in my setup.
I think it is not a good idea to buy products that 100% rely on other peoples proprietary software.
It should have been very simple to introduce 24 bit support otherwise.
The HW looks really good though. Well made. Many of the mods we're discussing here are already implemented.
I made it work and hooked it up to my desktop Adam A5x. As PS I was using the iPower straight ( including my flashy brandnew adapter
) into the PI2. I wasn't motivated to hook up an external supply, even though that's highly recommend by the MamboBerry folks.
Bottom line. That DAC sounds not bad at all. That was kind of expected.
I think MamboBerry claims to have the best sounding PI-DAC. (stock Pi-DACs
).Still -- I wouldn't keep or buy it.
First there's the driver issue and 2nd the clocking scheme.
I do prefer the DAC+ Pro master-clock setup and slaving the PI. Beside that the DAC+ Pro comes with a quality I2S output which makes the DAC more flexible.
The clocking scheme weakness is not exclusive to MamboBerry though, it basically excludes all other PI-DACs from my list as well.
Enjoy.
I'd like to add.
A slightly modified MamboBerry would IMO be well worth a consideration.
"slighthy modified" would mean:
An external 5V supply option would be a nice to have , that makes it more generic
instead of using 6V AC/9V DC.
The clocking scheme can be improved -- I mentioned that. That might be more complex for the Sabre though. I don't now that.
That clock thing goes in line with a driver update. They might not be able to use the DAC+ driver though. That'll make it more complex for them.
Their webpage and documentation is IMO a mess. Some small startup companies forget that this stuff is a key success factor.
And pricewise at today €90 they are IMO too expensive in comparison to the DAC+ Pro and others, even if I count some extra Euros on top of the DAC+ Pro for the additional caps in. But. Lets not talk about +/- 10€/$ here and there.
Though that's what makes the HifiBerry a success. Rather low price, reasonable quality (workmanship and sound) , high volumes.
A slightly modified MamboBerry would IMO be well worth a consideration.
"slighthy modified" would mean:
An external 5V supply option would be a nice to have , that makes it more generic
instead of using 6V AC/9V DC.
The clocking scheme can be improved -- I mentioned that. That might be more complex for the Sabre though. I don't now that.
That clock thing goes in line with a driver update. They might not be able to use the DAC+ driver though. That'll make it more complex for them.
Their webpage and documentation is IMO a mess. Some small startup companies forget that this stuff is a key success factor.
And pricewise at today €90 they are IMO too expensive in comparison to the DAC+ Pro and others, even if I count some extra Euros on top of the DAC+ Pro for the additional caps in. But. Lets not talk about +/- 10€/$ here and there.
Though that's what makes the HifiBerry a success. Rather low price, reasonable quality (workmanship and sound) , high volumes.
@Soundcheck,
Very good information. THANKS!
And I'm very glad you didn't damage your speakers, amp, or the DAC. I did kill a pair of Spica TC-50's and a Counterpoint SA-12 with an oscillating opamp 20 years ago... and now I check operation with a meter and a pair of cheap headphones first! (and sometimes scope, but I'm often too impatient!).
Some comments...
On the clock power I was planning to replace the local bypass caps (C9 & C13 I believe) with PPS SMD caps and piggyback a couple of Oscons across them. Short of a separate 3.3v rail for the digital & clocks, that is about the best one can do, AFAIK.
And on the output filter, my first set of mods to the board will include replacing at least the filter caps with PPS film SMD caps... I'd NEVER use ceramics for signal filtering, I don't care HOW much better they are today than in the past!
On the Mamboberry DAC, thanks for the report & info. I know mine is on the way, but it'll be some weeks still. I've been looking at the pictures to make educated guesses on what I'd like to try with it.
I've also had several runs of email conversations with them asking about the DAC and making some suggestions to them, largely about the clocking scheme. There are two things that could help in that realm... first, the source for the HiFiBerry driver is available, so that could be a good starting point. Second, synchronous operation of the ES9023 chips is pretty easy, just feed it an appropriate clock signal for the datastream (22/24 or 44/49) and it runs synchronously, same as the up-market ESS chips.
I do like that the Mamboberry uses some of the really good TI TPS7A47 & TPS7A49-series regulators. I am surprised they put such a low limit on the input voltage (either DC or AC) as these are fairly wide-input-range chips (36v!). Thermal concerns, I'm guessing.
On it using the 16bit HiFiBerry driver, it will likely work ok with other DAC drivers that are just feeding I2S out of the Pi... maybe try the IQAudio driver?
Then to defend the Mamboberry a bit, the 6vac/9vdc setup is what you get IF you want to do a good power supply that feeds both the Pi & the DAC built-in on the DAC Hat. They have what I would suggest is the simplist and cheapest way to do a fully linear-supplied Pi/DAC Hat combo.
And of course, the associated diodes (I bet they are good fast-recovery or Schotky) and caps drive up the cost a bit... along with two regulator chips, one costing $5USD and the other $3USD versus the single $1USD chip in the HFBD+P. Those higher cost components (very clear in the pictures I've got) along with what I suspect is a much smaller run quantity compared to HFB causes a lot of the cost differences, I suspect. Not that I wouldn't want to pay less for it, but I can see what drives at least some of the additional cost.
Finally, I continue to agree with your assessment of really liking the HFBD+P's clocking arrangements, feeding a bitclock into the Pi to use as the reference clock is the way to go for best sound. But as you've shown, power is important too... I really wish someone would include both!
Greg in Mississippi
P.S. @Abraxalito, thanks for your comments on the voltage-out DAC paralleling, what they said didn't add up to me, but I admit I am not the most learned in this realm.
P.P.S. I just noticed the DurioSound DACs use the TPS7A47 series regulators too!
Very good information. THANKS!
And I'm very glad you didn't damage your speakers, amp, or the DAC. I did kill a pair of Spica TC-50's and a Counterpoint SA-12 with an oscillating opamp 20 years ago... and now I check operation with a meter and a pair of cheap headphones first! (and sometimes scope, but I'm often too impatient!).
Some comments...
On the clock power I was planning to replace the local bypass caps (C9 & C13 I believe) with PPS SMD caps and piggyback a couple of Oscons across them. Short of a separate 3.3v rail for the digital & clocks, that is about the best one can do, AFAIK.
And on the output filter, my first set of mods to the board will include replacing at least the filter caps with PPS film SMD caps... I'd NEVER use ceramics for signal filtering, I don't care HOW much better they are today than in the past!
On the Mamboberry DAC, thanks for the report & info. I know mine is on the way, but it'll be some weeks still. I've been looking at the pictures to make educated guesses on what I'd like to try with it.
I've also had several runs of email conversations with them asking about the DAC and making some suggestions to them, largely about the clocking scheme. There are two things that could help in that realm... first, the source for the HiFiBerry driver is available, so that could be a good starting point. Second, synchronous operation of the ES9023 chips is pretty easy, just feed it an appropriate clock signal for the datastream (22/24 or 44/49) and it runs synchronously, same as the up-market ESS chips.
I do like that the Mamboberry uses some of the really good TI TPS7A47 & TPS7A49-series regulators. I am surprised they put such a low limit on the input voltage (either DC or AC) as these are fairly wide-input-range chips (36v!). Thermal concerns, I'm guessing.
On it using the 16bit HiFiBerry driver, it will likely work ok with other DAC drivers that are just feeding I2S out of the Pi... maybe try the IQAudio driver?
Then to defend the Mamboberry a bit, the 6vac/9vdc setup is what you get IF you want to do a good power supply that feeds both the Pi & the DAC built-in on the DAC Hat. They have what I would suggest is the simplist and cheapest way to do a fully linear-supplied Pi/DAC Hat combo.
And of course, the associated diodes (I bet they are good fast-recovery or Schotky) and caps drive up the cost a bit... along with two regulator chips, one costing $5USD and the other $3USD versus the single $1USD chip in the HFBD+P. Those higher cost components (very clear in the pictures I've got) along with what I suspect is a much smaller run quantity compared to HFB causes a lot of the cost differences, I suspect. Not that I wouldn't want to pay less for it, but I can see what drives at least some of the additional cost.
Finally, I continue to agree with your assessment of really liking the HFBD+P's clocking arrangements, feeding a bitclock into the Pi to use as the reference clock is the way to go for best sound. But as you've shown, power is important too... I really wish someone would include both!
Greg in Mississippi
P.S. @Abraxalito, thanks for your comments on the voltage-out DAC paralleling, what they said didn't add up to me, but I admit I am not the most learned in this realm.
P.P.S. I just noticed the DurioSound DACs use the TPS7A47 series regulators too!
Last edited:
You're welcome! Anyone claiming they're most learned in the field of what sounds best should be humoured, its a huge field to understand.
Seems to me something of a numbers game is being played with DAC SNR. Given that 99.9% or more of music is RBCD (i.e. 16bit) the recording is going to dominate the SNR - for 16bits its going to be -93dB at the very best due to dither and very few recordings get down that low. So to be talking about DAC numbers below -96dB doesn't make any sense to me, except where digital volume controls are used. Having a lower noise DAC is useful then.
I don't actually have one, but AFAIK, the Mamboberry is just another ES9023 based HAT board with a couple of "premium" passive components.... Which driver are you using that has a 16 bit limitation? The original hifiberry_dac/pcm5102a codec combo (dtoverlay=hifiberry-dac) which most people use with the ES9023 HAT's doesn't, IIRC. 16/24/32 bit....
But then you must be talking about using some other driver, I guess. The reference to "proprietary"....
@clivem
I used the hifiberry-dac module which comes with the stock ARCH 4.1 kernel. At least it worked.
I then compiled the 4.4 git kernel. That one wouldn't work at all with the MamboBerry. I mean, the module was loaded, the DAC was seen by Alsa, but caused very strange static noise when starting playback.
That's what I mean with proprietary. HifiBerry won't test their drivers/modules against all those PiggyBack manufacturers. Obviously it can't be in their interest to shape their
drivers to push other peoples businesses.
If they introduce changes to the code, other DACs might be offline with the next kernel update. That's the risk behind buying etc. MamboBerrys.
I used the hifiberry-dac module which comes with the stock ARCH 4.1 kernel. At least it worked.
I then compiled the 4.4 git kernel. That one wouldn't work at all with the MamboBerry. I mean, the module was loaded, the DAC was seen by Alsa, but caused very strange static noise when starting playback.
That's what I mean with proprietary. HifiBerry won't test their drivers/modules against all those PiggyBack manufacturers. Obviously it can't be in their interest to shape their
drivers to push other peoples businesses.
If they introduce changes to the code, other DACs might be offline with the next kernel update. That's the risk behind buying etc. MamboBerrys.
Last edited:
In a word, no. That only affects the (dtoverlay=hifiberry-dacplus) hifiberry_dacplus/pcm512x driver combo.
The strange thing for me, is that I have 5 or so (other brands, not MamboBerry) ES9023 DAC's working right now with the rpi-4.4.y (dtoverlay=hifiberry-dac) hifiberry_dac/pcm5102a codec combo.
Code:
[08:16:52.045455] log_formats:321 supported formats: S16_LE, S24_LE, S32_LE
Last edited:
I looked again into the MamboBerry bitdepth issue.
What's not available in the driver (ARCH stock 4.1 kernel) is S24_3LE!
Basically the 24bit 3 byte standard format of audio files.
If you run that format against the DAC it'll fail. That's what I ran into and been referring to.
Basically the audio app needs to convert to S32_LE to get the 24bit audio going.
OR
16 and 32 bit test files work!
###
I'll try again the 4.4 kernel later on. Perhaps I can figure out what's wrong with that in my setup.
What's not available in the driver (ARCH stock 4.1 kernel) is S24_3LE!
Basically the 24bit 3 byte standard format of audio files.
If you run that format against the DAC it'll fail. That's what I ran into and been referring to.
Basically the audio app needs to convert to S32_LE to get the 24bit audio going.
Code:
aplay -f S24_3LE -c 2 -D hw:0,0 whitenoise24.wav
Playing WAVE 'whitenoise24.wav' : Signed 24 bit Little Endian in 3bytes, Rate 44100 Hz, Stereo
aplay: set_params:1233: Sample format non available
Available formats:
- S16_LE
- S24_LE
- S32_LEOR
Code:
aplay -f S32_LE -c 2 -r 44100 -D plughw:0,0 whitenoise24.wav
Warning: format is changed to S24_3LE
Playing WAVE 'whitenoise24.wav' : Signed 24 bit Little Endian in 3bytes, Rate 44100 Hz, Stereo
aplay: set_params:1297: Unable to install hw params:
ACCESS: RW_INTERLEAVED
FORMAT: S24_3LE
SUBFORMAT: STD
SAMPLE_BITS: 24
FRAME_BITS: 48
CHANNELS: 2
RATE: 44100
PERIOD_TIME: (125011 125012)
PERIOD_SIZE: 5513
PERIOD_BYTES: 33078
PERIODS: 4
BUFFER_TIME: (500045 500046)
BUFFER_SIZE: 22052
BUFFER_BYTES: 132312
TICK_TIME: 016 and 32 bit test files work!
###
I'll try again the 4.4 kernel later on. Perhaps I can figure out what's wrong with that in my setup.
Last edited:
Oh right, I see. I haven't looked at code, but don't think that S24_3LE is directly supported, just 4 byte S24_LE, so I get what you are saying.
Just out of interest, and I suspect would work to at least get audio playback... Have you tried "-D plughw:0,0" as the device with aplay. I know you are not going to see that as anything like a solution, as I suspect you (in particular) will want to avoid the plughw, preferring direct to the device. I'll try and find time next week to have a look at the code. It may be really simple to add support for S24_3LE on the driver side of things. Don't know. Need to look.
Just a very quick update...
I put my Squeezebox-ish setup into my big system this evening.
I'm happy to say it all works flawlessly, at least so far.
First up was the R-Pi/HiFiBerry DAC+ Pro. I power it with a fairly serious linear supply using an LT3083 & about 18,000uf of capacitance. I've been running the unit some on my bench the last week or so and after about an hour warm up, I sat down to listen.
Impression was 'not bad, better than I thought'. But I clearly heard what I describe as 'digitalitis', an edge or at least a subliminal impression that has me uncomfortable after 5-10 minutes and wanting to turn off the music. Switching back to my super-modified Panasonic DVD player was a relief and it was clearly better in all areas.
I also listened to my setup with another Pi feeding I2S to a Soekris DAM DAC (stock at this point, but fed from another set of pretty massive supplies with +-47,000uf raw dc filtering, but un-regulated as yet). It was immediately better than the HFBD+P and a bit better than the modified DVD player. But the big surprise was when I split the power... I am a huge believer in getting power as clean as possible. Back in the day of early computer audio and the cMP playback setups, cics, the guy who worked out and developed the cMP concept, configuration, and software suggested using two ATX power supplies, one for the 'cleaner' side (main ATX power to the motherboard along with the harddrive) and one for the 'dirtier' side (P4 CPU power and any other peripherals). cics suggested the two sides should be plugged into separate AC circuits. I found this useful back then (even though I used linear supplies instead of computer ATX supplies) and have kept the configuration as I've improved my AC power setup and gone through some different sources. Currently I have a PS Audio P10 on the 'clean' side and there is significant parallel filtering upstream of it and the 'dirty' connection which goes to a separate circuit on the sub-panel.
I explain all of this because when I moved the AC connection for the Pi (and the input side of the DAM DAC's isolators) from the P10 to the 'dirty' outlets, I got a significant additional lift. At this point, ALL I wanted to do was listen! Instead I tore myself away after 45 minutes to go eat dinner and then start some power tweaks to the HFBD+P. But what I heard has me planning to go directly to a 3 power-supply setup, one for the DAC analog side, one for the DAC digital side, and a final one for the Pi. With the power supplies I have, it will be easy to put the Pi on the 'dirty' side and the DAC on the 'clean' side. I don't expect as dramatic of a difference, as I've only heard minor changes doing this in the past where all the other setups did have not had isolation and reclocking on the I2S stream like the DAM DAC does. I am guessing that is what made it more significant and dramatic of a difference than in the past.
But that notwithstanding, I figure why not go for broke first.
The rest of the setup is the before-mentioned modified DVD player (initial mods by Ric Schultz at EVS Homepage, last summer I added 2 small linear supplies to it providing separate regulated +-15v for an upgraded output stage and two separate regulated 3.3v supplies for the DAC digital and analog feeds with the drive and controls still fed by the Schultz-modified SMPS), an S&B TX-102 TVC, a pair of Ric Schultz B&O-IcePower-based amps, and a pair of Eminent Technology LFT-VIIIs with upgraded crossover components.
More after I poke around on the HFBD+P tonight and tomorrow, likely after I get it back in the system sometime in the coming week.
Greg in Mississippi
P.S. The LMS server side is also configured as a good setup, with 5 separate linear supplies (all using either LT3080 or LT3083 regulators and about 18,000uf/supply), one for the small Zotac server box (with the music on a 128Gb MicroSD card), one that I used for the Pi/HFBD+P tonight, but is designed to power the secondary music storage (either a 256Gb USB stick or an SSD), one each for the small switch and a small WiFi router that is used as a control access point and the final one, currently un-used, for a small monitor. No SMPS raw power supplies at all (of course there are several DC-DC switchers in the components) and it was all plugged into the 'dirty' AC outlets.
As the thing has been working pretty nicely headless with all of the control via the LMS webpage, I may re-configure the monitor supply for something else.
I put my Squeezebox-ish setup into my big system this evening.
I'm happy to say it all works flawlessly, at least so far.
First up was the R-Pi/HiFiBerry DAC+ Pro. I power it with a fairly serious linear supply using an LT3083 & about 18,000uf of capacitance. I've been running the unit some on my bench the last week or so and after about an hour warm up, I sat down to listen.
Impression was 'not bad, better than I thought'. But I clearly heard what I describe as 'digitalitis', an edge or at least a subliminal impression that has me uncomfortable after 5-10 minutes and wanting to turn off the music. Switching back to my super-modified Panasonic DVD player was a relief and it was clearly better in all areas.
I also listened to my setup with another Pi feeding I2S to a Soekris DAM DAC (stock at this point, but fed from another set of pretty massive supplies with +-47,000uf raw dc filtering, but un-regulated as yet). It was immediately better than the HFBD+P and a bit better than the modified DVD player. But the big surprise was when I split the power... I am a huge believer in getting power as clean as possible. Back in the day of early computer audio and the cMP playback setups, cics, the guy who worked out and developed the cMP concept, configuration, and software suggested using two ATX power supplies, one for the 'cleaner' side (main ATX power to the motherboard along with the harddrive) and one for the 'dirtier' side (P4 CPU power and any other peripherals). cics suggested the two sides should be plugged into separate AC circuits. I found this useful back then (even though I used linear supplies instead of computer ATX supplies) and have kept the configuration as I've improved my AC power setup and gone through some different sources. Currently I have a PS Audio P10 on the 'clean' side and there is significant parallel filtering upstream of it and the 'dirty' connection which goes to a separate circuit on the sub-panel.
I explain all of this because when I moved the AC connection for the Pi (and the input side of the DAM DAC's isolators) from the P10 to the 'dirty' outlets, I got a significant additional lift. At this point, ALL I wanted to do was listen! Instead I tore myself away after 45 minutes to go eat dinner and then start some power tweaks to the HFBD+P. But what I heard has me planning to go directly to a 3 power-supply setup, one for the DAC analog side, one for the DAC digital side, and a final one for the Pi. With the power supplies I have, it will be easy to put the Pi on the 'dirty' side and the DAC on the 'clean' side. I don't expect as dramatic of a difference, as I've only heard minor changes doing this in the past where all the other setups did have not had isolation and reclocking on the I2S stream like the DAM DAC does. I am guessing that is what made it more significant and dramatic of a difference than in the past.
But that notwithstanding, I figure why not go for broke first.
The rest of the setup is the before-mentioned modified DVD player (initial mods by Ric Schultz at EVS Homepage, last summer I added 2 small linear supplies to it providing separate regulated +-15v for an upgraded output stage and two separate regulated 3.3v supplies for the DAC digital and analog feeds with the drive and controls still fed by the Schultz-modified SMPS), an S&B TX-102 TVC, a pair of Ric Schultz B&O-IcePower-based amps, and a pair of Eminent Technology LFT-VIIIs with upgraded crossover components.
More after I poke around on the HFBD+P tonight and tomorrow, likely after I get it back in the system sometime in the coming week.
Greg in Mississippi
P.S. The LMS server side is also configured as a good setup, with 5 separate linear supplies (all using either LT3080 or LT3083 regulators and about 18,000uf/supply), one for the small Zotac server box (with the music on a 128Gb MicroSD card), one that I used for the Pi/HFBD+P tonight, but is designed to power the secondary music storage (either a 256Gb USB stick or an SSD), one each for the small switch and a small WiFi router that is used as a control access point and the final one, currently un-used, for a small monitor. No SMPS raw power supplies at all (of course there are several DC-DC switchers in the components) and it was all plugged into the 'dirty' AC outlets.
As the thing has been working pretty nicely headless with all of the control via the LMS webpage, I may re-configure the monitor supply for something else.
Thx for the update Greg.
Yep, power is everything. The DAC analog and digital side could probably also be powered with two 3.3V LiFePo4s. I used these before with good results.
My ultimate goal though is to keep things as simple as possible. My current setup comes very close to it. I think I've never been closer. Adding two batteries would make things more complex again.
I e.g. ditched my TX102 a long time ago ( I still have it in the basement).
I try to allign DAC output voltage, amp gain and speaker sensitivity instead. Mainly to limit the digital volume control impact.
All external VCs or preamps add their signature. I try to stay away from them.
Did I mention that I also have some IcePower ASP1000 sitting idle. I also sold my Hypex stuff. My current Anaview AMS0100-2300 is the best amp I ever had. As you might know Anaview is a spinoff of IcePower.
Looking forward to your results.
Yep, power is everything. The DAC analog and digital side could probably also be powered with two 3.3V LiFePo4s. I used these before with good results.
My ultimate goal though is to keep things as simple as possible. My current setup comes very close to it. I think I've never been closer. Adding two batteries would make things more complex again.
I e.g. ditched my TX102 a long time ago ( I still have it in the basement).
I try to allign DAC output voltage, amp gain and speaker sensitivity instead. Mainly to limit the digital volume control impact.
All external VCs or preamps add their signature. I try to stay away from them.
Did I mention that I also have some IcePower ASP1000 sitting idle. I also sold my Hypex stuff. My current Anaview AMS0100-2300 is the best amp I ever had. As you might know Anaview is a spinoff of IcePower.
Looking forward to your results.
Response, Part 1
Soundcheck,
Thx for the responses!
On the off-topic stuff first, I also have a couple of high-quality switch stepped attenuators. My main one (Daven broadcast-quality silver-contact switch with a nude Vishay series element and Vishay 102 shunt elements) is used in my downstairs system and clearly beats the TX102... but the S&B unit is a good close 2nd AND I need to deal with various source levels in that system (not so much in the downstairs system, which is more static). But I agree the method you've chosen is the right direction for the best sound!
On the ASP1000, the EVS units I have are based on that module. He did a good high-impedence front end buffer and some other tweaks to the module. I later figured out the Bel Canto MkII upgrade where they add pre-SMPS rectification and DC filtering and feed the modules 385v or so DC. This was a significant improvement to the modules and made them close to competitive with my modified NC400's, which are in the downstairs system. AND I'll have to check out the Anaview amps. I do have a pair of Lazy Cat's First One 1.3 modules running and they do beat both EVS B&O & the NC400, but are out of the system at present waiting some mod-work. AFTER I finish up this spat of digital source builds, I'll turn to resurrecting my phono setup & doing some amp builds/mods.
Back to the HFBD+P, I did get the first round of mods done last night and have verified operation, but not listened in the system. I have it setup for splitting the power, but have not done so yet... I'll listen first with a single power feed, then remove the 0-ohm resistor and feed 5v separately to the Pi and the HFBD+P board. Then third round is to cut the 3.3v feed from the Pi to the DAC's digital side AND remove/bypass the on-board 3.3v regulator (a good unit, but not stellar). I'll feed both with seperate 3.3v feeds from an OPC LT3042 board, which uses four of the better-spec'd LT3042 regulator chips in parallel on each half to give good current output (just under 1A).
And given that I'll have it setup for 3.3v feeds, I will also be able to easily try it with two LiFePO4 batts too.
On the specific DAC board mods, I replaced ALL of the ceramics on the analog side of the DAC (power and filtering both) with PPS SMD caps. Where I needed more capacitance than the 1uf of the largest PPS SMD that would fit, I stacked two or three. In one place I added a small 33uf Oscon for additional capacitance where the original cap was 10uf. And I added 560uf Nichicon polymers at the input to the analog and to the digital sections of the DAC, along with one at the 5v rail of the Pi.
At this point I left the clock bypasses alone... that's for another day.
The mods went fast (I had all the parts in my parts bins from other projects) and I was happy it worked ok with no issues!
I'll see how it sounds after some run-in time today.
I've attached pix of the board after I replaced the ceramics with PPS, another with the electrolytics added, a third of the OPC LT3042 regulator board, and a fourth of my LMS server setup (the linear supplies are under the 'mezzanine' with the 'nude' (someday to be heat-shrinked) switch, the power supply for the WiFi Router as an access point is visible, but the router itself is outside of the picture.
And finally, you can see I am not into simplicity! My main goal with this set of digital source builds is to learn the sonic impacts and effectiveness of certain setups, such as using a simple source like the R-Pi versus my previous full motherboard source (which required a fairly massive ATX linear supply to sound it's best, even after a LOT of SW & Op System tweaking) and things like isolation and reclocking such as that on the DAM DAC and in Ian's FIFO and Acko's S03. I'm planning to try to get each setup to sound as good as I can reasonably do... and my current end-point target for the HFBD+P setup will be fed via 3-4 separate supplies (Pi, DAC digital, DAC analog, and possibly a separate feed for the clock supply) PLUS bypassing the on-board filters by upsampling to 384. Curious where I can get it to with all this.
Part 2 Later after I listen to it!
Greg in Mississippi
P.S. I measured each cap after removing and before replacing. The original and replacements were as follows:
C1 & C5 2.2uf -> 2 x 1uf
C3 & C6 .1uf -> .1uf
C7 & C8 2.2nf - 2.2nf
C10 10uf -> 2x 1uf + 33uf Oscon
C12 10uf -> 3x 1uf + 560uf Nichicon
C11 Added 560uf Nichicon
5v Hdr Added 560uf Nichicon
Soundcheck,
Thx for the responses!
On the off-topic stuff first, I also have a couple of high-quality switch stepped attenuators. My main one (Daven broadcast-quality silver-contact switch with a nude Vishay series element and Vishay 102 shunt elements) is used in my downstairs system and clearly beats the TX102... but the S&B unit is a good close 2nd AND I need to deal with various source levels in that system (not so much in the downstairs system, which is more static). But I agree the method you've chosen is the right direction for the best sound!
On the ASP1000, the EVS units I have are based on that module. He did a good high-impedence front end buffer and some other tweaks to the module. I later figured out the Bel Canto MkII upgrade where they add pre-SMPS rectification and DC filtering and feed the modules 385v or so DC. This was a significant improvement to the modules and made them close to competitive with my modified NC400's, which are in the downstairs system. AND I'll have to check out the Anaview amps. I do have a pair of Lazy Cat's First One 1.3 modules running and they do beat both EVS B&O & the NC400, but are out of the system at present waiting some mod-work. AFTER I finish up this spat of digital source builds, I'll turn to resurrecting my phono setup & doing some amp builds/mods.
Back to the HFBD+P, I did get the first round of mods done last night and have verified operation, but not listened in the system. I have it setup for splitting the power, but have not done so yet... I'll listen first with a single power feed, then remove the 0-ohm resistor and feed 5v separately to the Pi and the HFBD+P board. Then third round is to cut the 3.3v feed from the Pi to the DAC's digital side AND remove/bypass the on-board 3.3v regulator (a good unit, but not stellar). I'll feed both with seperate 3.3v feeds from an OPC LT3042 board, which uses four of the better-spec'd LT3042 regulator chips in parallel on each half to give good current output (just under 1A).
And given that I'll have it setup for 3.3v feeds, I will also be able to easily try it with two LiFePO4 batts too.
On the specific DAC board mods, I replaced ALL of the ceramics on the analog side of the DAC (power and filtering both) with PPS SMD caps. Where I needed more capacitance than the 1uf of the largest PPS SMD that would fit, I stacked two or three. In one place I added a small 33uf Oscon for additional capacitance where the original cap was 10uf. And I added 560uf Nichicon polymers at the input to the analog and to the digital sections of the DAC, along with one at the 5v rail of the Pi.
At this point I left the clock bypasses alone... that's for another day.
The mods went fast (I had all the parts in my parts bins from other projects) and I was happy it worked ok with no issues!
I'll see how it sounds after some run-in time today.
I've attached pix of the board after I replaced the ceramics with PPS, another with the electrolytics added, a third of the OPC LT3042 regulator board, and a fourth of my LMS server setup (the linear supplies are under the 'mezzanine' with the 'nude' (someday to be heat-shrinked) switch, the power supply for the WiFi Router as an access point is visible, but the router itself is outside of the picture.
And finally, you can see I am not into simplicity! My main goal with this set of digital source builds is to learn the sonic impacts and effectiveness of certain setups, such as using a simple source like the R-Pi versus my previous full motherboard source (which required a fairly massive ATX linear supply to sound it's best, even after a LOT of SW & Op System tweaking) and things like isolation and reclocking such as that on the DAM DAC and in Ian's FIFO and Acko's S03. I'm planning to try to get each setup to sound as good as I can reasonably do... and my current end-point target for the HFBD+P setup will be fed via 3-4 separate supplies (Pi, DAC digital, DAC analog, and possibly a separate feed for the clock supply) PLUS bypassing the on-board filters by upsampling to 384. Curious where I can get it to with all this.
Part 2 Later after I listen to it!
Greg in Mississippi
P.S. I measured each cap after removing and before replacing. The original and replacements were as follows:
C1 & C5 2.2uf -> 2 x 1uf
C3 & C6 .1uf -> .1uf
C7 & C8 2.2nf - 2.2nf
C10 10uf -> 2x 1uf + 33uf Oscon
C12 10uf -> 3x 1uf + 560uf Nichicon
C11 Added 560uf Nichicon
5v Hdr Added 560uf Nichicon
Attachments
Last edited:
Part 2
With 8 hours or so of burn-in, the modified HFBD+P with a single 5v supply was significantly improved in lack of spurious noise, bass, and dynamics. There was still a hint of 'digitalitis', but it hit me after 10-15 minutes, not 2-3. Much more listenable and musical.
Next to provide 5v from separate supplies to the Pi/DAC digital/clocks and the DAC analog sections. I should be able to get to that by next weekend (I have a busier than typical week).
Thanks Soundcheck for leading the way!
Greg in Mississippi
With 8 hours or so of burn-in, the modified HFBD+P with a single 5v supply was significantly improved in lack of spurious noise, bass, and dynamics. There was still a hint of 'digitalitis', but it hit me after 10-15 minutes, not 2-3. Much more listenable and musical.
Next to provide 5v from separate supplies to the Pi/DAC digital/clocks and the DAC analog sections. I should be able to get to that by next weekend (I have a busier than typical week).
Thanks Soundcheck for leading the way!
Greg in Mississippi
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.