which regulated power design type for a DAC?

[Nada]

Power regulation is widely considered to be crucial for a DAC. Its such a complex area for  a beginner I need help.

 

I guess ideally there will be separate regulated supplies for the:

 

signal input (5v and 3.3V clock = 2)

DAC digital stages and DAC Analogue stages (so for the TDA1541 +5V / -5V  / -15V )

Analogue stage (L @ R = 2 x2 )

 

So nine regulators are indicated. Cost is going to be a factor x9.

 

This guide looks excellent:

 

Comparison of low power supply regulators. includes measurements and controlled listening tests (New Jersey Audio Club) by Jack Walton,Linear Audio Volume 4

 graphs in color :

Fig 5, positive regulators Line Rejection

Fig 6, negative regulators Line Rejection

Fig 7, positive regulators Output Noise

Fig 8, positive regulators Output Impedance

Fig 9, negative regulator Output Impedance

 

 

What would you choose?

 

 

 

 

[Zaphod Beeblebrox]

It's hard to go past cascaded LM317/LM337 regs.

Let's say you need +15 Volts.

Start with a (say) LM317 set at 20 Volts. Then run an LM317 set at 15 Volts to achieve your final result.

LM317/337 regs are amongst the best bang for your Buck and are dead easy to use. FAR superior to most other 3 terminal regs.

[Nada]

It's hard to go past cascaded LM317/LM337 regs.

Let's say you need +15 Volts.

Start with a (say) LM317 set at 20 Volts. Then run an LM317 set at 15 Volts to achieve your final result.

LM317/337 regs are amongst the best bang for your Buck and are dead easy to use. FAR superior to most other 3 terminal regs.

 

I wonder if running three is even better then two or does the slew rate get clobbered?

[Zaphod Beeblebrox]

I wonder if running three is even better then two or does the slew rate get clobbered?

IMO, there's not much point, as the regulation is already pretty good with two cascaded regs. You could run a capacitance multiplier after the main filter caps though. Cheap and cheerful and does an excellent job.

[hochopeper]

317/337 with well designed PCB and local decoupling capacitance would be my choice most of the time, except for XO and DAC analogue current output supply. Some more creative buffered voltage reference and super regulators can be useful there though getting them to work at the lower voltages can be tricky.

 

The designs of capacitor banks that add large amounts of inductance between the storage and the load don't make a lot of sense IMO. With caps it's far more about how you use them and certainly not a more is better situation.

[A J]

There are better solutions available now than the LM317/337 which IMHO have a less than best of class noise floor for sensitive digital stages. Have a gander around DIYAUDIO.COM where you will find plenty of ideas, there are a number of kitset PCB's such as the Salas shunt, and links to ready made solutions - particularly with the lower voltage (3.3v etc) 

Zaphs concept of cascaded regs makes sense, although I wouldn't cascade from a differential +/- supply such as that used for driving an opamp, you want to keep these very tightly matched.

 

I'm surprised Nada that you call yourself a "beginner" - you know more than you let on haha, don't pull the wool over our eyes....

 

Heres a reasonable 4X low voltage PSU for around $100 - I know other DIY-ers here have used these to good effect.

Edited October 27, 2013 by A J

[hochopeper]

I think I've said it before but I should say it again .... those DIYINHK kits won't get anywhere near the numbers they're claiming for performance. They are pulling datasheet specs and printing it on the PCB like they think it's some sort of badge of honour, they're not even following all of the datasheet recomendations in the PCB design, not something you can mod your way around later either. The PCB design and decoupling are poor. Those regulators even with the decorative heatsinks won't be able to carry the power that they quote either. Do the sums on the thetaJC and include the thermal properties of the heatsink and those things will do nothing for the performance, the part is designed to use the PCB as heatsink yet they haven't given sufficient thermal mass for anything but maybe 2/10ths to 3/10ths of the 1A they say you should be able to run off them. The noise performance also drops off as you put more current through them. The TPS7A is a VERY good part, but that's a very poor implementation of it.

[Luc]

Interruption:  God I wish I knew what you lot were talking about(seriously) as it's just eye watering stuff at the moment.

 

 

End of interruption.

 

[Cafad]

Interruption:  God I wish I knew what you lot were talking about(seriously) as it's just eye watering stuff at the moment.

 

 

End of interruption.

 

Good to hear it's not just me then Luc.  I got all the way to post #4 before I had to reach for the paracetamol.  Don't stop though guys, we'll both be back to take notes once the Panadol kicks in.

[Nada]

The listening test results for the regulators in the first post  are very annoying.

 

The three best sounding regs in that particular test set up bore no correlation with any of the measurements.

 

The differences in the listening test were marked but unpredictable.

 

Looks like the only thing to do is try 10 different regs in any application  and see what sounds best for that set-up 

[hochopeper]

Sorry Luc and Cafad and others who I'm about to give stomach ulcers for overuse of painkillers, but here goes my braindump on regulator selection.

 

I've got a copy of the Linear Audio book where that test was done and written up and includes outcomes of the listening tests. It's an interesting read but the listening tests are done with a line stage with low PSRR as an attempt to best expose the PSU. It's an interesting read but not something that can be directly applied to the situation where Nada is talking about. The measurements are excellent but really demonstrate how oversimplified published specifications for audio gear is, the consumer really doesn't always have access to the information that tells them anything at all about the performance of the gear they're buying.

 

I've used LM317, LT1083, LT1963, TPS7a4700 and Sjostrom SSR01 with success before ... the 317 and 1963 are great value, while the others can benefit from some intelligent PCB design and selection of decoupling and bulk capacitance, the TPS7a series really wants some care in the implementation of the PCB, as much as people might tell you otherwise, I've soldered these by hand with a normal soldering iron so they're still DIY friendly and you don't need to have more than a pair of tweezers, some good solder and flux. The Salas reg needs some care with the parts selection and the load and the heatsinking selection. The super regulator is prone to be unstable if you have a load that is overly capacitive so it also requires some care in that regard. There are a lot of regulators that are mentioned in audiophile circles that are actually very poor when given any sort of relevant testing. It's a tough gig selecting PSU for the very sensitive parts of a DAC and for the last few % I'd be going with other choices for oscillator. The tendency for audiophile designers to put shunt regulation everywhere is a personal bugbear of mine, for a stable load (like oscillator and DAC analogue supply for example) they're a poor choice as the balance of noise performance vs output impedance is all upside down and they're often worse performance and add more heat to the enclosure. 

 

It is important to consider the bandwidth of the load and the fact that the load's PSRR changes with freq also so it really needs to be considered in conjunction with the PSU's PSRR and then somehow rationalised to input noise and how much that will correspond with on the output to get a mix of priorities in selecting the regulator. IMO impedance of connection between the capacitance local to the load makes the biggest impact on PSU impedance as seen by the load. The grounding of the entire system is critical to noise performance. 

 

For Nada, I've got some other suggestions for DAC digital supply and clocks, I'd rather keep them offline though. For analogue I'd use Sjostrom super regulators for the last few % of performance or LT1963 and it's mirror -ve part (can't remember part number off the top of my head) for a moderate cost option. Or adjustable LM317/LM337 for +ve/-ve with the option to add sjostrom regulators later and use those regulators as pre-regulation stage. If I was designing something from scratch I might do it slightly differently but I'd have to give it a lot more thought so from the options above that's my thoughts. It's easy to understand how some DACs become a PSU + DAC two box design ... the PSU gets bulky.

 

Chris

Edited October 27, 2013 by hochopeper

[Nada]

[Addicted to music]

http://www.altronics.com.au/index.asp?area=item&id=K5501

Here's a cheap way to get +\- 15v cheaply and its local. $20.00 max. This kit was also used in a SC preamp, it was also in a SC article in how to modify it to get other other output voltages you need.

Addition: it uses the 317/337 reg as previously mentioned by ZB and supplied with the SC ultra low distortion preamp using BB 2134.

Edited October 27, 2013 by pchan

[acg]

Nice thread Nada, very timely for me too.  I hope you don't mind a quick interjection...

 

I have some digital components that I wish to supply with linear power.  The first is a modded USB card that needs external power to function, and seems to be responsive to the cheap lab linear power that I can put towards it at the moment (compared to an ATX psu).  Even though it is a USB card at no stage will the USB3 ports be powered (I think USB spec allows up to 900mA) and my dac does not accept nor look for power.  I have measured the maximum draw by the card at 0.02A (20mA).  What recommendations would you guys field for a high quality 5V 100mA linear supply.  I've looked at the Sjostrom and it looks doable.  The modder of the USB card recommends battery before linear power but I just think batteries have warts on 'em.  I reckon low noise, high PSRR and a fast transient response are highly desirable.

 

Nada, as a fan of Audio-GD I'm surprised you did not add this puppy or this puppy to your list (guys - feel free to discuss them as well).

 

 

Cheers,

 

Anthony

Edited October 30, 2013 by acg