I am testing different PSUs with the same DAC (that has spec of THD+N at -107 1Khz/Unweigted o db)
Using a cheap SMPS , we get -98
Using Nirvana SMPS with earthed output we get -107
Using Shanti....we get -108.1!! Earthing/not does not change the number
I have seen previously on audiosciencereview that a better PS does not change THD numbers. I will ask Amir to test that finding with Shanti .
Funny thing is that noise of both Shanti and Nirvana are close at 0-20Khz(200nv to 80nv so low its almost immaterial) using AP and even at 5Mhz (we see noise from 4-12uV , normal in higher bandwith and with Spectrum analyzer that has less precision)
Can Soundecheck be right about impedance and actually being able to measure the effect on THD+N ??
I will study more , but so far we have measurable differences using good PSUs on a good..ish DAC
Using a cheap SMPS , we get -98
Using Nirvana SMPS with earthed output we get -107
Using Shanti....we get -108.1!! Earthing/not does not change the number
I have seen previously on audiosciencereview that a better PS does not change THD numbers. I will ask Amir to test that finding with Shanti .
Funny thing is that noise of both Shanti and Nirvana are close at 0-20Khz(200nv to 80nv so low its almost immaterial) using AP and even at 5Mhz (we see noise from 4-12uV , normal in higher bandwith and with Spectrum analyzer that has less precision)
Can Soundecheck be right about impedance and actually being able to measure the effect on THD+N ??
I will study more , but so far we have measurable differences using good PSUs on a good..ish DAC
There is another aspect.
You'll step over it if you look at the lab psu datasheets.
I brought it up earlier.
Load Transients
We're on DC. Yep. Every load, digital or analog, has different requirements - obviously. A single PSU sees all of them at once.
Luckily there are plenty of buffers attached to the rails. They are supposed to cope with it.
However. All these suckers do interfere somehow - they sit on the same rail
and there's just one supplier.
Load Transient Recovery Time
It basically defines the time that it takes to recover from a load attack.
If you read this document, you'll see that the quality of a PS can very well associated to this parameter.
100us is considered a good value in above article.
The earlier referenced Tektronix even comes with <50us and shows a much smaller settling band voltage.
For sure it's not all about noise.
I'm pretty sure Nirvana/Shanti will do well on that one.
Just try to measure it.
We IMO simply need to have more parameters to get into a position to explain what we hear and experience and how it can be associated to power supplies.
You'll step over it if you look at the lab psu datasheets.
I brought it up earlier.
Load Transients
We're on DC. Yep. Every load, digital or analog, has different requirements - obviously. A single PSU sees all of them at once.
Luckily there are plenty of buffers attached to the rails. They are supposed to cope with it.
However. All these suckers do interfere somehow - they sit on the same rail
and there's just one supplier.
Load Transient Recovery Time
It basically defines the time that it takes to recover from a load attack.
If you read this document, you'll see that the quality of a PS can very well associated to this parameter.
100us is considered a good value in above article.
The earlier referenced Tektronix even comes with <50us and shows a much smaller settling band voltage.
For sure it's not all about noise.
I'm pretty sure Nirvana/Shanti will do well on that one.
Just try to measure it.
We IMO simply need to have more parameters to get into a position to explain what we hear and experience and how it can be associated to power supplies.
Just a remark.
I very well understand that the transient recovery document talks about regulation and feedback related implications.
But even without feedback, as implemented in Shanti, there'll be transient related impact. And that might even differ from load (static + dynamic) to load situation.
The document outlines a nice example of the mobile phone industry, where the huge transients of a "power amplifier" (headphone amp/DAC-Avcc/outputstage?) impacts the "bias" (base load - digital) stability. We basically could call it a feedback loop: analog->digital.
No question. Regulation with feedback slows things down.
A PS without feedback should then be at least as fast as these high quality lab PS examples.
Again. I think it'd be very interesting to see what's going on.
I very well understand that the transient recovery document talks about regulation and feedback related implications.
But even without feedback, as implemented in Shanti, there'll be transient related impact. And that might even differ from load (static + dynamic) to load situation.
The document outlines a nice example of the mobile phone industry, where the huge transients of a "power amplifier" (headphone amp/DAC-Avcc/outputstage?) impacts the "bias" (base load - digital) stability. We basically could call it a feedback loop: analog->digital.
No question. Regulation with feedback slows things down.
A PS without feedback should then be at least as fast as these high quality lab PS examples.
Again. I think it'd be very interesting to see what's going on.
Last edited:
Yes there will be transient impact .
First check this white paper
We can clearly see how transients are reduced with a supercap on the output. The supercaps have no recovery time. They supply the transient with no delay (as long as in the limits of its capacitance)
Second , on the Shanti PCB we have about 47.000 uF of capacitance = 47mF (3A rail ) = 0.047F . So any transient will first draw from supercaps and then from massive bank. Meanwhile yes the voltage will slightly dip and magnetic the transformer will start supplying it.
From and architecture point of view , Shanti was designed since the beginning with transients in mind. Local decoupling (reservoir) will make any transient look like a small gentle curve ...because of its architecture.
First check this white paper
We can clearly see how transients are reduced with a supercap on the output. The supercaps have no recovery time. They supply the transient with no delay (as long as in the limits of its capacitance)
Second , on the Shanti PCB we have about 47.000 uF of capacitance = 47mF (3A rail ) = 0.047F . So any transient will first draw from supercaps and then from massive bank. Meanwhile yes the voltage will slightly dip and magnetic the transformer will start supplying it.
From and architecture point of view , Shanti was designed since the beginning with transients in mind. Local decoupling (reservoir) will make any transient look like a small gentle curve ...because of its architecture.
Ioan,
I'm very excited about your new supply.
My experiences and those of others using LiFePO4, Supercap, and Ultracap supplies have been very positive. Most have suggested a similar mechanism for the good sonics with these supplies, that of these charged devices having an inherent very low output impedance that does not rely on active circuits with global feedback. BUT most implementations have also had some limitations:
- The drawback of slowly declining output voltage for a limited on-time before requiring recharge when using pre-charged, large-capacitance devices and relying on a relatively low discharge rate OR ssue of post-regulation to maintain a constant output voltage, but blunting the sonic improvements.
- The complexity and cost of using 2 banks of the devices with one bank being charged while the other is in use.
- The possible and sometimes perceived sonic hit of constant charging to a pre-set voltage.
AND many of these solutions have required a pre-charge period before the target voltages were reached and they could be used.
To be fair, ALL of these implementations can be made to work very well AND sound very good. BUT they either have the cons as above or the con of increased costs.
Your solution seems to provide the benefits with fewer drawbacks and greater utility / convenience than most other implementations using similar charged devices. AND at a reasonable cost.
ALSO my recent moves in my Katana power setup going from the no-longer available OPC dual 5V 4||LT3042 output boards to MPAudio dual 5V 3||LT3045 output boards to 2 MPAudio 5V 6||LT3045 output boards brought good increases in overall SQ with each step. Each successive supply should lower the output impedance and noise, so I expect your Shanti, which you already optimized for these parameters, should provide similar benefits.
I am looking forward to being able to purchase one. AND even moreso, a +-15V version, which is a more difficult and likely more expensive proposition using these charged devices.
Greg in Mississippi
P.S. I reported on some experiments with Ultracap supplies in posts #s 74 & 126 of this thread: Getting the best out of Allo.com's new Katana DAC...
Some of those other users of large-capacity charged devices:
- Ian Canada's LiFePO4 supply: Develop ultra capacitor power supply and LiFePO4 battery power supply
- Uptone Audio's LPS-1 & LPS-1.2: UltraCap™ LPS-1.2 – UpTone Audio
- The DIY'ers at TirNaHiFi.org's use of float-charged LiFePO4 supplies and more recently float-charged Ultracap supplies as realized in their community DIY DAC project using a SDCard player and their own implementation of a PCM1794 DAC chips similar to the DDDAC project... these are spread across a number of threads on their boards in the General and Computer Audio sections
- A couple of heroic implementations of a dual-bank Ultracap supply: R2R DA M1 Ultracap powered build | Headphone Reviews and Discussion - Head-Fi.org
- John Keny's LiFePO4 and Ultracap supplies as featured at USB audio | Ciunas Audio and also discussed on TirNaHiFi.org and here: https://pinkfishmedia.net/forum/threads/supercapacitor-power-supply-tour.223681/page-4
I'm very excited about your new supply.
My experiences and those of others using LiFePO4, Supercap, and Ultracap supplies have been very positive. Most have suggested a similar mechanism for the good sonics with these supplies, that of these charged devices having an inherent very low output impedance that does not rely on active circuits with global feedback. BUT most implementations have also had some limitations:
- The drawback of slowly declining output voltage for a limited on-time before requiring recharge when using pre-charged, large-capacitance devices and relying on a relatively low discharge rate OR ssue of post-regulation to maintain a constant output voltage, but blunting the sonic improvements.
- The complexity and cost of using 2 banks of the devices with one bank being charged while the other is in use.
- The possible and sometimes perceived sonic hit of constant charging to a pre-set voltage.
AND many of these solutions have required a pre-charge period before the target voltages were reached and they could be used.
To be fair, ALL of these implementations can be made to work very well AND sound very good. BUT they either have the cons as above or the con of increased costs.
Your solution seems to provide the benefits with fewer drawbacks and greater utility / convenience than most other implementations using similar charged devices. AND at a reasonable cost.
ALSO my recent moves in my Katana power setup going from the no-longer available OPC dual 5V 4||LT3042 output boards to MPAudio dual 5V 3||LT3045 output boards to 2 MPAudio 5V 6||LT3045 output boards brought good increases in overall SQ with each step. Each successive supply should lower the output impedance and noise, so I expect your Shanti, which you already optimized for these parameters, should provide similar benefits.
I am looking forward to being able to purchase one. AND even moreso, a +-15V version, which is a more difficult and likely more expensive proposition using these charged devices.
Greg in Mississippi
P.S. I reported on some experiments with Ultracap supplies in posts #s 74 & 126 of this thread: Getting the best out of Allo.com's new Katana DAC...
Some of those other users of large-capacity charged devices:
- Ian Canada's LiFePO4 supply: Develop ultra capacitor power supply and LiFePO4 battery power supply
- Uptone Audio's LPS-1 & LPS-1.2: UltraCap™ LPS-1.2 – UpTone Audio
- The DIY'ers at TirNaHiFi.org's use of float-charged LiFePO4 supplies and more recently float-charged Ultracap supplies as realized in their community DIY DAC project using a SDCard player and their own implementation of a PCM1794 DAC chips similar to the DDDAC project... these are spread across a number of threads on their boards in the General and Computer Audio sections
- A couple of heroic implementations of a dual-bank Ultracap supply: R2R DA M1 Ultracap powered build | Headphone Reviews and Discussion - Head-Fi.org
- John Keny's LiFePO4 and Ultracap supplies as featured at USB audio | Ciunas Audio and also discussed on TirNaHiFi.org and here: https://pinkfishmedia.net/forum/threads/supercapacitor-power-supply-tour.223681/page-4
Last edited: