I am surprised! Even the NDK NZ2520SD has this feature! How could I not notice this.....
I will try difference between Expresso and NDK NZ2520SD (I am absolutely not convinced Expresso would sound better) so I will try this!!
Thanks for pointing me in the right direction
cool, would these just drop right in, no modifications?
Unless they come in different sizes that I'm not familiar with, not quite that easy.
Same pin-out, vastly different footprint.
Here are two NDKs on top of one of the HFBD+P clocks.
I agree the NDK would be better. But not an easy replacement unless the pads under the existing clocks are sized to accommodate differently-sized clocks. This I doubt.
Still, I still suspect other issues.
Sorry for being absent for the past couple of weeks, been busy at work and in life. I should be able to catch up on replies this week.
Greg in Mississippi
Attachments
Thanks guys for all the external linear power supply tips. I'm about to experiment with a Pi3 and DAC+ Pro outputting I2S to my Buffalo II DAC. A couple of questions:I finally managed to feed DAC+Pro external with 3.3V.
Indeed the bridge wire from pin 17 to 3.3V next to TCXO did the trick, soldered it
to the cap, just as Greg did in post 106. This works!
It gained in SQ again, very happy with it.
BUT:
It's sad they use ExpressO TCXO's which we can not exchange with better ones,
at least, not that easy.
Since the ExpressO TCXO's are switched on and off from software (driver), we can't
use normal TCXO's like the NDK NZ2520SD which I am very fond of.
ExpressO FXO-HC73 which are used for DAC+Pro have very mediocre specs.
They have not so good Phase noise performance
ExpressO FXO-HC73 NDK NZ2520SD
1Hz ??? -85 dB
10Hz -65 dB -113 dB
100Hz -100 dB -140 dB
1kHz -120 dB -153 dB
10kHz -120 dB -157 dB
100kHz -125 dB -158 dB
This is why I prefer NDK NZ2520SD.
Maybe I will create a double relais board which will take use of the ON/OFF signal
to trigger the relais to use. In that case I could use the NDK's.
Given that I have two linear power supplies available what would you advise is the best combination for external power? 5V down the GPIO powering the PI is a given. What about the other? 3V3?
Also, is the 5V DAC analog power mandatory when only using I2S?
cheers,
Bruce
Gave it a try! I'm still always surprised when you can get more out of digital... another surprising improvement. This had me up for awhile just listening.
Next step for me is build a few tiny APD150 daughter boards, and truly separate the power.
Did a video for you guys
https://www.youtube.com/watch?v=JxEdd6GJCNA
Thanks so much!
You're right! Bent away pin 1. For some reason I thought there was just one 3.3v pin.
Appended the video. Listening now, luckily it's unchanged (improvement remains)--
Last edited:
BTW have you thought about bypassing the USB input in your DL3 and sticking I2S from the DAC+ Pro straight into it's DAC?
Yes, would love to rip out the usb chip and tap into the i2s... but it's my main Dac, don't want to risk breaking it. Also it has those Cullen IV upgrades, may complicate things. I'd do it if it was a simple tap into it (maybe it is), but bet there are control signals etc, I'd be guessing.
Ok, yeah those are tiny. I bet layout is critical on those clocks-- can't just wire them in.
Digi+
Any thoughts on decoupling the digi+ WM8804 from the 3.3v pi, similar to the dac+ that I did here:
https://www.youtube.com/watch?v=JxEdd6GJCNA
(Using ADP150's or a direct 3.3v supply)
The WM8804 has a PLL supply and Digital supply, is it worth separating these?
this is that U2 regulator on the Dac:
ADP150 Datasheet and Product Info | Analog Devices
Just using linear power thru the GPIO is a big improvement (already do that)
Thanks!
Any thoughts on decoupling the digi+ WM8804 from the 3.3v pi, similar to the dac+ that I did here:
https://www.youtube.com/watch?v=JxEdd6GJCNA
(Using ADP150's or a direct 3.3v supply)
The WM8804 has a PLL supply and Digital supply, is it worth separating these?
this is that U2 regulator on the Dac:
ADP150 Datasheet and Product Info | Analog Devices
Just using linear power thru the GPIO is a big improvement (already do that)
Thanks!
Attachments
I soldered those NDK NZ2520SD a few times, yes, it's almost brain surgery, but doable.
This was my first attempt with very thin wire, but they broke off all the time
because they were too stiff
https://drive.google.com/open?id=0B5ivzpgUT3UYUDRFM3o0Z29vRDQ
Then I decided to try a different method...
https://drive.google.com/open?id=0B5ivzpgUT3UYTlVDeVp0YU9MWG8
This worked for me, but as said before, it's a kind of brain surgery method....
This method works best:
https://drive.google.com/open?id=0B5ivzpgUT3UYSEtkQ0VEZW5ETHM
It's just a simple, low cost, pcb. Put some solder on the 4 pads, put XO on it,
fix it on one side with some adhesive tape, press on top and solder them one by one.
This worked for me.
Last edited:
Very agreed. I've seen this done with flying wires, but I really like the idea of using an available protoboards. I have to find some of these!
I suspect one could cut up those protoboards and make clock adapter boards from them that will solder directly in place of the stock HFBD+P (and other DACs) clocks!
I don't think you'd have any problems either doing flying wires or an adapter board as above.
Gotta love OSHPark! OTOH, the issue with doing a custom board of our own is getting a driver that will enable the Pi to generate the I2S stream using the bitclock from the DAC board. I suspect that U3 chip on the HFBD+P provides the information that the HiFiBerry driver needs to tell whether it is working with their DAC that does this or the earlier HFB DACs. That may be hard to fake.
OTOH, looking at the SabreBerry32, it does not appear to do the same thing, I suspect the driver assumes a bitclock and LRclock feed from the DAC card. So we MIGHT be able to use it's driver with a custom board.
OTOOH, I like what I see on the SabreBerry32. I'd like to see a more sophisticated power setup as on the MamboBerry, with its extensive filtering and two VERY high quality industry-best regulators (It really does set the standard for Pi DAC card power supply setups from what I've seen). AND I'd REALLY like to see isolation and reclocking between the DAC card and the inherently noisy Pi. I can work with the first... and live with the latter for now, as it would take a significant restructure of the interface between the Pi and the DAC card to make isolation and reclocking happen. I do like their use of the good NDK clocks and the almost industry standard ESS DAC chip. I will likely get one of these to play around with too.
While I do have devices that use TI DAC chips to good effect (my modified Sony HAP Z1-ES and Panasonic S47 DVD Player, for example), my understanding is that the best results are obtained with the PCM5122 on the HFBD+P when you provide it with a music feed that has been upsampled to 384kHz using a better filter so that it bypasses the on-board filters.
Thoughts for today.
Greg in Mississippi
P.S. I've gone to direct out from the R2R ladder on my Soekris setups and getting the best performance with them I've experienced so far... and again without any of the common mods to them yet. I had not done this at first as I was sure my setup with only good passive attenuators between any source and my amps would need more output volume than available from the Soekris direct output. Not true, I don't have a lot of spare volume anymore, but do have plenty for even lower-than-normal level recordings like the CD transfers of the Sheffield Direct-to-Disk recordings. VERY glad I tried this, my Soekris setup will now be a lot simpler going forward!!!
BTW, I did try the balanced output from the Soekris... it sounded very bad in my setup, almost broken compared to a single-ended setup which is handily surpassed by the R2R direct. If that holds true for other setups (and I have no reason to doubt that), anyone not using the direct R2R feed is not hearing what that DAC can do.
I also have put the Mamboberry back into the system for a short while and it still holds up to what I said previously.
And I have a go-forward path on further mods to the HFBD+P... to be detailed in a future post.
I'd definently talk to hifiberry first, they did lots of work on that driver, and just in general how they feel about people making compatible boards. For all I know maybe they want people to use it and contribute to improving it.
hello, i´ve been reading this tread with great enthusiasm, i ´m new to electronics... and i need some help. a bought a ak4495 dac thas has i2s input:
1 GND
2 MCK
3 LRCK
4 BCK
5 DATA
and i want to connecto the i2s out from hifiberry dac pro to ak board...
is it ok to do it like this?
from p4
1 GND - GND
2 MASTER CLOCK - MCK
3 BIT CLOCK - ?????BCK??
4 FRAME CLOCK - ????? LRCK??
5 DATA- DATA
i have the mamboberry and use last archphile version . it sounds very natural...
1 GND
2 MCK
3 LRCK
4 BCK
5 DATA
and i want to connecto the i2s out from hifiberry dac pro to ak board...
is it ok to do it like this?
from p4
1 GND - GND
2 MASTER CLOCK - MCK
3 BIT CLOCK - ?????BCK??
4 FRAME CLOCK - ????? LRCK??
5 DATA- DATA
i have the mamboberry and use last archphile version . it sounds very natural...
@Soundcheck,
First, I apologize again for the slowness of my responses... work and some other activities have kept me busy the last few weeks.
On the Mamboberry with two supplies:
1. The Mamboberry board has a jumper connection. If the jumper is in place, the Pi is powered from the Mamboberry's supplies via the expansion header 5v pins. If disconnected, the Pi needs to be separately powered. They showed this in the previous iteration of their website, but I didn't find it mentioned now. I asked them via email about doing this and they confirmed it was designed to allow the DAC and Pi to be separately powered and did not caution me away from that.
2. While I do have the Mamboberry DAC Hat and Pi powered from entirely separate supplies (down to separate transformers and power cords), I do take care to make sure their final grounding point (sort of a star-ground) is at the DAC/Pi unit, so that they stay at the same ground potential. In my experience this is a common configuration, I've only taken it a step farther by separating the power cords too. I used this to good effect back in the middle 90s with my massive modification to a Magnavox CDB-650 (5 supplies, 3 power cords) and more recently my fully linear cMP/cPlay PC-based player (13 supplies at last count, if I keep it in service it will go up to 14 supplies, with 5 power cords). And the Sony HAP Z1-ES stock has 3 transformers and 6 supplies. While I did not increase the number of supplies when I modified it, I did wire in separate power cords for the digital side versus D-A & analog out side.).
When I tested both the MamboBerry and the HiFiBerry DAC+ Pro, I started running them basically stock with a single supply. It was only after getting a feel for the sound that way that I went to multiple supplies. IN EVERY CASE separately powering the Pi sounded better in my setup. In my experience, that is always true, but I like to confirm it and not leave a stone un-turned.
Also, IMHO, the MamboBerry power supply setup is the best among the R-Pi DAC hats I've used and looked at (as far as can be told from pictures and experience). That it is designed to take AC from an appropriate transformer to drive an on-board linear supply that in the basic configuration feeds both the DAC and the Pi helps ensure a poor-sounding SMPS will not be used with the unit. AND the supply is fairly sophisticated and well-designed, with what appear to be good rectifier diodes, an extensive set of filtering capacitors, and great-sounding best-of-class regulators from TI (the same ones John Swenson prefers for his designs such as the Regen, the uRendu, and the Bottlehead DAC). And they use a 2nd regulator cascaded to for the DAC to somewhat isolate it from the electrically-noisy Pi.
This supply is a strong driver of the unit cost compared to other units. For example the two regulators they use in single quantities are $5 and $3... the one used on the HFBD+P is $1. Only the DurioSound Pro units begin to approach this level of sophistication and they use one of the same regulators to power their DAC chip).
Where I think there is room for improvement on the MamboBerry is to power the clock separately from the DAC. And of course use two clocks with the DAC Hat in master mode to feed the Bitclock to the Pi as the I2S signal clock source... but that would require an extensive re-design AND a new driver... or one could just get the SabreBerry32, which while it does not have as sophisticated of a power supply setup, does run in master mode AND uses the highest-quality clocks of any of these units.
And in my experience, upgrading power supplies is a lot easier to do and do fairly right than restructuring the clocking and I2S signal feeds.
Greg in Mississippi
@Soundcheck,
It sounds like you've done some great work optimizing your setup!
Sorry to say, I'm still using the stock PiCorePlayer. I want to get through the HW changes first before I tackle the SW/OpSys.
At this time, I only plan one more round of changes to the HFBD+P. I consider the MamboBerry a success as it is now and while I may tweak it a bit in the future, that is not a high priority.
After I do the next round of tweaks to the HFBD+P, I'll start looking back through your suggestions here... and start asking questions where I need to do so.
THANKS for all of your hard work wrestling with the HW & OpSys.
Greg in Mississippi
@Soundcheck,
Some very good points.
I'd agree on the layout... not impressive. OTOH, I'm not sure which is better, close clocks or good clocks! I guess I'm going to have to find out if the price is not too bad.
I'm more concerned about the power supply situation, which appears pretty basic. But as I mentioned, I see that as something that can be addressed as opposed to the clock and I2S signal situation, which is much harder to upgrade and make it work well.
I don't have the datasheet for the ESS9018Q2C. I'll have to get that and check what output modes it has, but if only current as the older ESS018, I'd agree that running this in voltage output mode is not optimal.
And I've heard that comment on a well-implemented 9023 vs a poor 9018 one... and my experience is that is true.
Still, the SabreBerry DOES have better clocks than anything else I know of. And unlike the DurioSound and IQAudio DACs along with the HiFiBerry non-Pro units, it DOES have clocks which IMHO HAS to be better than the PCM5102/5122 PLL mode.
We'll have to see how much they want for it!
Greg in Mississippi
Last edited: