Optimum Decoupling of Digital ICs

[Guido Tent]

Also worth a look are LT1963 regs.
Very low noise (14uV rms 10Hz~100kHz)

Terry

let us see, that is about 50 nV/Sqrrt Hz

at least better than the LM317 stuff.....

regards

 

[jewilson]

Logic Power Supply Noise

Guido,

If you had looked at my post #186 you would have seen the low noise regulator that was attached for the 8412.

There was no point in telling Setmenu to implement a circuit that he does not under stand. The idea was to point him to some basic information on regulators and let him do the homework. I think he was doing well.

Also, really let us see, “you said we need about 5 nV/Sqrrt Hz,” for a noise on the digital supplies, are you kidding? Getting that kind of noise performance from a supply powering digital logic, when there are clocks and asynchronous data in
the 2-10 MHz range, that is super difficult. Even 50 nV/Sqrrt Hz supply noise will be very difficult without superior grounding decoupling and one very good layout. Next, digital logic families are designed to run in an environment where there is a lot of noise.

While it is important to have low noise for the receiver there is some point of diminishing returns. Obviously if the signal is small at the receiver inputs, it really helps to have the best noise performance we can get or we need to switch to the ASE/EBU or optical interface. Trying to obtain that type noise performance is not always practical; it is a good design goal.

Maybe you have some design ideas that you would like to share with us on the best way to implement a low noise receiver and power supply.

Cheers

 

[fmak]

Re: Logic Power Supply Noise

Guido,

If you had looked at my post #186 you would have seen the low noise regulator that was attached for the 8412.

There was no point in telling Setmenu to implement a circuit that he does not under stand. The idea was to point him to some basic information on regulators and let him do the homework. I think he was doing well.

Also, really let us see, “you said we need about 5 nV/Sqrrt Hz,” for a noise on the digital supplies, are you kidding? Getting that kind of noise performance from a supply powering digital logic, when there are clocks and asynchronous data in
the 2-10 MHz range, that is super difficult.
------------------------------------------------------------------------
He is not kidding. His clock with PS meets this over a 1 MHz bandwidth connected to a Sony SCD670. Guido qualifies this with a 100kHz babdwidth in his post.

Yes, there is a need to pount out that the LTs are better, and yes, I have measured threm.

 

[jewilson]

Fmak,

I have not seen any performance information on Guido clocks and power supplies. I have heard numerous people speak very positive about them, including Jocko. In addition, I agree with you that the LT regulators or good and better than most. However, to get the kind of noise performance Guido was talking about requires a discreet high performance design and a dam good layout and components.

Now when Guido stated that about power supplies, we all have different ideas of what he’s talking about. Just like, when I was try to explain some basics to someone else.

So, if you trying to accomplish getting a medium scale digital high speed logic power supply to perform to the levels Guido was stating good luck. On the other hand if you have a few pico gates and few SOIC flip flop maybe that is possible depend on the logic types, layout and other good design practices. In addition, just because we have this super regulator does not mean it will perform so in a crummy environment.

Cheers

 

[setmenu]

So my 'low noise' ADP3301 regs are not low noise after all
From what I have been reading here, they now seem positively noisy



Setmenu

 

[fmak]

So my 'low noise' ADP3301 regs are not low noise after all
From what I have been reading here, they now seem positively noisy

-------------------------------------------------------------------------
It's a great pity that so many low noise regulators touted as such by vendors have such high noise levels. Many clock units actually use 7805s, some of which are not bad (40 uV or so) whilst others are rubbish.

 

[jean-paul]

Fmak, was it you that that measured the 7805 from JRC with the outcome that it was very quiet ?

From memory the 7805's from Motorola are quiet too.

 

[fmak]

Fmak, was it you that that measured the 7805 from JRC with the outcome that it was very quiet ?

From memory the 7805's from Motorola are quiet too.

-----------------------------------------------------------------
Yes, I have measured several with low noise, about 30-40 uV/1 MHz banfdwidth.

 

[setmenu]

So it seems that 20-40uv is considered low noise by reg manufacturers?

And if I wish to go lower I would need to build some discreet design, [which would need to be small!] but without technical experience or a suitably optimized circuit to install them in, I am unlikely to see this high performance realized?

Therefore back to square one, my noisy 'low noise' regs.
I hope the dac does not sound too awful then, assuming it works in the first place



Setmenu

 

[fmak]

So it seems that 20-40uv is considered low noise by reg manufacturers?

And if I wish to go lower I would need to build some discreet design, [which would need to be small!] but without technical experience or a suitably optimized circuit to install them in, I am unlikely to see this high performance realized?

Therefore back to square one, my noisy 'low noise' regs.
I hope the dac does not sound too awful then, assuming it works in the first place

------------------------------------------------------------------------------
Audiocom sells a supereregulator with low noise. However, when driving a clock, this produces harmonics all over the place! You really need to measure; caveat emptor!

 

[jewilson]

It is possible to use some OSCON caps with a PI network to lower the noise of a 7805 regulator. As Fmak stated the many of the Linear Technology have much better performance and if your going to run this from batteries you should look for a low drop out regulator.

You might look at these regulators from LT LT1083 and LT1033. These are older design but still have good performance they are also pin compatible with the 317 and 337.

If your looking to build a super low noise regulator hear is a thread that discusses several designs

http://www.diyaudio.com/forums/showthread.php?threadid=20937&goto=newpost

 

[setmenu]

Hi jewilson
I have bitten the bullet and ordered some LT1963 regs from the Linear Technology website.
They seem to be the quiet and tolerant of cap types..
Plus I have a chioce of packages, so no teeny weeny sot packs
to drive me mad

I guess I need to find out how to best implement them now!
I was originally thinking of mounting my regs as close as possible to the supply pins and using a single value low esr cap .
This based on reading that the performance of modern caps being
such that paralleling them is unnecessary and represent a higher risk of resonances occurring??
There does seem to be a multitude of views on this matter though


Cheers

Setmenu

 

[jewilson]

It a good practice to decouple the high frequencies with a .1 or a .01 ceramic XR7 type. The OSCON will filter the low frequencies. Put the regulator as close to the chip as possible but be careful not to run the digital traces next to any regulator adjustment pins.

 

[setmenu]

I was thinking Oscon on the receiver etc and a Black gate on the
DAC chip. not really sure of the best values to use here though.
Regarding the Black gates ,it seems there is the opinion that
paralleling these with ceramic does not sound too nice?
[probably the least of my worries eh?]

Thanks for your help


Setmenu

 

[Guido Tent]

Fmak,

I have not seen any performance information on Guido clocks and power supplies. I have heard numerous people speak very positive about them, including Jocko. In addition, I agree with you that the LT regulators or good and better than most. However, to get the kind of noise performance Guido was talking about requires a discreet high performance design and a dam good layout and components.

Now when Guido stated that about power supplies, we all have different ideas of what he’s talking about. Just like, when I was try to explain some basics to someone else.

So, if you trying to accomplish getting a medium scale digital high speed logic power supply to perform to the levels Guido was stating good luck. On the other hand if you have a few pico gates and few SOIC flip flop maybe that is possible depend on the logic types, layout and other good design practices. In addition, just because we have this super regulator does not mean it will perform so in a crummy environment.

Cheers

Hi

Some specs can be found on my website (www.TentLabs.com)

On inverters: At the transistion, the supply rejection is about 6 dB. Half of the supply noise can be found at the output. Ofcourse that noise modulates the decion moment of the next gate, hence jitter is genererated.

For this reason, a clean supply reduces jitter.

The jitter of complex ICs is a complex collection of many contributions: Supply noise is among them, but more issues play a role here. Digital filters used in front of DACs are famous for generating jitter, which is one reason non oversampling sounds much better.

cheers

 

[Peter Daniel]

I finally picked up some 33/16 BG N caps from Partsconnexions and tried them in my DAC. I had 4.7 BG N previously on input driver and receiver chips. Well, I don't think that 33u is a better choice actually. My initial impressions are thickening of sound (tubey like), less resolution. pace and speed. There is less presence of highs (and they become less "shimmering" and silky smooth).

I went back to 4.7 caps. Maybe 10u would be a better choice.

I'm still keeping 100u on a DAC.

 

[setmenu]

Hi Peter
Do you run separate regulators for the analog and digital
V+ pins in the CS8412 chip?
I have seen designs with dual and single supplies.
I would love to run only one as this would help keep the design
compact.
But if this compromises things [further] then two it will have to be!

?


Cheers

Setmenu

 

[Peter Daniel]

I run two separate regulators and since they come in TO-92 package I don't regard it as excessive complication. I don't know what difference a single regulator would do though.

 

[Guido Tent]

Hi Peter
Do you run separate regulators for the analog and digital
V+ pins in the CS8412 chip?
I have seen designs with dual and single supplies.
I would love to run only one as this would help keep the design
compact.
But if this compromises things [further] then two it will have to be!

?


Cheers

Setmenu

Hi

Seperate these supplies, especially the analogue should be very clean, low noise

The digital one should be well decoupled, using series ferrite

cheers

 

[Oli]

Guido will be proud of me!

Guido,

You'll be be very proud of me- I spent some time removing all those nasty ceramic capacitors from my Dac schematic. I also removed my multiple grounds, I placed ferrites in the supply lines and resistors in the digital lines.

I am still looking at my decoupling options- I have noticed Black Gate nx-hiq capacitors. AudioCom seem to use these in their op-amp modules. I have a few questions...

1) I was thinking about using 47uF nx-hiq for my dac and op-amps. Does this seem suitable?

2) Do I need such a high value for the digital logic? Will 0.1nF be enough to provide localised power- remember the ICs supply lines are decoupled with ferrites.

3) Are ferrites okay to use in op-amp supply lines?