diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   The Well Regulated Power Supply (http://www.diyaudio.com/forums/power-supplies/212476-well-regulated-power-supply.html)

andrea_mori 11th May 2012 01:29 PM

The Well Regulated Power Supply
 
Hi all,

thread starts in parallel with this other
http://www.diyaudio.com/forums/digit...jitter-45.html
to discuss the right way to build an adeguate power supply for that project,
but also to discuss the right way to build high performance regulated power supply for DIY DAC.

In the last time I worked on project simulation and finally I developed my idea. The project at a glance:
- 1° stage: CLCRC filter
- 2° stage: bootstrapped series with pre-regulation (a revised Jung super regulator)
- 3° stage: feed-forward shunt regulation
- hi-grade components

The project is op-amp based, so I investigated if there was reason to build a discrete operational amplifier, but finally I concluded that it is better to use a monolithic OA, such as AD797, AD825, OPA627, OPA132 or so. Another story about Stage IV or buffer.

Now I'm working on simulation, then I'll share my project and I'll start to build a prototype.

Any comments or suggestions would be greatly appreciated.

Andrea

qusp 12th May 2012 03:16 AM

nice one!! I would have started a thread, but I couldnt do it justice with the time I have available. well i'm certainly interested. can we please have at least 2-4 (minimum) of the regulators with power transistors or followers? perhaps this could be a snap in module for the standard lower current version.

i'm not sure you quite get where I was coming from though, we or at least I, was talking about networks of 25-50 regulators for the ever-growing complexity of the modern dac and it has little to do with the fifo project; which only requires a single regulated DC voltage input or at the most an additional dedicated supply for the clock board and its flipflops. the greater system under discussion involves far more than this.

the highest performance, yet compact solution i've come across so far is more like the Flea type regulator involving a filtered very low noise monolithic reference with a high performance opamp driver/divider. networks of 50 super regulators and matching raw DC supplies are neither needed nor viable IMO. For me the system involves a combination of LiFePO4 batteries and chargers, IC based post regulators, AC input preregs, flea type regs and shunt regs.

PET-240 12th May 2012 10:57 AM

Interesting qusp that you note that different supplies in different sections yield the best sound, whilst this is a can of worms because most everyone will argue for the sake of it, which is mostly useless diatribe, can you expand on what you have found works best in your system?
Please note others reading this, I'm not wanting to have arguments about who likes what, listening impressions are subjective, but at least having it documented we may see patterns in differing powersupplies and what sound they produce, this may provide a further way to tune the system to attain what the listener feels suits them best.
Andrea, crack on brother, have read most of your site with wonder, thanks for sharing your findings, look forward to your discoveries.
OT, may chat to you about dc coupling yet........

Thanks,

Drew.

dtproff 12th May 2012 03:28 PM

I am curious what the design goals for this project are? What noise levels are you seeking and why the extra design complexity?

I normally try and use an architecture that has an inherently good PSRR so that the PSU is not as big an issue. That being said, things like sag for punch on the bass are much harder to measure and describe/specify.



Tony

andrea_mori 12th May 2012 04:58 PM

Quote:

Originally Posted by dtproff (Post 3020937)
I am curious what the design goals for this project are? What noise levels are you seeking and why the extra design complexity?

I normally try and use an architecture that has an inherently good PSRR so that the PSU is not as big an issue. That being said, things like sag for punch on the base are much harder to measure and describe/specify.



Tony

Not a particular complexity:
- 1rst stage, CLCRC filter, to well decoupling regulator from AC and dreastically reduce ripple
- 2nd stage, series regulator, to maintain a steady voltage decoupling supply from load
- 3rd stage, shunt regulator, to reduce noise
- hi-grade components, to add themselves minimum noise, eg. feed-forward shunt needs a precise and stable current sensor resistor, so I prefer to use a 0,02 ppm metal foil resistor rather than a metal oxide resistor

BTW all your comments and suggestions are welcome, i'll keep them in the right consideration developing this project.

Andrea

raul_77 12th May 2012 05:25 PM

Hi Andrea,

This are a some years old design of a low noise power supply, the interesting are the topology of the pcb:

Super Buffer

But this article has confused me. They get very low noise and distortion with a simple zener regulated power supply. Well, test is easy and cheap.

Link para Electrónicos

Click on the links and see the articles in english, also see the page two of the LM833 pdf.

Regards.

qusp 12th May 2012 05:34 PM

Quote:

Originally Posted by dtproff (Post 3020937)
I am curious what the design goals for this project are? What noise levels are you seeking and why the extra design complexity?

I normally try and use an architecture that has an inherently good PSRR so that the PSU is not as big an issue. That being said, things like sag for punch on the base are much harder to measure and describe/specify.



Tony

for me it was sheer number, only about 25-30% were of absolute highest performance. but i'm talking about a 4 balanced channel sabre dac (each chip needs 7 supplies) with the fifo i2s buffer, dual master clock and spdif input receiver, multichannel usb->i2s convertor, 2 stereo balanced IV stages, MCU, monitor, remote control (wifi and RS232), charger and control circuitry. no getting around the complexity of such a system

PET-240 13th May 2012 07:24 AM

Mmmmmm remote via wifi and 232. Very interesting.... Can you pm a link please?
Yes that is going to be one very full box....... Can see why regs are important, esp those slimline ones you have......
I don't have any design ability regarding sand powersupplies so I shall read, learn and probably ask what others consider dumb questions.....I was taught that the dumbest question is the one not asked.....

analog_sa 14th May 2012 08:53 AM

Maybe this thread's title should be changed - wasn't the idea to discuss specifically regulator types suitable for high performance digital circuits?

I have little doubt that at least the subjective requirements for analogue circuit regulators are very different from those meant to power digital circuits. How this relates to objective parameters is anyone's guess. Per example i wouldn't seriously consider flea-type regulators for analogue circuits but they may be great for digital. Dac chips are a special case as they often require separate supplies for the analogue and digital parts.

Obviously some compromise will be needed due to the number of regulators and space constraints. Maybe a diy version of the popular discrete 3-terminal replacements would be a good idea with shunt regulators only in really critical stages.

andrea_mori 14th May 2012 12:57 PM

Quote:

Originally Posted by analog_sa (Post 3022654)
Maybe this thread's title should be changed - wasn't the idea to discuss specifically regulator types suitable for high performance digital circuits?

I have little doubt that at least the subjective requirements for analogue circuit regulators are very different from those meant to power digital circuits. How this relates to objective parameters is anyone's guess. Per example i wouldn't seriously consider flea-type regulators for analogue circuits but they may be great for digital. Dac chips are a special case as they often require separate supplies for the analogue and digital parts.

Obviously some compromise will be needed due to the number of regulators and space constraints. Maybe a diy version of the popular discrete 3-terminal replacements would be a good idea with shunt regulators only in really critical stages.

I agree with you that different circuits requires different power supplies.
Btw if you had an ideal supply you could use it either for analogue than digital circuits. Obviously ideal supply does not exists and I have no claim to build it.
But if we accept some compromises we can try to build a good regulator as for analogue as for digital circuits.

I believe that a CLCRC filter is a good way to decoupling regulator from AC, either for analogue than for digital, so I think we can use it in both case.

The second stage, I planned to use a series regulator like Jung super regulator, but now I'm thinking seriously to implement a simple series regulator based on op amp like AD797 or LT1018 and precision reference like AD580 or AD780, preceded by a monolithic regulator like LT1085. From LT1085 datasheet: Line Regulation 0.025%, Ripple Rejection 75 db, RMS Output Noise 0.003%. I think this three terminal regulator is a good choise for pre-regulation and can supply directly op amp, whereas AD797 has a PSRR of at least 120 db. Also AD797 has an Input voltage noise of 0.9 nV/sqrtHz, that guarantees a very low noise level in regulation.
So, I think that's a good starting point either from analogue than digital.

The shunt stage, I'm thinking to use a feed-forward shunt that guarantees a supply noise rejection of about 26 db at lower frequencies.

Maybe this could be a good design to use either for analogue than for digital stage.

I think that in addition to a good design, components choise make the difference.

Andrea


All times are GMT. The time now is 11:32 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2