Hello all.
I am a long way behind you guys in knowledge and development so please bare with me as I ask this for advice.
I have Ian's isolator and I plan to power the Rpi with 3 separate regs for the 5v, 3.3v and 1.8v. Should I still use the Rpi 5v input on the isolator or go straight to the board.? I shall be removing the standard fit regulator .
Thankyou in advance
I am a long way behind you guys in knowledge and development so please bare with me as I ask this for advice.
I have Ian's isolator and I plan to power the Rpi with 3 separate regs for the 5v, 3.3v and 1.8v. Should I still use the Rpi 5v input on the isolator or go straight to the board.? I shall be removing the standard fit regulator .
Thankyou in advance
@jimk04,
I suggest you check out this thread:
Mezzanine Power board for Raspberry Pi
AND the board he sells here:
Raspberry Pi Power
I've used them to convert 2 RPi 2B's over to all-linear power with VERY good results. I posted a couple of pix in the DIYAudio thread above.
NOTE that the 2B is the latest RPi with which you can do this. See this article for more information on some of the changes implemented in the 3 and later RPi's and why you can't EASILY convert them to linear power:
More power to your Pi - Raspberry Pi
You can find more info on this modification to the RPi on these 2 pages:
http://www.easyaudiokit.com/bekkan/Raspberry/Raspberry.html
and
http://www.easyaudiokit.com/bekkan/Raspberry/Raspberry2.html
As for your question, I suggest going straight to the RPi board. On the first RPi I modified this way, I connected the 5V, 3.3V, and 1.8V to the appropriate input / output capacitors around the now-removed switching DC-DC converter chip. On the 2nd one, I used test points on the bottom of the board for 2 of the voltages and an appropriate cap for the other voltage... sorry I don't remember which for which. BUT going this way kept my connecting wires a little shorter and neater while still using insertion points near to the origin points.
BUT by going to the board, you keep the voltage insertion points close to the original origin points.
I hope this helps! It is a well worthwhile modification. OTOH, Allo will soon be releasing their USBridge Signature which will do the same in a more sophisticated manner... of course for a price yet-to-be publicized.
Greg in Mississippi
P.S. While this is an involved and advanced DIY project, most can be done pretty easily with care and planning. The hard part was removing the DC-DC converter chip. I found it best to use a fairly serious soldering iron on top of the chip package after pre-heating the area with my SMD rework hot air gun. I did ruin one RPi while removing that chip, so it took me 3 RPi's to get 2 operational in this mode.
I suggest you check out this thread:
Mezzanine Power board for Raspberry Pi
AND the board he sells here:
Raspberry Pi Power
I've used them to convert 2 RPi 2B's over to all-linear power with VERY good results. I posted a couple of pix in the DIYAudio thread above.
NOTE that the 2B is the latest RPi with which you can do this. See this article for more information on some of the changes implemented in the 3 and later RPi's and why you can't EASILY convert them to linear power:
More power to your Pi - Raspberry Pi
You can find more info on this modification to the RPi on these 2 pages:
http://www.easyaudiokit.com/bekkan/Raspberry/Raspberry.html
and
http://www.easyaudiokit.com/bekkan/Raspberry/Raspberry2.html
As for your question, I suggest going straight to the RPi board. On the first RPi I modified this way, I connected the 5V, 3.3V, and 1.8V to the appropriate input / output capacitors around the now-removed switching DC-DC converter chip. On the 2nd one, I used test points on the bottom of the board for 2 of the voltages and an appropriate cap for the other voltage... sorry I don't remember which for which. BUT going this way kept my connecting wires a little shorter and neater while still using insertion points near to the origin points.
BUT by going to the board, you keep the voltage insertion points close to the original origin points.
I hope this helps! It is a well worthwhile modification. OTOH, Allo will soon be releasing their USBridge Signature which will do the same in a more sophisticated manner... of course for a price yet-to-be publicized.
Greg in Mississippi
P.S. While this is an involved and advanced DIY project, most can be done pretty easily with care and planning. The hard part was removing the DC-DC converter chip. I found it best to use a fairly serious soldering iron on top of the chip package after pre-heating the area with my SMD rework hot air gun. I did ruin one RPi while removing that chip, so it took me 3 RPi's to get 2 operational in this mode.
@minivan,
I was having similar thoughts to what @wlowes posted when I saw his post. I didn't post because his was very complete. BUT he didn't suggest what I do below, hence my post now.
I'm not quite sure what @bigpandahk means by 'by pass the LT3042 rectification circuit'. I'm guessing he meant regulation circuit. BUT what happens when you feed the 3.3V into the 5V input of the stock FiFoPi is that the 3.3V is fed to the 5V inputs of the on-board regulators (some of which are LT3042, but not all... look for some posts by @Markw4 and me earlier in this thread for more details).
These regulators, like most linear regulators, will pass that 3.3V onto their output. BUT since they are not actually regulating while still passing through the regulators, both the current limit of the regulator chips will still be active AND the output impedence of the 3.3V will be very poor. I found that running the FiFoPi this way, while it worked, was the poorest sounding power option of all that I tried (and I tried a bunch!).
AND I think the issue you're having is that the current-limiting of the regulators is not allowing sufficient current from the 3.3V battery to your Pulsar clock.
Your best next step here is to add the soldered-links on the back of the board to ACTUALLY bypass the regulators. That ensures the low output impedance of the battery is preserved along with the full current capacity. BUT be sure to never feed 5V there once you have physically bypassed the regulators OR you will destroy your FiFoPi! I color-code my connectors to help me remember which FiFoPi I have modified and which I haven't.
I bet that will correct your issue.
Greg in Mississippi
I was having similar thoughts to what @wlowes posted when I saw his post. I didn't post because his was very complete. BUT he didn't suggest what I do below, hence my post now.
I'm not quite sure what @bigpandahk means by 'by pass the LT3042 rectification circuit'. I'm guessing he meant regulation circuit. BUT what happens when you feed the 3.3V into the 5V input of the stock FiFoPi is that the 3.3V is fed to the 5V inputs of the on-board regulators (some of which are LT3042, but not all... look for some posts by @Markw4 and me earlier in this thread for more details).
These regulators, like most linear regulators, will pass that 3.3V onto their output. BUT since they are not actually regulating while still passing through the regulators, both the current limit of the regulator chips will still be active AND the output impedence of the 3.3V will be very poor. I found that running the FiFoPi this way, while it worked, was the poorest sounding power option of all that I tried (and I tried a bunch!).
AND I think the issue you're having is that the current-limiting of the regulators is not allowing sufficient current from the 3.3V battery to your Pulsar clock.
Your best next step here is to add the soldered-links on the back of the board to ACTUALLY bypass the regulators. That ensures the low output impedance of the battery is preserved along with the full current capacity. BUT be sure to never feed 5V there once you have physically bypassed the regulators OR you will destroy your FiFoPi! I color-code my connectors to help me remember which FiFoPi I have modified and which I haven't.
I bet that will correct your issue.
Greg in Mississippi
Last edited:
@jimk04
With my audio pc modding experience I found that most noise comes from poor clocks itself and/or poor power to the clocks. rpi clocks are poor quality and are powered by 3.3V switcher. So first I would tackle that. That is the biggest gain you will get.
I would go with this board as first:
Raspberry Pi 3B+ modified clock crystal TCXO warm-filled crystal oscillator upgrade crystal oscillator + clean linear 5V to it (this board already has two 3.3V LDO on it, one per clock).
Then power the rpi with clean 5V.
With my audio pc modding experience I found that most noise comes from poor clocks itself and/or poor power to the clocks. rpi clocks are poor quality and are powered by 3.3V switcher. So first I would tackle that. That is the biggest gain you will get.
I would go with this board as first:
Raspberry Pi 3B+ modified clock crystal TCXO warm-filled crystal oscillator upgrade crystal oscillator + clean linear 5V to it (this board already has two 3.3V LDO on it, one per clock).
Then power the rpi with clean 5V.
Last edited:
@TioFrancotirador,
Of course replacing the DC-DC converters on the RPi also upgrades the power to the on-board clocks... so it will improve their performance. OTOH, that does not touch the clock quality.
I like the board you linked. How much is it? Have you used it? What were your experiences?
I'm even more interested in how it might be used with the soon-to-be released Allo USBridge Signature. They have designed and implemented a carrier board for an RPi Compute Module with all low-noise linear-regulators and an additional USB chip to correct the earlier RPi USB/Ethernet resource conflicts. AFAIK, their carrier board implements one of the clocks, the other is on the Compute Module. While the implementation will be a bit more challenging, the potential benefits should be even greater.
Thanks for mentioning this!
Greg in Mississippi
Of course replacing the DC-DC converters on the RPi also upgrades the power to the on-board clocks... so it will improve their performance. OTOH, that does not touch the clock quality.
I like the board you linked. How much is it? Have you used it? What were your experiences?
I'm even more interested in how it might be used with the soon-to-be released Allo USBridge Signature. They have designed and implemented a carrier board for an RPi Compute Module with all low-noise linear-regulators and an additional USB chip to correct the earlier RPi USB/Ethernet resource conflicts. AFAIK, their carrier board implements one of the clocks, the other is on the Compute Module. While the implementation will be a bit more challenging, the potential benefits should be even greater.
Thanks for mentioning this!
Greg in Mississippi
@Greg
It costs ~50$
I ordered it just before Rpi4 was released. This board is dedicated for rpi3b+ (one clock 19,2Mhz for cpu, the other 25Mhz for LAN/USB chip).
Now in Rpi4 there are actually three clocks: CPU 19,2Mhz, LAN 25Mhz, USB 25Mhz. But if you only use it for streaming/network and no USB dac nor using USB for music files then you can skip replacing USB clock.
I have this board and plan it to use it on top of rpi4 together with Ian's gear, so those two clocks are enough.
I expect having same results as got in Audio PC. Once you replace your clocks on mother board for H81chipset(including USB) and Lan then power it with clean PSU, the other mods, which I also tried, like:
- mother board recapping
- clean ATX power to the rest of the components
- addtional dedicated audio USB card
- additional dedicated audio LAN card
- running system from RAM
- Windows Server tweaking
- powering SSD with clean 5V
are really of much, much, much less importance ...
So, actually if one wants to use rpi4 as streamer and connect USB dac in to it then you need those three clocks that I mentioned.
Allo USBridge Signature - is like dedicated USB card for PC ... I found that eliminating problems at its root is better
I used to use Allos sparky usb bridge - not even close to my dedicated audio pc setup.
It costs ~50$
I ordered it just before Rpi4 was released. This board is dedicated for rpi3b+ (one clock 19,2Mhz for cpu, the other 25Mhz for LAN/USB chip).
Now in Rpi4 there are actually three clocks: CPU 19,2Mhz, LAN 25Mhz, USB 25Mhz. But if you only use it for streaming/network and no USB dac nor using USB for music files then you can skip replacing USB clock.
I have this board and plan it to use it on top of rpi4 together with Ian's gear, so those two clocks are enough.
I expect having same results as got in Audio PC. Once you replace your clocks on mother board for H81chipset(including USB) and Lan then power it with clean PSU, the other mods, which I also tried, like:
- mother board recapping
- clean ATX power to the rest of the components
- addtional dedicated audio USB card
- additional dedicated audio LAN card
- running system from RAM
- Windows Server tweaking
- powering SSD with clean 5V
are really of much, much, much less importance ...
So, actually if one wants to use rpi4 as streamer and connect USB dac in to it then you need those three clocks that I mentioned.
Allo USBridge Signature - is like dedicated USB card for PC ... I found that eliminating problems at its root is better
I used to use Allos sparky usb bridge - not even close to my dedicated audio pc setup.
Last edited:
DSD Upsampling
Very nice products and projects.
I have my order filled, but like to have some confirmation that they will work in my system.
I'd like to use them with Roon/Hqplayer.
Upsampling all music to dsd512
send it to endpoint RPi4 (HQPlayer NAA or Ropieee Roon Bridge)
from what I read: I don't need a driver for linux and if I like to I can use the 'HifiBerry' one. is that true?
Can I use the max sample rates (PCM 768 and DSD1024) with RPi4?
I hope the answers are yes and im looking forward to share my impressions with you!
Very nice products and projects.
I have my order filled, but like to have some confirmation that they will work in my system.
I'd like to use them with Roon/Hqplayer.
Upsampling all music to dsd512
send it to endpoint RPi4 (HQPlayer NAA or Ropieee Roon Bridge)
from what I read: I don't need a driver for linux and if I like to I can use the 'HifiBerry' one. is that true?
Can I use the max sample rates (PCM 768 and DSD1024) with RPi4?
I hope the answers are yes and im looking forward to share my impressions with you!
FYI
https://www.diyaudio.com/forums/vendor-s-bazaar/327794-output-transformers-dacs-16.html#post5878430
I am wonders who does it compare it now to Katana.
Would be lovely if Ian wrote THD compensation program for this output HAT
https://www.diyaudio.com/forums/vendor-s-bazaar/327794-output-transformers-dacs-16.html#post5878430
I am wonders who does it compare it now to Katana.
Would be lovely if Ian wrote THD compensation program for this output HAT
I am wonders who does it compare it now to Katana.
Would be lovely if Ian wrote THD compensation program for this output HAT
Katana has separate AVCC supplies for AVCC_L and AVCC_R. That makes stereo separation much better.
There is also the matter of the shield board Ian is trying out now. Katana (with optional isolator), has better shielding between boards and of certain components, but worse default power supplies. A big limitation using RPi is the maximum DSD sample rate supported. However, Ian's dac can accept high sample rate DSD from a USB board which also gets rid of all the RPi noise issues (which is not an option for Katana). That leaves only some stray coupling between the boards and a tiny bit of RF leakage from the DAC outputs to the output stage inputs. Personally, I would also like to have the output stage differential summing more optimized for OPA1612 rather than OPA1622. Potentially it could be a killer dac, perhaps in the next design iteration.
Last edited:
Actually hifiberry is limitited to 192kHz/24bit only, so you can't.
@Ian,
Maybe you could design simple adapter board to your fifo pi for most popular USB receivers:
Amanero and XMOS (DSD1024) from diyinhk (I love this one!).
This would greatly enhance your target audience and could squeeze out last drop of hifi juice from your gears
Last edited:
Here's some clarifications to what @Markw4 wrote above:
- When Ian's DAC boards are used with Ian's LiFePO4 supply for which they were designed and optimized, there is very little negative impact on the stereo separation by use of a common AVCC power feed between the right and left DAC channels. When used with a typical LDO or better actively regulated power supply which are Markw4's preferred supply setup, there is. I have used Ian's DAC boards both ways and with Ian's LiFePO4 supply, there is just a tiny bit of drop in soundstage width and depth compared to my Katana setups.
- Noise generated by the attached RPi and its associated power supply, processing, and networking is something a user can control. Use an SMPS to power the RPi instead of a good linear supply and use an RPi 3 or 4 with WiFi turned on instead of using wired ethernet (or possibly even an RPi 2B with no on-board WiFi) and you are creating a high-RF environment for the FiFoPI and DAC card. Proper grounding plays a role in this too. I do suspect there is some benefit to further shielding the FiFoPi clocks from the DAC board similar to how Allo has a shield can over the clocks on the Katana. But one can make implementation choices that reduce and largely eliminate electrical and RF noise generated in the RPi.
@Markw4, I appreciated your contributions to the art of getting the best out of these small ESS-based DAC cards such as your suggestion of providing a means to adjust the THD compensation for these boards and all of the other good info in your large ES9038Q2M DAC board thread. AND I am curious on your suggestions to optimize the output stage differential summing stage to better use the OPA1612 there... what do you have in mind?
BUT to make such blanket damning statements about Ian's DACs when AFAIK you have not used them as they were designed and intended, but as testbeds for your own modifications, seems a bit irresponsible to me. If you need to make such blanket statement in the future, please qualify them.
Greg in Mississippi
P.S. To @TioFrancotirador's comments on Ian's gear versus the Katana, I have run them both in a number of configurations across 2 systems (which even though they have very similar gear, highlight different areas of performance due to their setups and rooms). I find them roughly comparible when using NDK SDA clocks. My best combos include Ian's IVStd modified with lower filtering cutoffs and using Sparkos (and possibly Burson and LKS) discrete opamps along with the OPA1612 AND Ian's transformer-coupled output using the now unavailable Onetics transformers.
Each one does some things better than the other and other things worse. But to my ears and preferences and in my setups, comparible.
I have a pair of @Bisesek's transformers on the way (shipped this week) and will report on them here when I have them in hand.
Last edited:
Oké thank you! I’m reading thru all the pages but it’s hard to filter all right info out of it..
I understood that I could use a generic i2s driver?
Will the rpi4 output DSD512 via fifopi/dual mono 9038q2m or isn’t the rpi4 capable.. I’m still not certain
Very OT I know but I figure a god place to ask.....and I have tried elsewhere before but I have more confidenece here...
How do yo turn your Rpi streamers on and off?
I have my Rpi and DaC hat powered by 2 xs Salas L adaptor for now....using Ians Isolator.
Ians manual stats about power on sequence, mine all boots up at once but no issues .
my Q to yo is what s the suggested way to turn on and off the Rpi system?.....I simply pull the mains wall plug AC once I have paused the music. plug back in and al seems to be well.....am I on borrowed time? Should I be powering down otherwise?
How do yo turn your Rpi streamers on and off?
I have my Rpi and DaC hat powered by 2 xs Salas L adaptor for now....using Ians Isolator.
Ians manual stats about power on sequence, mine all boots up at once but no issues .
my Q to yo is what s the suggested way to turn on and off the Rpi system?.....I simply pull the mains wall plug AC once I have paused the music. plug back in and al seems to be well.....am I on borrowed time? Should I be powering down otherwise?
I'll defer to Rpi experts for best approach to shutdown. I'll chime in with what I do. I almost never power it down. When I do it is same as you. Kill mains without a formal Linux shutdown. DAC's, clocks etc do best when constantly powered. If you think about it, you don't shut down your router. It is just another Linux processor with some software on it. We leave them on all the time and when we do shut them down it is the good ol power cord pull.