Hello to All!
My DAC has 3 very small value (.047uf or less) ceramic
caps between the circuit GND and the third wire ground
(connected through the metal chassis). Would I be able
to further reduce ground plane noise if I added another
similar value ceramic cap between GND near major digital
processing chips and the third wire earth ground connection?
Thanks in advance!
Fastcat
My DAC has 3 very small value (.047uf or less) ceramic
caps between the circuit GND and the third wire ground
(connected through the metal chassis). Would I be able
to further reduce ground plane noise if I added another
similar value ceramic cap between GND near major digital
processing chips and the third wire earth ground connection?
Thanks in advance!
Fastcat
Hi fastcat95,
You are on the right track. Cutting the noise and re-routing ground planes / leads is the only way to get good results when modifying Hi-Fi equipment. Only a low noise environment together with low jitter oscillators, simple design, short signal path and high quality components will bring the unit to High End levels.
Usually there are 0.01uF with 0.1uF cap in parallel between ground planes and chassis. However, earthing the chassis to mains ground is not a good idea for many reasons, especially if you have more then one unit’ chassis grounded to mains ground AND they (units) are connected together (which they are via interconnects) AND they are plugged in to separate mains outlets. The earth pins in those separate mains outlets are usually not at the same potential, which causes current to run - and it is very unclean current!!! This applies only if the mains ground is DC coupled to chassis, which happens very often. In your case, you should check to see how your mains earth is connected to chassis. The quietest ground point is power supply ground point where the filter capacitors and voltage regulation is. Sometimes this two ground points are separated to further decrease the noise. Ideally we should have THIS ground potential EVERYWHERE throughout the digital / analog section (each section should have its own power supply ground feed routed to it). There are units with PCB stacked one on top of the other where it is possible to feed THIS power supply ground wherever it is nodded. So, instead of using chassis as a reference, maybe you should try to feed your ground from a truly low noise environment to the place where you need it most, like oscillators, upsemplers, uPC and DAC's. Using mains ground is NOT a good idea.
Some units have chassis fixed to analog ground at RCA's ground point (RCA ground pin) and then 0.1 with 0.01uF caps located throughout, at many places feed the noise to the chassis. So, your chassis sits at very low noise potential thanks to low analog ground noise.
Sorting-out the noise problems usually takes a bit of time, investigation, reading and understanding of current flows and runs.
Some digital sections, like UPSEMLER chips, DAC chips, FILTER chips have localised ground around them, which is connected to the "rest" of the ground via ferrite beads. This is ideal opportunity to “DC feed” some nice clean ground to these localised ground sections and achieve fantastic results. Of course, remove the ferrite beads!!!
You have more then enough hints here to make something good out of your unit.
NOTE: lifting the mains ground off the chassis is DANGEROUS. Unit has to have properly design transformers and mains runs if you are thinking to do this!
Extreme_Boky
You are on the right track. Cutting the noise and re-routing ground planes / leads is the only way to get good results when modifying Hi-Fi equipment. Only a low noise environment together with low jitter oscillators, simple design, short signal path and high quality components will bring the unit to High End levels.
Usually there are 0.01uF with 0.1uF cap in parallel between ground planes and chassis. However, earthing the chassis to mains ground is not a good idea for many reasons, especially if you have more then one unit’ chassis grounded to mains ground AND they (units) are connected together (which they are via interconnects) AND they are plugged in to separate mains outlets. The earth pins in those separate mains outlets are usually not at the same potential, which causes current to run - and it is very unclean current!!! This applies only if the mains ground is DC coupled to chassis, which happens very often. In your case, you should check to see how your mains earth is connected to chassis. The quietest ground point is power supply ground point where the filter capacitors and voltage regulation is. Sometimes this two ground points are separated to further decrease the noise. Ideally we should have THIS ground potential EVERYWHERE throughout the digital / analog section (each section should have its own power supply ground feed routed to it). There are units with PCB stacked one on top of the other where it is possible to feed THIS power supply ground wherever it is nodded. So, instead of using chassis as a reference, maybe you should try to feed your ground from a truly low noise environment to the place where you need it most, like oscillators, upsemplers, uPC and DAC's. Using mains ground is NOT a good idea.
Some units have chassis fixed to analog ground at RCA's ground point (RCA ground pin) and then 0.1 with 0.01uF caps located throughout, at many places feed the noise to the chassis. So, your chassis sits at very low noise potential thanks to low analog ground noise.
Sorting-out the noise problems usually takes a bit of time, investigation, reading and understanding of current flows and runs.
Some digital sections, like UPSEMLER chips, DAC chips, FILTER chips have localised ground around them, which is connected to the "rest" of the ground via ferrite beads. This is ideal opportunity to “DC feed” some nice clean ground to these localised ground sections and achieve fantastic results. Of course, remove the ferrite beads!!!
You have more then enough hints here to make something good out of your unit.
NOTE: lifting the mains ground off the chassis is DANGEROUS. Unit has to have properly design transformers and mains runs if you are thinking to do this!
Extreme_Boky
Extreme_Boky said:
Hi Extreme_Boky.
What you are saying here is that ideal grounding is that every chip should have their own seperate grounds. Theese grounds are wired back to the groundpoint between the filter capasitors, wich then will act as a "star" ground?
Am I right?
What about return paths for the digital signals between chips, like the I2S signals? Should theese travel through this new star ground or should they be connected directly between the grounds of the involving chips?
Thanks!
Tor Martin
Use a solid plane ground. Star grounds don't belong in single-ended digital logic systems.
Between any two chips that are interconnected with high speed single-ended logic (eg, any connected I2S devices), you want the ground between them to have as little inductance as possible. And ideally, I2S and clock signals should have an uninterrupted ground path directly below them, so that the return current travels directly below the traces without spreading out elsewhere and 'contaminating' grounds elsewhere. Best way to accomplish this is a solid ground plane.
In a star ground arrangement, the return current will have to travel through the 'star', which can excite considerable ground voltage differentials between the different devices. In a bad situation, this can make one chip's '0' output look like -0.5V on another chip, or a '1' output look like 5.5V instead of 5V... and I've seen boards fail in this exact scenario.
Between any two chips that are interconnected with high speed single-ended logic (eg, any connected I2S devices), you want the ground between them to have as little inductance as possible. And ideally, I2S and clock signals should have an uninterrupted ground path directly below them, so that the return current travels directly below the traces without spreading out elsewhere and 'contaminating' grounds elsewhere. Best way to accomplish this is a solid ground plane.
In a star ground arrangement, the return current will have to travel through the 'star', which can excite considerable ground voltage differentials between the different devices. In a bad situation, this can make one chip's '0' output look like -0.5V on another chip, or a '1' output look like 5.5V instead of 5V... and I've seen boards fail in this exact scenario.
gmarsh said:
I posted an item from EDN perhaps 1/2 year ago -- seems that it's also best to avoid vias for the high speed circuitry -- a little nub of inductance just where you don't need it.
Personally, the fastest digital circuit I have run is a 20MHz PIC (which isn't that fast) so I can't tell whether the "via" is a real problem.
Well, if we are referring to digital sections - star topology is not important, it is important to feed the noise to the ground plane by shortest way possible in to lowest impedance possible (almost never the case!). So, we have to isolate this noise from other IC's. No need to rout high currents, just noise. I have noticed that some digital sections could be separated and fed very low noise ground from the digital power supply via ribbon cables - I use 10mm wide silver ribbons for this purpose and very short runs. Is this star topology - probably yes, but limited to only 2 or 3 blocks. uPC and LCD sections' ground in some units have already been isolated from the rest of the ground plane by means of ferrite beads. I measured the noise around these chips and was shocked by the amount of noise. The only thing that was stopping this noise polluting the rest of the ground plane were ferrite beads.
I removed the ferrite beads and fed the low noise ground via short run of thick and wide silver ribbons (2-3 in parallel, different width, because the combination sounded better!) to this localised ground. The noise had disappeared, but more interestingly, the noise around filter, DAC and upsempler chip was lowered as well without doing much at all.
Now with low ground noise, high quality decoupling caps made all the difference because they had LOW NOISE ground available to decouple the noise from power supplies / generated by clocks / gates…
So, is it a star topology - probably yes.
Does it disturb DC potential (reference 0) - no way! (actually - it makes it more stable - much more stable)
Does it provide much quieter ground for decoupling to actually WORK for a change?- YES.
Does it isolate noisy stages - YES (!!!)
Now, the point where to feed low noise ground from your digital power supply via ribbon cables to localised ground should be where the clock(s) is(are). uPC - chose ceramic resonators. Upsemlers/ DAC’s/ Filters - chose Xtals.
So, in reality you should use very short and very high quality ribbons and check the original design to see if there is possibility to do this mod. Ground DC reference improves, it does not deteriorate.
There is better solution to this problem – just check PC motherboards and how they distribute current and provide ground potential where needed…
I am usually stuck with single-sided PCB’s – ouch!
The result sound-wise? You have to try this yourself…
Extreme_Boky
I removed the ferrite beads and fed the low noise ground via short run of thick and wide silver ribbons (2-3 in parallel, different width, because the combination sounded better!) to this localised ground. The noise had disappeared, but more interestingly, the noise around filter, DAC and upsempler chip was lowered as well without doing much at all.
Now with low ground noise, high quality decoupling caps made all the difference because they had LOW NOISE ground available to decouple the noise from power supplies / generated by clocks / gates…
So, is it a star topology - probably yes.
Does it disturb DC potential (reference 0) - no way! (actually - it makes it more stable - much more stable)
Does it provide much quieter ground for decoupling to actually WORK for a change?- YES.
Does it isolate noisy stages - YES (!!!)
Now, the point where to feed low noise ground from your digital power supply via ribbon cables to localised ground should be where the clock(s) is(are). uPC - chose ceramic resonators. Upsemlers/ DAC’s/ Filters - chose Xtals.
So, in reality you should use very short and very high quality ribbons and check the original design to see if there is possibility to do this mod. Ground DC reference improves, it does not deteriorate.
There is better solution to this problem – just check PC motherboards and how they distribute current and provide ground potential where needed…
I am usually stuck with single-sided PCB’s – ouch!
The result sound-wise? You have to try this yourself…
Extreme_Boky
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.