Right, which is why it might be good to verify the noise measurement FFT is actually in units of 'volts per root hertz,' and not just 'volts.' I only ask to check because it is possible for the units to come out wrong on an AP analyzer if enough care is not taken with the FFT settings.
Interesting test.
We hooked the Katana to our LPS (uController and Isolator)
We get around 110Thd+N normally in this config m with 2x 5V smps (ones we are selling on the website)
However prior to this we tried different types of power including IFI , Nirvana (older PCB) and linear. While we had fluctuations , its was around + - 0.5db (max)
With the new LPS , THD+N jumped to -111.250
This is the first time we observe a better THD+N with better PSU. (unmistakably )
We hooked the Katana to our LPS (uController and Isolator)
We get around 110Thd+N normally in this config m with 2x 5V smps (ones we are selling on the website)
However prior to this we tried different types of power including IFI , Nirvana (older PCB) and linear. While we had fluctuations , its was around + - 0.5db (max)
With the new LPS , THD+N jumped to -111.250
This is the first time we observe a better THD+N with better PSU. (unmistakably )
First of all. Impressive progress...
...on the scope.
Congrats.
I'm wondering though if these basic "audio device" characteristics are that relevant at all on power supplies.
Of course a certain (noise) threshold need to be kept.
And it's nice to read about 100nV noise (whatever kind of noise under whatever conditions are meant here).
The big Q: Does it have to be 100nV!?!? Is it relevant at all?
I rather doubt it.
Especially inside such a "pre-stage" PS it could very well mean massive overkill, considering the numerous filters and
regulation stages following on the rails inside the audio devices, which clean up
the tiny mess (which is basically proven by your (Allo) own audio device measurements btw.)
Like the results of our wild tweakers around here, the actual impact on sound seems more like a coincidence to me.
At least it can't be explained properly. No matter how much prose gets written and no matter how many devices/parts get swapped.
What is it than!?!?
How about transient performance? E.g. Transient recovery time, rise times etc.
Basically how long it'll take to refill the buckets.
Parameters like that change with the parasitics around. Basically every mR,mH etc. on the rail counts.
Of course there are other parameters that add to the game like feedback loops.
As a matter of fact there is equipment around to measure these transient parameters. Nobody seems to be using that gear though.
And I think it's not hard to guess that e.g. an iFi with a 1.5m cable and very low buffer caps will perform much worse on that transient account
than several hundered/thousand uF low esr (super)caps right in front of the audio device. That's been documented/reported numerous times.
I havn't seen a lot of discussions around that subject. That's why I bring it up. I have to admit though I never spent much time in the DIYA-PS section.
Would be nice to see power supplies getting tested and valued on potentially more relevant parameters.
Enjoy.
...on the scope.
Congrats.
I'm wondering though if these basic "audio device" characteristics are that relevant at all on power supplies.
Of course a certain (noise) threshold need to be kept.
And it's nice to read about 100nV noise (whatever kind of noise under whatever conditions are meant here).
The big Q: Does it have to be 100nV!?!? Is it relevant at all?
I rather doubt it.
Especially inside such a "pre-stage" PS it could very well mean massive overkill, considering the numerous filters and
regulation stages following on the rails inside the audio devices, which clean up
the tiny mess (which is basically proven by your (Allo) own audio device measurements btw.)
Like the results of our wild tweakers around here, the actual impact on sound seems more like a coincidence to me.
At least it can't be explained properly. No matter how much prose gets written and no matter how many devices/parts get swapped.
What is it than!?!?
How about transient performance? E.g. Transient recovery time, rise times etc.
Basically how long it'll take to refill the buckets.
Parameters like that change with the parasitics around. Basically every mR,mH etc. on the rail counts.
Of course there are other parameters that add to the game like feedback loops.
As a matter of fact there is equipment around to measure these transient parameters. Nobody seems to be using that gear though.
And I think it's not hard to guess that e.g. an iFi with a 1.5m cable and very low buffer caps will perform much worse on that transient account
than several hundered/thousand uF low esr (super)caps right in front of the audio device. That's been documented/reported numerous times.
I havn't seen a lot of discussions around that subject. That's why I bring it up. I have to admit though I never spent much time in the DIYA-PS section.
Would be nice to see power supplies getting tested and valued on potentially more relevant parameters.
Enjoy.
Last edited:
Would it be correct to say it is an audiophile SBC, compatible with all Allo devices which are used so far with either RPI or Sparky and a replacement for Sparky?
Matt
Hi Klaus
I think thats the million dollar questions . is 100nV relevant ? I , like you, will jump and say no....but mind you , you are right a second time. ""considering the numerous filters and
regulation stages following on the rails inside the audio devices, which clean up
the tiny mess (which is basically proven by your (Allo) own audio device measurements btw.) ""
Katana has a lot of filtering. We have at least one LDO per rail and of course plenty of PI filters on a each rail.
So why the change of THD+N .
I have attached here the thd+n taken today. To me it looks like the THD part has not changed (check the same graph from audioschiencereview) what changes is the N part.
I have a theory , that noise on PSU will always find a way on the output. For example resonance of PI filters will amplify certain frequencies.Tracks and capacitors are PI filters. Vias and caps another PI filter..in fact you have hundreds of PI filters (undumped)
Most have resonance in hundreds of Khz some in audioband
If less noise is feed to a DAC...less noise on the output.
I will check tomorrow with the Boss.
I think thats the million dollar questions . is 100nV relevant ? I , like you, will jump and say no....but mind you , you are right a second time. ""considering the numerous filters and
regulation stages following on the rails inside the audio devices, which clean up
the tiny mess (which is basically proven by your (Allo) own audio device measurements btw.) ""
Katana has a lot of filtering. We have at least one LDO per rail and of course plenty of PI filters on a each rail.
So why the change of THD+N .
I have attached here the thd+n taken today. To me it looks like the THD part has not changed (check the same graph from audioschiencereview) what changes is the N part.
I have a theory , that noise on PSU will always find a way on the output. For example resonance of PI filters will amplify certain frequencies.Tracks and capacitors are PI filters. Vias and caps another PI filter..in fact you have hundreds of PI filters (undumped)
Most have resonance in hundreds of Khz some in audioband
If less noise is feed to a DAC...less noise on the output.
I will check tomorrow with the Boss.
Its a fully featured SBC using the RPI compute module with an accent on USB clean power rails , dejittering and buffering clocks. RPI module is feed all power rails (4.7V , 3.3v , 1.8V )using independent LDOs . Power to USB hub is cleaned 3 times , power to the clock of USB 3 times and buffer sits between them. Ethernet is brought using an Axis IC (capable of Gb to USB 3.0 but of course we are using 2.0 hub) . Axis IC is powered using 6 independent power rails (USB 3.3 and 1.2, Ethernet 3.3 and 1.2 and core logic3.3 and 1.2)
Even SD card power rail gets a LDO ..
At last Ethernet GB is cleaned using the latest CMC from pulse thats is qualified for auto environments
What we have done different this time...is capacitor dampening. We used over 30 electrolytic on board to dampen the PI filters (before and after inductor) and also as reservoirs of energy to decouple each rail from another .
Decoupling was beefed up as per the great book of Henry W. Ott (Electromagnetic compatibility engineering)
Basically a SBC (again based on RPI compute module) with clean...everything.
cdsgames,
thank you,
and the existing USBridge or Digione would clean this SBC further?
Does it support higher rate DSD?
Matt
prety good huh.. hopefully these filters will not take a toll on the compactness of this device.
hopefully the OS will be Rpi compatible build. I prefer the Pi distros becuase they are stable and get continous update.
Usbridge HAT (circuit) is included on the PCB. If spdif is needed , then Digione is needed.
It will not be "compact" not possible to have the cake and eat it too.
Yes we are striving to have it compatible with all distri. Still I this time I think its a bit premature to talk about it . Will reopen the discussion in 1 month.
Thanks.
Anyway, I have seen some reviews powering MC of Katana with battery pack. Is it recomended? Do you have measurements for this?
Does anyone here tried to use the ifi Micro iUSB3.0 in the whole Katana?
No WIFI , but compatible with WIFI and BT dongles that have better specs then RPI 3B+