No Mark. There's more to that story.
Taking out the filters is one thing. If you'd then feed HireZ your "distortions"
issue simply evaporates. Ess recommends to feed 8*FS if you go OSF-B.
And even if you'd feed 44.1 there might even be again just a tradeoff or a compromise you'll face: "DSP related losses vs. little distortions". Depending on the setup the magnitude of impact would probably even look different.
Simply by looking at the awful curve of that 9038 minimum phase filter gets my head spinning. On top of that comes the filter/DSP associated losses. Nope what you hear is not fake clarity in OSF-B.
I ran my 51xx DACs this way for years! And guess what. I loved it.
And my own 64bit resampler increased "clarity" even further. DSP quality to me is quite a topic.
The key is to have that OSF-B option at hand. Let the users decide. From that perspective Ians DAC is the much better choice.
As discussed over at Ians, I guess you followed that discussion, you'll loose the on-dac software volume control if you run OSF-B. But who cares!?!?
***
I'm quite sure that true-sync is a different feature then that what you describe. With true sync you'd loose DSD/DoP capabilities. I'm not aware that Katana offers such a mode. But I'm not 100% sure.
***
If you think your opamps face RFI/EMI issues and these have audible impact
you might consider a transformer output stage. See Bisesiks great HAT for Ians DAC. On top you'd get rid of all this annoying opamp stuff and +-15V PS crap.
All I can say is, that I really enjoy my no frill "USBless" lowest power RPI Wifi setup.
Enjoy.
Taking out the filters is one thing. If you'd then feed HireZ your "distortions"
issue simply evaporates. Ess recommends to feed 8*FS if you go OSF-B.
And even if you'd feed 44.1 there might even be again just a tradeoff or a compromise you'll face: "DSP related losses vs. little distortions". Depending on the setup the magnitude of impact would probably even look different.
Simply by looking at the awful curve of that 9038 minimum phase filter gets my head spinning. On top of that comes the filter/DSP associated losses. Nope what you hear is not fake clarity in OSF-B.
I ran my 51xx DACs this way for years! And guess what. I loved it.
And my own 64bit resampler increased "clarity" even further. DSP quality to me is quite a topic.
The key is to have that OSF-B option at hand. Let the users decide. From that perspective Ians DAC is the much better choice.
As discussed over at Ians, I guess you followed that discussion, you'll loose the on-dac software volume control if you run OSF-B. But who cares!?!?
***
I'm quite sure that true-sync is a different feature then that what you describe. With true sync you'd loose DSD/DoP capabilities. I'm not aware that Katana offers such a mode. But I'm not 100% sure.
***
If you think your opamps face RFI/EMI issues and these have audible impact
you might consider a transformer output stage. See Bisesiks great HAT for Ians DAC. On top you'd get rid of all this annoying opamp stuff and +-15V PS crap.
All I can say is, that I really enjoy my no frill "USBless" lowest power RPI Wifi setup.
Enjoy.
Last edited:
Benchmark DAC-3 uses OSF bypass and they implement their own volume control -- inside the FPGA that implements their custom oversampling (interpolation) filter.
Of course, people can do as they wish. I don't like the PCM filters in Sabre dacs either, but filters should be there for reconstruction to work properly. That's why I use AK4137 to convert all PCM to DSD512. The DSD interpolation filters in Sabre dacs are much better sounding and quite tolerable. Best sound is that way, not with no filter at all and playing PCM.
Also, when people are attracted to false clarity it means there are other problems in the reproduction system, most typically problems that are making it sound muddied and dull in the midrange. Everyone who hears how a system is really supposed to sound always likes it better than false clarity, but many never get the chance to compare, unfortunately.
I tried to explain how to get the best sound of of Katana, and also for Ian's FIFO-Pi dac setup. A few tried it with Katana and liked it. Their problems are solved. I don't think anyone tried the fixes for FIFO-Pi dual ES9038Q2M dac, except maybe Ian. I see he came up with some of his own fixes that are similar, including a shield board and a filter at the input of the output stage.
In the meantime, I'm off working on AK4499. Leaves Sabre dacs behind. Time for change is coming in the world of dacs over the next year or two, some changes sooner, some later.
Of course, people can do as they wish. I don't like the PCM filters in Sabre dacs either, but filters should be there for reconstruction to work properly. That's why I use AK4137 to convert all PCM to DSD512. The DSD interpolation filters in Sabre dacs are much better sounding and quite tolerable. Best sound is that way, not with no filter at all and playing PCM.
Also, when people are attracted to false clarity it means there are other problems in the reproduction system, most typically problems that are making it sound muddied and dull in the midrange. Everyone who hears how a system is really supposed to sound always likes it better than false clarity, but many never get the chance to compare, unfortunately.
I tried to explain how to get the best sound of of Katana, and also for Ian's FIFO-Pi dac setup. A few tried it with Katana and liked it. Their problems are solved. I don't think anyone tried the fixes for FIFO-Pi dual ES9038Q2M dac, except maybe Ian. I see he came up with some of his own fixes that are similar, including a shield board and a filter at the input of the output stage.
In the meantime, I'm off working on AK4499. Leaves Sabre dacs behind. Time for change is coming in the world of dacs over the next year or two, some changes sooner, some later.
Last edited:
As I said. Simply feed 352k8 or 384 resampled by a HQ resampler and you'll be fine. Live can be that easy.
If Allo offers us a new firmware and a driver update we all can and probably will try (and enjoy) it.
***
To me DAC chips are very close to become commodity products.
The recipes of making a good DAC device are known to most manufacturers.
Even 50$ devices claim femto seconds jitter and <-110dB THD/N nowadays.
IMO since years we keep spinning in circles around here about the same subjects.
And to be honest. I could have lived with my 51xx DAC forever. Yep. I'm getting older. Actually my 1543 DAC sounded really great more then 10years ago. And it still does.
And my 55€ 1387 Audiophonics HAT has a sound signature that keeps me questioning any delta-sigma implementation. Yep - even a Katana.
Don't waste your time on the next great thing! I'd be surprised if something
new would be worth the effort.
Enjoy.
PS: I also owned Benchmark DACs and RME DACs. The great thing about these devices was that I could nicely resell these. Try that with a so called "DIY" device.
If Allo offers us a new firmware and a driver update we all can and probably will try (and enjoy) it.
***
To me DAC chips are very close to become commodity products.
The recipes of making a good DAC device are known to most manufacturers.
Even 50$ devices claim femto seconds jitter and <-110dB THD/N nowadays.
IMO since years we keep spinning in circles around here about the same subjects.
And to be honest. I could have lived with my 51xx DAC forever. Yep. I'm getting older. Actually my 1543 DAC sounded really great more then 10years ago. And it still does.
And my 55€ 1387 Audiophonics HAT has a sound signature that keeps me questioning any delta-sigma implementation. Yep - even a Katana.
Don't waste your time on the next great thing! I'd be surprised if something
new would be worth the effort.
Enjoy.
PS: I also owned Benchmark DACs and RME DACs. The great thing about these devices was that I could nicely resell these. Try that with a so called "DIY" device.
Last edited:
"""Of course"""" there's no such thing .... and Katana had it since day 1.
OhhhK. Sharp as a pebble.
Great.
Than you guys are still behind on OSF-B. (Not that we didn't discuss this
long ago. )
Thanks, looked at your blog, appreciate the detailed instructions.
It looks like I can just turn off USB with this:
0:2 -- Controls all four PI-USB ports (not the Ethernet)
I'm using ethernet instead of wifi, but maybe I'll try wifi and see if I can tell a difference one way or another.
Randy
Hi Mark,
I wonder if you could point me to the post(s) where these were explained please?
Many thanks
John
John,
The stuff about FIFO_Pi and the dual ES9038Q2M dac is spread around over several pages as I found things while working on the dac. I doubt I found everything possible, but I found things that reduced distortion enough that I could start to narrow in on remaining issues. One post in that series is back around: https://www.diyaudio.com/forums/pc-...essions-tweaks-mods-hints-40.html#post5803366
Regarding Katana, I tried a few things shortly after it came out as I was a reviewer at the time. I used a simple +-15v and +5 linear supply to power katana and the output stage. The 'measurement' output board sounds better, IMHO (calling the other one 'sound quality' was likely a mistake, also IMHO) The Allo isolator was used, and about 100uf of Wima mks4 film caps in parallel from each of the +-15v rails to ground. Cap values in each parallel block were 10+10+22+33+33 (all values in uf). The linear power supply was modded a bit as follows: All rectifier diodes were replaced with Hexfred types, and the +15v regulator was changed to LT1083. https://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-332.html#post5623968
One thing that is very important during all the testing is that power line and ground contamination issues were handled by the best power conditioner there is, Monster HTPS 7000 MK II. Some discussion of that going on right now in the Blowtorch thread: https://www.diyaudio.com/forums/the...wtorch-preamplifier-iii-2930.html#post5929120
A Benchmark DAC-3 was used as the preamp, and a Benchmark AHB2 was the power amp.
All equipment was powered though the one power conditioner. With very clean grounds and power ground isolation between all my equipment, there is no need for line-level transformers to improve sound quality. In fact, I can always hear the added distortion of signal transformers, but in most situations where people don't have the best power conditioners line transformers will probably help sound quality.
The foregoing is sort of brief pointer(s) to some of what I wrote about. The stuff about FIFO_Pi and Ian's dac you may have to sort through. I think it ended with the boards separated, the input to the output stage filtered via a folded ribbon cable, and 3x of ADM7150 type regulators used to power dac. Input voltage to the LDOs was increased to 8v, and the one for AVCC was loaded with an additional 33-ohm resistor. Those changes improve LDO regulation and dac sound quality, although opamp-regulator AVCC supplies still sound better. I also tried feeding FIFO_Pi with upsampled DSD from HQplayer by feeding I2S from a USB board into the I2S pins on FIFO_Pi GPIO input connector. OPA1612 for opamps. That whole combination of things gave the best sound, and no need for line transformers. However, I know further improvements could have been made. I stopped working on it because it does not have dual AVCC power supply inputs, one for each L/R channel, so the stereo effect is weak. I will say that despite knowing more could have been done with the output stage to improve it, it still sounded quite good.
Another area where there may still be problems is something I also mentioned, which is that the sockets for plug-in clocks soon loose their contact springiness which can be expected to increase jitter. Also, u.fl cables are not good if they are installed and removed more than once or twice. They also quickly loose their initial contact force and even at best, the braid shield of the RF cables is only crimped on by two flimsy tabs, not soldered or properly crimped. Thus, u.fl cables may be well suited to model planes and such, but not so suitable for dacs where we care about low femto-second jitter. SMA connectors are much better, good up to 18GHz, and they needed to be screwed together and properly torqued to assure a dependable ground connection. Even being slightly less that fully torqued to spec can introduce audible jitter, so you can probably guess what u.fl connectors and dip clock sockets must be doing.
The stuff about FIFO_Pi and the dual ES9038Q2M dac is spread around over several pages as I found things while working on the dac. I doubt I found everything possible, but I found things that reduced distortion enough that I could start to narrow in on remaining issues. One post in that series is back around: https://www.diyaudio.com/forums/pc-...essions-tweaks-mods-hints-40.html#post5803366
Regarding Katana, I tried a few things shortly after it came out as I was a reviewer at the time. I used a simple +-15v and +5 linear supply to power katana and the output stage. The 'measurement' output board sounds better, IMHO (calling the other one 'sound quality' was likely a mistake, also IMHO) The Allo isolator was used, and about 100uf of Wima mks4 film caps in parallel from each of the +-15v rails to ground. Cap values in each parallel block were 10+10+22+33+33 (all values in uf). The linear power supply was modded a bit as follows: All rectifier diodes were replaced with Hexfred types, and the +15v regulator was changed to LT1083. https://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-332.html#post5623968
One thing that is very important during all the testing is that power line and ground contamination issues were handled by the best power conditioner there is, Monster HTPS 7000 MK II. Some discussion of that going on right now in the Blowtorch thread: https://www.diyaudio.com/forums/the...wtorch-preamplifier-iii-2930.html#post5929120
A Benchmark DAC-3 was used as the preamp, and a Benchmark AHB2 was the power amp.
All equipment was powered though the one power conditioner. With very clean grounds and power ground isolation between all my equipment, there is no need for line-level transformers to improve sound quality. In fact, I can always hear the added distortion of signal transformers, but in most situations where people don't have the best power conditioners line transformers will probably help sound quality.
The foregoing is sort of brief pointer(s) to some of what I wrote about. The stuff about FIFO_Pi and Ian's dac you may have to sort through. I think it ended with the boards separated, the input to the output stage filtered via a folded ribbon cable, and 3x of ADM7150 type regulators used to power dac. Input voltage to the LDOs was increased to 8v, and the one for AVCC was loaded with an additional 33-ohm resistor. Those changes improve LDO regulation and dac sound quality, although opamp-regulator AVCC supplies still sound better. I also tried feeding FIFO_Pi with upsampled DSD from HQplayer by feeding I2S from a USB board into the I2S pins on FIFO_Pi GPIO input connector. OPA1612 for opamps. That whole combination of things gave the best sound, and no need for line transformers. However, I know further improvements could have been made. I stopped working on it because it does not have dual AVCC power supply inputs, one for each L/R channel, so the stereo effect is weak. I will say that despite knowing more could have been done with the output stage to improve it, it still sounded quite good.
Another area where there may still be problems is something I also mentioned, which is that the sockets for plug-in clocks soon loose their contact springiness which can be expected to increase jitter. Also, u.fl cables are not good if they are installed and removed more than once or twice. They also quickly loose their initial contact force and even at best, the braid shield of the RF cables is only crimped on by two flimsy tabs, not soldered or properly crimped. Thus, u.fl cables may be well suited to model planes and such, but not so suitable for dacs where we care about low femto-second jitter. SMA connectors are much better, good up to 18GHz, and they needed to be screwed together and properly torqued to assure a dependable ground connection. Even being slightly less that fully torqued to spec can introduce audible jitter, so you can probably guess what u.fl connectors and dip clock sockets must be doing.
Last edited:
Thanks for the info Mark. Lots to work through now...
So far I have only explored the power supplies on the Katana, settling on AMB Audio Sigma11 supplies for all the 5 volt rails, and the Sigma 22 for the pos/neg15. They are discrete supplies and unconditionally stable into any capacitive load. Previously the best I had tried were Jung super-regs but these are compromised by the low ESR capacitors on most digital equipment and have to be "de-tuned" to be made stable, but still better than LT3042/5 designs.
John
So far I have only explored the power supplies on the Katana, settling on AMB Audio Sigma11 supplies for all the 5 volt rails, and the Sigma 22 for the pos/neg15. They are discrete supplies and unconditionally stable into any capacitive load. Previously the best I had tried were Jung super-regs but these are compromised by the low ESR capacitors on most digital equipment and have to be "de-tuned" to be made stable, but still better than LT3042/5 designs.
John
First, let's answer some questions and make a few general comments...
@alexkosha, consider the Allo USBBridge Signature to be an audiophile RPi with:
- A lot of high-quality regulators and filtering separately supplying each chip's power rails
- An added Ethernet chip to resolve the on-chip IO Ethernet/USB conflicts along with filtering on the Ethernet input to reduce noise fed back into the USBBrdgSig board
- An additional USB connection configured to provide a high-quality USB connection for use with a USB-connected DAC similar to the output side of Uptone Audio Regens and ISORegens and the various Sonore Rendus and SoTM USB-output devices
- An RPi 3+ compute module as the unit's brains for RPi kernel compatibility
- A fully RPi-compliant mounting footprint (within the required larger motherboard due to all of the added functions and features) including an RPi-compliant GPIO header
As such it will only draw a little more than a standard RPi (partly due to the additional loss of the linear regulators and partly due to the additional chips). It does not have a built-in WiFi interface, so you have to add that via a USB WiFi dongle or if you use wired Ethernet like I do and want WiFi access for control purposes, a connection to a WiFi router or access point in your network.
I use mine so far as a I2S-connected DAC endpoinat only, with a wired Ethernet connection. I use either an Allo Shanti 5V/3.5A output or a somewhat modified K&K Audio 12 watt (5V/2A) low-voltage power supply to power mine. Voltages into the USBBridge Sig have ranged from 4.9V to 5.2V per these small voltage monitor I will use to watch the voltage over time:
bayite 3 Wire 0.36" DC 0~30V Digital Voltmeter Gauge Tester Blue LED Display Panel Mount Car Motorcycle Battery Monitor Volt Voltage Meter with Reverse Polarity Protection Pack of 5: Amazon.com: Industrial & Scientific
Distro is piCorePlayer, player software Squeezelite, as I've used since I got into RPi computer audio a few years back.
On use of the Isolator, I'll post about that soon.
@John Luckins, many thanks for your comments, useful as always. I have AMB listed in the possible regulator source list for the Katana in the first post of this thread. I haven't tried them myself, but have read a number of good comments on them from people I trust like you. Good to know you put them in the same category as Jung super-regs, if I read your comments correctly.
There's a lot of other stuff to respond to here in recent posts, but I want to be clear on what I post and how I say it. I find it useful to distinguish between various 'types' of comments I post to be sure they are taken in an appropriate context... Those types are:
- Facts... things you'll find in a datasheet and are generally not in dispute.
- Impressions or experiences... How I preceive something(s), typically in the context of listening. Always filtered through my own personal hearing mechanisms (ears and brain) and sonic preferences.
- Opinion or judgement... MY assessment of what I've perceived. To be clear, this is mine, not anyone else's and not fact. To be taken as such, just like anyone else's.
- Purpose and/or context and/or perspective... Why I'm doing something and/or within what framework I'm doing it and/or from what viewpoint.
AND as a preamble to some following posts (today and in the next few days) the purpose of MY work with and most of my posts about the Allo Katana in this thread are to see how good I can get the Katana to sound while playing mostly my ripped RBCD digital library. I have a very few higher-res PCM and DSD recordings, but they are not what I play every day.
That may change at some point, I am interested in high-quality upsampling and DSD conversion, but currently don't have the infrastructure to do such.
Having provided that context and background.... let's get into...
Greg in Mississippi
@alexkosha, consider the Allo USBBridge Signature to be an audiophile RPi with:
- A lot of high-quality regulators and filtering separately supplying each chip's power rails
- An added Ethernet chip to resolve the on-chip IO Ethernet/USB conflicts along with filtering on the Ethernet input to reduce noise fed back into the USBBrdgSig board
- An additional USB connection configured to provide a high-quality USB connection for use with a USB-connected DAC similar to the output side of Uptone Audio Regens and ISORegens and the various Sonore Rendus and SoTM USB-output devices
- An RPi 3+ compute module as the unit's brains for RPi kernel compatibility
- A fully RPi-compliant mounting footprint (within the required larger motherboard due to all of the added functions and features) including an RPi-compliant GPIO header
As such it will only draw a little more than a standard RPi (partly due to the additional loss of the linear regulators and partly due to the additional chips). It does not have a built-in WiFi interface, so you have to add that via a USB WiFi dongle or if you use wired Ethernet like I do and want WiFi access for control purposes, a connection to a WiFi router or access point in your network.
I use mine so far as a I2S-connected DAC endpoinat only, with a wired Ethernet connection. I use either an Allo Shanti 5V/3.5A output or a somewhat modified K&K Audio 12 watt (5V/2A) low-voltage power supply to power mine. Voltages into the USBBridge Sig have ranged from 4.9V to 5.2V per these small voltage monitor I will use to watch the voltage over time:
bayite 3 Wire 0.36" DC 0~30V Digital Voltmeter Gauge Tester Blue LED Display Panel Mount Car Motorcycle Battery Monitor Volt Voltage Meter with Reverse Polarity Protection Pack of 5: Amazon.com: Industrial & Scientific
Distro is piCorePlayer, player software Squeezelite, as I've used since I got into RPi computer audio a few years back.
On use of the Isolator, I'll post about that soon.
@John Luckins, many thanks for your comments, useful as always. I have AMB listed in the possible regulator source list for the Katana in the first post of this thread. I haven't tried them myself, but have read a number of good comments on them from people I trust like you. Good to know you put them in the same category as Jung super-regs, if I read your comments correctly.
There's a lot of other stuff to respond to here in recent posts, but I want to be clear on what I post and how I say it. I find it useful to distinguish between various 'types' of comments I post to be sure they are taken in an appropriate context... Those types are:
- Facts... things you'll find in a datasheet and are generally not in dispute.
- Impressions or experiences... How I preceive something(s), typically in the context of listening. Always filtered through my own personal hearing mechanisms (ears and brain) and sonic preferences.
- Opinion or judgement... MY assessment of what I've perceived. To be clear, this is mine, not anyone else's and not fact. To be taken as such, just like anyone else's.
- Purpose and/or context and/or perspective... Why I'm doing something and/or within what framework I'm doing it and/or from what viewpoint.
AND as a preamble to some following posts (today and in the next few days) the purpose of MY work with and most of my posts about the Allo Katana in this thread are to see how good I can get the Katana to sound while playing mostly my ripped RBCD digital library. I have a very few higher-res PCM and DSD recordings, but they are not what I play every day.
That may change at some point, I am interested in high-quality upsampling and DSD conversion, but currently don't have the infrastructure to do such.
Having provided that context and background.... let's get into...
Greg in Mississippi
Last edited:
Isolators...
First, some facts... Isolator are simply devices that transmit electric signals without sharing a common ground between the input side and the output side. Transformers are the most common isolators we encounter (for AC isolation in power supplies), with optoisolators being another common type in many electronic devices.
In the context of digital audio signals, isolators are very high-speed devices (110-150 Mbps for some commonly used for I2S isolation) that are designed to take a digital signal input and produce a faithful replica at the output without a ground connection between the input side and output side to limit noise from the input side contaminating the ground plane of the output side. Common vendors are Silicon Labs (Used in Allo's, Ian's, and Soekris products, among others) and NVE (favorites of audio engineer John Swenson for some of his early and middle work).
Because of the mechanisms used to produce the above result, they do 2 things to the signal... produce skew or delay in the signal both from input to output and channel to channel and add jitter to the output signals. Neither are desired. Because of this, IMHO, the best use for isolators is where there is an output-side master clock and either synchronous (real-time) or asynchronous (non-real-time, read the data into a buffer and reclock it to the output from that buffer) reclocking is performed on the I2S signals to reduce jitter before use by the DAC. Without this reclocking you will hear a slightly poorer signal from a jitter and timing perspective due to the isolator.
This is even more important in the RPi world, especially with a stock RPi, when the RPi clock is used to generate the I2S signal ('slave' mode). Then the I2S signal is inherently noisy (due to the RPi processing noise) and jittery (due to the RPi clock not being of a digital audio signal related clocking frequency, so non-integer division is used to generate the master clock in the RPi, an inherently jittery process). SO (again in my experience and IMHO) whether or not you use an isolator, with an RPi and a 'slave' mode DAC (most of them), you'll get better results using a reclocker like Allo's Kali or Ian's FiFoPi or similar.
Why would you use isolators? IMHO, basically to prevent noise from the input side from getting to the output side via the ground connection. IMHO, this is most important when the input side is noisy (a computer, like an RPi) and likely less so when the source is quiet, like a good SD Card player like the SDTrans384. In fact, with a good quiet I2S source (like an SD Card player), some (like the super-tweaks on TirNaHiFi.org) have found improvements bypassing isolators (such as those at the I2S inputs of the Soekris DAM 1021 DAC boards).
Before using Allo's Isolator 1.2 with the Katana (and their original Isolator with the Boss 1.2), MOST of my experiences with isolators have been in setups where the final stage before the DAC board/card included asynchronous reclocking. AND the source was a standard RPi 2B, generally with good power supplies.
In these cases with various I2S-connected DACs, my experiences have been that the isolators lowered the background noise I perceived, though sometimes with a perception of lowered dynamics. Either going to a better (read that as a lower-output-impedance) power supply or just slightly increasing the volume recovered the dynamics. My experiences and judgements of what I perceived as greater dynamics without the isolators always seemed to be greater background noise and hash that the isolators reduced. AND my preferences are HIGHLY skewed towards lower background noise.
I found a similar experience replacing DC-DC converters with linear regulators in various circuits, where noise caused by those converters sounded more dynamic. BUT ultimately, to my ears, noise is just noise.
In the case of the Allo Katana and Boss 1.2 DACs, the I2S output, which is generated via a 'master' mode process where a better quality clock signal from the DAC board is fed into the RPi to control the generation of the I2S signal going into the DAC, there is no reclocking performed after the isolators on the signals going both ways. So skew and jitter added by the isolators is not 'corrected' before the generation of the I2S signals in the RPi and again not corrected before the DAC uses the I2S signal.
Is that important? IMHO, it is a matter of tradeoffs and preferences... IF the I2S source is noisy (stock RPi with a typical power supply), the noise reduction by breaking the ground connection MAY be more important than the slightly more jittery I2S signal. AND my experience and assessment is that is true with stock RPis and typical to very good power supplies.
BUT that is definitely my opinion and others may judge and decide differently, especially with different systems.
My bottom line assessment... with stock RPi's, adding Allo's Isolator 1.2 is an improvement EVEN with really good power supplies on the stack. YMMV... My 'audio friend' was almost ambivilent on adding the Isolator 1.2, just barely coming down on the side of it being a good thing.
Whew! I know that was a lot... if you think it is too much, just TLDR it!
BUT this leads to a report from my 'audio friend...
Greg in Mississippi
First, some facts... Isolator are simply devices that transmit electric signals without sharing a common ground between the input side and the output side. Transformers are the most common isolators we encounter (for AC isolation in power supplies), with optoisolators being another common type in many electronic devices.
In the context of digital audio signals, isolators are very high-speed devices (110-150 Mbps for some commonly used for I2S isolation) that are designed to take a digital signal input and produce a faithful replica at the output without a ground connection between the input side and output side to limit noise from the input side contaminating the ground plane of the output side. Common vendors are Silicon Labs (Used in Allo's, Ian's, and Soekris products, among others) and NVE (favorites of audio engineer John Swenson for some of his early and middle work).
Because of the mechanisms used to produce the above result, they do 2 things to the signal... produce skew or delay in the signal both from input to output and channel to channel and add jitter to the output signals. Neither are desired. Because of this, IMHO, the best use for isolators is where there is an output-side master clock and either synchronous (real-time) or asynchronous (non-real-time, read the data into a buffer and reclock it to the output from that buffer) reclocking is performed on the I2S signals to reduce jitter before use by the DAC. Without this reclocking you will hear a slightly poorer signal from a jitter and timing perspective due to the isolator.
This is even more important in the RPi world, especially with a stock RPi, when the RPi clock is used to generate the I2S signal ('slave' mode). Then the I2S signal is inherently noisy (due to the RPi processing noise) and jittery (due to the RPi clock not being of a digital audio signal related clocking frequency, so non-integer division is used to generate the master clock in the RPi, an inherently jittery process). SO (again in my experience and IMHO) whether or not you use an isolator, with an RPi and a 'slave' mode DAC (most of them), you'll get better results using a reclocker like Allo's Kali or Ian's FiFoPi or similar.
Why would you use isolators? IMHO, basically to prevent noise from the input side from getting to the output side via the ground connection. IMHO, this is most important when the input side is noisy (a computer, like an RPi) and likely less so when the source is quiet, like a good SD Card player like the SDTrans384. In fact, with a good quiet I2S source (like an SD Card player), some (like the super-tweaks on TirNaHiFi.org) have found improvements bypassing isolators (such as those at the I2S inputs of the Soekris DAM 1021 DAC boards).
Before using Allo's Isolator 1.2 with the Katana (and their original Isolator with the Boss 1.2), MOST of my experiences with isolators have been in setups where the final stage before the DAC board/card included asynchronous reclocking. AND the source was a standard RPi 2B, generally with good power supplies.
In these cases with various I2S-connected DACs, my experiences have been that the isolators lowered the background noise I perceived, though sometimes with a perception of lowered dynamics. Either going to a better (read that as a lower-output-impedance) power supply or just slightly increasing the volume recovered the dynamics. My experiences and judgements of what I perceived as greater dynamics without the isolators always seemed to be greater background noise and hash that the isolators reduced. AND my preferences are HIGHLY skewed towards lower background noise.
I found a similar experience replacing DC-DC converters with linear regulators in various circuits, where noise caused by those converters sounded more dynamic. BUT ultimately, to my ears, noise is just noise.
In the case of the Allo Katana and Boss 1.2 DACs, the I2S output, which is generated via a 'master' mode process where a better quality clock signal from the DAC board is fed into the RPi to control the generation of the I2S signal going into the DAC, there is no reclocking performed after the isolators on the signals going both ways. So skew and jitter added by the isolators is not 'corrected' before the generation of the I2S signals in the RPi and again not corrected before the DAC uses the I2S signal.
Is that important? IMHO, it is a matter of tradeoffs and preferences... IF the I2S source is noisy (stock RPi with a typical power supply), the noise reduction by breaking the ground connection MAY be more important than the slightly more jittery I2S signal. AND my experience and assessment is that is true with stock RPis and typical to very good power supplies.
BUT that is definitely my opinion and others may judge and decide differently, especially with different systems.
My bottom line assessment... with stock RPi's, adding Allo's Isolator 1.2 is an improvement EVEN with really good power supplies on the stack. YMMV... My 'audio friend' was almost ambivilent on adding the Isolator 1.2, just barely coming down on the side of it being a good thing.
Whew! I know that was a lot... if you think it is too much, just TLDR it!
BUT this leads to a report from my 'audio friend...
Greg in Mississippi
Last edited:
My Audio Friend's latest report...
Some of you may have read about modifying an RPi (2B or early 3) to remove the onboard DC-DC conversion stage generating the 1.8V and 3.3V rails and instead using low-noise linear regulation for those rails. That is the topic of this thread where a special board available from LDOVR.com is used to make that conversion:
Mezzanine Power board for Raspberry Pi
I've done that and posted there and other places that my experience with using an RPi 2B modified this way is that it provided a significantly greater sense of background silence and improved detailing and resolution compared to a stock RPi. At that time, these were my best RPi sources and started to sound ALMOST as good as using my SDTrans384 SDCard player.
Recently, Allo released their USBBridge Signature. See my description of it a few posts back OR just go to Allo's website or Vendor thread on this and their Shanti power supply and Revolution DAC:
Shanti Dual LPS 5V/3A , 5V/1.5A
I posted my gushingly positive impressions of using them to replace one of these modified RPis under an Ian GB stack in post #588 and a stock RPi under my Katana setup in post #594.
IMHO, these take my RPi sourced-setups to a new high level!
Based on my impressions and prodding, my 'audio friend' ordered one and just installed it under his Katana stack today. Here's his impressions and assessment:
"As stated, I received the USBridge-Signature yesterday. I had ordered it with a microSD card loaded with MoOde Audio and it came with ver. 6.1. I booted up the USBridge-Signature with an Edimax USB dongle and did an in place update to ver. 6.2. An in place update to ver 6.3 isn’t available so, after struggle with Etcher on macOS Catalina - there’s a security permissions issue - I used another machine with Mojave and flashed the microSD card with a fresh download of ver. 6.3.
I listened to the system with the Raspberry Pi 3B+ with 6.3 under the Katana stack. I then put the USBridge-Signature into the system by simply removing and replacing the Raspberry Pi3B+ with the USBridge-Signature - the DAC HAT being the Allo Katana. Here’s my set up:
- slightly modified K&K low voltage power supply for the Pi and dirty side of the isolator board, the power connection on the GPIO header
- DIYINHK LT3042 with active rectification powering the Katana DAC board and the clean side of the isolator board, power through the provided on board connector on the DAC board
- DIYINHK LT3042 powering the Katana microcontroller board, power directly to the super cap
- Twisted Pear Placid HD Bipolar Power Supply set for +/-14.1 volts and 300mA per rail directly connected to the Katana op amp board via pads made available by removing the 4-pin header connector between the opamp board and the microcontroller board, the DC-DC convertor on the microcontroller board being turned off and regulation for the rails on the bottom of the microprocessor board thereby being completely bypassed
I heard an immediate across the board improvement. I easily heard things/details I’ve never heard before or things I knew were there before but were somehow obscured, became readily apparent. For example, muffled or jumbled lyrics were now clearly enunciated. Bass has more authority, and everything is more dynamic. OK - good!"
All good. BUT based on my previous post on Isolators, because the USBBridge Signature is a significantly lower-noise environment than a stock (or even my linear-regulation-modified) RPi AND in the case of using it with a Katana, the Isolator's added jitter and skew is not being addressed by a following reclocking stage, maybe the need for the Isolator would be diminished. He went on:
"I pulled the isolator board out.
If you know what jitter sounds like or brings or adds to a digital sonic presentation - well, that’s what goes away with removing the isolator. You now clearly hear the jitter the isolator adds. It’s gone. More detail, easier to access the musicians/singers. Images have more character and are more delineated. The position of things didn’t really change, but the images bleed together much less. A few recordings that were somewhat marginal before, yet I liked them, are now, well,…really good and make much more sense from a recording perspective. It’s more liquid, but not in a syrupy way. The music being more fluid might be a better description.
A friend said he’s been listening to Cream - Live at Royal Albert Hall. A good recording. I switched the phase back and forth - you can now hear which mics, etc. were out of phase in the mix relative to others - no way you could hear that before with this level of specificity.
All in all, much more life like, and accessing the music is much more quickly and readily accomplished with the USBridge-Signature than with the Raspberry Pi3B+. And where before, I could take or leave the isolator - it just seemed a little different - their’s no way I want it in there with the USBridge-Signature.
Magnitude-wise, in my guestimation, the USBridge-Signature is on par with the power supply change on the opamp board, and while it does many of the things a power supply change brings to the table, it does other or different things as well. “Clock related” things, things you don’t normally get with a power supply change.
Need more listening now!!!!!"
Ok, so now we have one answer to the question "Do you need an Isolator 1.2 for a Katana on a USBBridge Signature".
Pumped by his experience, I spent a few minutes today and pulled the Isolator 1.2 in my Katana stack. AND heard a similar set of SQ improvements. AND it is not going back into that stack!
Does this mean I would not recommend the Isolator 1.2 for all situations. Nope, same as before, I prefer it in place with a stock RPi. YMMV. BUT if you are looking for a significant upgrade for your Katana stack, consider a USBBridge Sig and if you have an Isolator 1.2, pull it out.
BTW, mounting the Katana without the Isolator board on the USBBridge Sig takes some care due to some filter caps on that board underneath the RPi footprint that will short against the Supercaps on the bottom of the Katana DAC board if you don't allow for that. See my comments in post #683 of the Allo Shanti/USBBridge Sig/Revolution thread linked above. I needed to add another fiber washer under each post for the additional clearance.
I'll finish tonight with a threat... more long posts coming!
Greg in Mississippi
Some of you may have read about modifying an RPi (2B or early 3) to remove the onboard DC-DC conversion stage generating the 1.8V and 3.3V rails and instead using low-noise linear regulation for those rails. That is the topic of this thread where a special board available from LDOVR.com is used to make that conversion:
Mezzanine Power board for Raspberry Pi
I've done that and posted there and other places that my experience with using an RPi 2B modified this way is that it provided a significantly greater sense of background silence and improved detailing and resolution compared to a stock RPi. At that time, these were my best RPi sources and started to sound ALMOST as good as using my SDTrans384 SDCard player.
Recently, Allo released their USBBridge Signature. See my description of it a few posts back OR just go to Allo's website or Vendor thread on this and their Shanti power supply and Revolution DAC:
Shanti Dual LPS 5V/3A , 5V/1.5A
I posted my gushingly positive impressions of using them to replace one of these modified RPis under an Ian GB stack in post #588 and a stock RPi under my Katana setup in post #594.
IMHO, these take my RPi sourced-setups to a new high level!
Based on my impressions and prodding, my 'audio friend' ordered one and just installed it under his Katana stack today. Here's his impressions and assessment:
"As stated, I received the USBridge-Signature yesterday. I had ordered it with a microSD card loaded with MoOde Audio and it came with ver. 6.1. I booted up the USBridge-Signature with an Edimax USB dongle and did an in place update to ver. 6.2. An in place update to ver 6.3 isn’t available so, after struggle with Etcher on macOS Catalina - there’s a security permissions issue - I used another machine with Mojave and flashed the microSD card with a fresh download of ver. 6.3.
I listened to the system with the Raspberry Pi 3B+ with 6.3 under the Katana stack. I then put the USBridge-Signature into the system by simply removing and replacing the Raspberry Pi3B+ with the USBridge-Signature - the DAC HAT being the Allo Katana. Here’s my set up:
- slightly modified K&K low voltage power supply for the Pi and dirty side of the isolator board, the power connection on the GPIO header
- DIYINHK LT3042 with active rectification powering the Katana DAC board and the clean side of the isolator board, power through the provided on board connector on the DAC board
- DIYINHK LT3042 powering the Katana microcontroller board, power directly to the super cap
- Twisted Pear Placid HD Bipolar Power Supply set for +/-14.1 volts and 300mA per rail directly connected to the Katana op amp board via pads made available by removing the 4-pin header connector between the opamp board and the microcontroller board, the DC-DC convertor on the microcontroller board being turned off and regulation for the rails on the bottom of the microprocessor board thereby being completely bypassed
I heard an immediate across the board improvement. I easily heard things/details I’ve never heard before or things I knew were there before but were somehow obscured, became readily apparent. For example, muffled or jumbled lyrics were now clearly enunciated. Bass has more authority, and everything is more dynamic. OK - good!"
All good. BUT based on my previous post on Isolators, because the USBBridge Signature is a significantly lower-noise environment than a stock (or even my linear-regulation-modified) RPi AND in the case of using it with a Katana, the Isolator's added jitter and skew is not being addressed by a following reclocking stage, maybe the need for the Isolator would be diminished. He went on:
"I pulled the isolator board out.
If you know what jitter sounds like or brings or adds to a digital sonic presentation - well, that’s what goes away with removing the isolator. You now clearly hear the jitter the isolator adds. It’s gone. More detail, easier to access the musicians/singers. Images have more character and are more delineated. The position of things didn’t really change, but the images bleed together much less. A few recordings that were somewhat marginal before, yet I liked them, are now, well,…really good and make much more sense from a recording perspective. It’s more liquid, but not in a syrupy way. The music being more fluid might be a better description.
A friend said he’s been listening to Cream - Live at Royal Albert Hall. A good recording. I switched the phase back and forth - you can now hear which mics, etc. were out of phase in the mix relative to others - no way you could hear that before with this level of specificity.
All in all, much more life like, and accessing the music is much more quickly and readily accomplished with the USBridge-Signature than with the Raspberry Pi3B+. And where before, I could take or leave the isolator - it just seemed a little different - their’s no way I want it in there with the USBridge-Signature.
Magnitude-wise, in my guestimation, the USBridge-Signature is on par with the power supply change on the opamp board, and while it does many of the things a power supply change brings to the table, it does other or different things as well. “Clock related” things, things you don’t normally get with a power supply change.
Need more listening now!!!!!"
Ok, so now we have one answer to the question "Do you need an Isolator 1.2 for a Katana on a USBBridge Signature".
Pumped by his experience, I spent a few minutes today and pulled the Isolator 1.2 in my Katana stack. AND heard a similar set of SQ improvements. AND it is not going back into that stack!
Does this mean I would not recommend the Isolator 1.2 for all situations. Nope, same as before, I prefer it in place with a stock RPi. YMMV. BUT if you are looking for a significant upgrade for your Katana stack, consider a USBBridge Sig and if you have an Isolator 1.2, pull it out.
BTW, mounting the Katana without the Isolator board on the USBBridge Sig takes some care due to some filter caps on that board underneath the RPi footprint that will short against the Supercaps on the bottom of the Katana DAC board if you don't allow for that. See my comments in post #683 of the Allo Shanti/USBBridge Sig/Revolution thread linked above. I needed to add another fiber washer under each post for the additional clearance.
I'll finish tonight with a threat... more long posts coming!
Greg in Mississippi
The problem with the USBBridge signature is that it was already outdated at the time of launch (due to unexpected PI4 launch) .
The Allo intention with this product is quite understandable. Bad luck, bad timing, though.
There are IMO several weak spots on the Allo product (PI3 based) and I'm not talking about
the power rails. These weak spots alone would hold me back buying such a device.
And now with the still rather new PI4 at hand, at still $40, I have to say there is no looking back to good old PI3 times.
The PI4 comes with so many new features, a fast PCI-E lane, separate networking stack, USB3, 64bit support asf.
Of course you'll also have onboard WLAN and BT5.0. On top you can run whatever (64bit)-OS you want. And more...
But as usual, these are just my 2cts.
And don't forget. It's the DAC who's supposed to make the music and to fight issues induced by the transport.
If the transport causes sound degradation it's IMO the DAC which needs to be questioned.
My Katana has a Pi4 feeding it nowadays. And you bet, it's gonna stay that way for a while.
I keep the fingers crossed that Allo sells enough units to become profitable on the SBC side - to be able to stay in the market.
If the DAC business is a tough business, I'd say the computer/SBC business is ten times tougher.
Enjoy.
The Allo intention with this product is quite understandable. Bad luck, bad timing, though.
There are IMO several weak spots on the Allo product (PI3 based) and I'm not talking about
the power rails. These weak spots alone would hold me back buying such a device.
And now with the still rather new PI4 at hand, at still $40, I have to say there is no looking back to good old PI3 times.
The PI4 comes with so many new features, a fast PCI-E lane, separate networking stack, USB3, 64bit support asf.
Of course you'll also have onboard WLAN and BT5.0. On top you can run whatever (64bit)-OS you want. And more...
But as usual, these are just my 2cts.
And don't forget. It's the DAC who's supposed to make the music and to fight issues induced by the transport.
If the transport causes sound degradation it's IMO the DAC which needs to be questioned.
My Katana has a Pi4 feeding it nowadays. And you bet, it's gonna stay that way for a while.
I keep the fingers crossed that Allo sells enough units to become profitable on the SBC side - to be able to stay in the market.
If the DAC business is a tough business, I'd say the computer/SBC business is ten times tougher.
Enjoy.
Last edited:
My Katana has a Pi4 feeding it nowadays. And you bet, it's gonna stay that way for a while.
Could you describe your RPi4 configuration with Katana more accurately?
How did you solve the power supply ? Do you use WiFi or Ethernet ?
Any additional changes to piCorePlayer ?
Usbridge Sig is not competing with any RPI
If you care about CPU speed , ethernet and type of connection it has to CPU...then a RPI4 is a good idea. Yeap we measured the noise...and its in mV (USB) , i2s lines are even noisier than rpi3
If you want to listen to music on a device that subjectively sounds better and that can also show improvement on machine testing (THD+N) , at up to 512 DSD, USbridge Sig is rather good choice.
Please folks , understand that everyone has an opinions ,sometimes with no understanding of hardware or electronics .
If you care about CPU speed , ethernet and type of connection it has to CPU...then a RPI4 is a good idea. Yeap we measured the noise...and its in mV (USB) , i2s lines are even noisier than rpi3
If you want to listen to music on a device that subjectively sounds better and that can also show improvement on machine testing (THD+N) , at up to 512 DSD, USbridge Sig is rather good choice.
Please folks , understand that everyone has an opinions ,sometimes with no understanding of hardware or electronics .
@Greg Thanks for the follow-up above from you and your friend. I'm using the Isolator 1.2 with the Boss HAT in one of secondary roon end-points and works well enough. I haven't popped for the USBBridge Sig yet, as I'm thoroughly enjoying my heavily modded Ian DAC stack. Old music sounds new again. No more fatiguing sound.
Greg, thank you a lot for all these details. It is exceptionally well placed pieces of info. I’ll get that board and I’ll test it soon.
However, I have some challenge with its size. My enclosure has limited space and don’t know how I’ll integrate it in.
Another think is how to measure Voltage at designator. I could not find any way to get it from Pi. So, I assumed some drop and adjusted based on PSU out.
Do you see any test pads that can be used for voltage measurements on USBridge Sig board?
I'm running (my desktop/development system ) currently a
* PI4
* Shanti (I do also use some other PS such as iPower, since I do no like the Shanti power-up/power-down behavior - IMO there's a relay missing - see IANs supply)
* Katana (I do also use some other less demanding DACs)
Then I'm running
* Arch Linux ARM - I recently finished building my full 64bit installation (rt-kernel and userland)
* squeezelite
* onboard WLAN - IMO better then any wired solution
* and my tunings - pretty much as described on my blog.
Cheers