Yes! They are totally different! In low-frequency (audio) circuits, generally only the resistance of the ground connection matters, so star or bus grounding works well. In digital circuits, inductance dominates, so it is important to have small loops. Every high frequency trace should be traveling over a ground plane. You don't want large slots or voids in your plane, because your return currents will have to detour around them and the geometry often causes them to act like antennae.5th element said:
Are you putting the deoupling caps directly on the carrier boards? The reason I ask is there do not appear to be caps close to the pads on the board. I have had trouble with the commercial adapters before because the inductance is high and you are far from the ground plane no matter what you do. It is better to make your own adapter with a solid ground plane underneath, which also connects to the main ground plane. I have had good success with this.
Here is Dr. Johnson's website.
http://www.sigcon.com/index.htm
OK the tracks on the left IC that are duplicates are not meant to be there!
The top left chip is a PCM2902 USB interface chip which extracts an SPDIF stream from a computer. So I gave it a separate ground if its going to be wired to the computer ground.
The bottom left is a SPDIF receiver chip the DIR 1703 which then outputs the relavent signal to the DAC chip which you have correctly identified as the middle chip.
The copper round the bottom left is the 1703, digi ground, and the copper to the right is the analogue ground. The tracks you can see going from the 1703 to the 1798 (DAC) are the BKC SKC DATA and
This is almost probably never going to see a 24bit 192 khz signal in its life, max probably only 16bit 48khz.
OK if I join the analogue and digital grounds of the DAC, which ground do I connect them to? digi or ana?
Is it a really really BAD idea to have both the copper on the top and bottom as a shared ground?
Obviously I dont know what a HUGE impact all this is going to have on sound. Can you give me some idea? Will what I have now give me a good SNR and low distortion. Would the nit picking and perfecting give THAT much of a benafit.
The top left chip is a PCM2902 USB interface chip which extracts an SPDIF stream from a computer. So I gave it a separate ground if its going to be wired to the computer ground.
The bottom left is a SPDIF receiver chip the DIR 1703 which then outputs the relavent signal to the DAC chip which you have correctly identified as the middle chip.
The copper round the bottom left is the 1703, digi ground, and the copper to the right is the analogue ground. The tracks you can see going from the 1703 to the 1798 (DAC) are the BKC SKC DATA and
This is almost probably never going to see a 24bit 192 khz signal in its life, max probably only 16bit 48khz.
OK if I join the analogue and digital grounds of the DAC, which ground do I connect them to? digi or ana?
Is it a really really BAD idea to have both the copper on the top and bottom as a shared ground?
Obviously I dont know what a HUGE impact all this is going to have on sound. Can you give me some idea? Will what I have now give me a good SNR and low distortion. Would the nit picking and perfecting give THAT much of a benafit.
tiroth said:
I wont have good success. One reason I am getting the bought ones is I wont be able to make a successful SSOP PCB. Yes it is all decoupled.
High frequency - Is this the clock signal and the three data signals that go to the dac? If so I can rearrange the copper a little bit to make this happen.
OK! I went away and fiddled a little bit more!
This reduced the number of tracks on the copper top to a minimum possible. Also there is now copper under or on top of all the traces appart from a couple which are just power lines.
I have a feeling that all the grouding could be wired up on the top layer leaving the bottom one free. The only problem with this is that some traces which go on the top layer will not have an "earth" below them, but will have copper present.
Tiroth - I am obviously going to use the adapters which I have purchased. If I find I have a problem with them then I will put copper foil on the underside of the adapter and then link it to the ground plane, this should work in a similar way I would think.
This reduced the number of tracks on the copper top to a minimum possible. Also there is now copper under or on top of all the traces appart from a couple which are just power lines.
I have a feeling that all the grouding could be wired up on the top layer leaving the bottom one free. The only problem with this is that some traces which go on the top layer will not have an "earth" below them, but will have copper present.
Tiroth - I am obviously going to use the adapters which I have purchased. If I find I have a problem with them then I will put copper foil on the underside of the adapter and then link it to the ground plane, this should work in a similar way I would think.
Attachments
Congrats, you are making great strides in improving the design. I think you have improved it to the point where it will definitely work from a signal integrity standpoint, even if there are some noise issues. Some problems are unavoidable with leaded packages (the long slots formed by the pins) but I am guessing at least some of those pins are ground, so that should help.
The left, lower IC has a bunch of traces choking off the connection to ground under the package. That might bear some looking into.
Always think loops--there is no such thing as "just" a power trace--if the power is going to something that is doing high speed switching, guess where the switching current is coming from.
As for the ground connection, if the left half in your PCB is the digital side of the DAC and the right the analog, then you could probably just join the ground all along the underside of the DAC.
The left, lower IC has a bunch of traces choking off the connection to ground under the package. That might bear some looking into.
Always think loops--there is no such thing as "just" a power trace--if the power is going to something that is doing high speed switching, guess where the switching current is coming from.
As for the ground connection, if the left half in your PCB is the digital side of the DAC and the right the analog, then you could probably just join the ground all along the underside of the DAC.
Well, I can tell you that I certainly don't have the experience to make a '"perfect" design. The problem is that it is relatively easy to spot major routing goofs, but the answer you are looking for ("perfect" design) could only be pointed out by someone sitting down, laying out and routing the design themself. It is an iterative process generally, you keep trying to improve it until it is optimal.
The best advice I could offer is keep the groundplane as continuous as possible, all on one layer, and use as many SMD parts as possible to avoid interruptions and take advantage of closer spacing and lower lead inductance. Pay very close attention to decoupling and minimize trace lengths.
The best advice I could offer is keep the groundplane as continuous as possible, all on one layer, and use as many SMD parts as possible to avoid interruptions and take advantage of closer spacing and lower lead inductance. Pay very close attention to decoupling and minimize trace lengths.
In "Supply Decoupling" Guido Tent argues for not using power planes. One of the reasons given is that a power plane might resonate with a ground plane.
A power plane in close proximity to a ground plane will form a capacitor. This capacitor behaves as a capacitor to very high frequencies and might therefore make a second small bypass capacitor unnecessary. The capacitor formed by the two planes might of cause resonate with the inductance formed by the seris inductance of the ordinary bypass capacitor and/or with the IC's power pin and internal power supply wires.
However, if ICs are connected to the power plane with damping resistors the resonance can be curbed. A serial resistor does of cause result in an unvelcome voltage drop, so ferrite beads having a high loss at the resonance frequency might be a better choice.
I fully agree with Guido's use of local decoupling capacitors, but I am not entirely convinced by his dismissal of power planes. The risk of a resonance also excists when low ESR capacitors are employed and the cure seems to be the same in booth cases: ferrite beds.
I do not say that Guido is wrong, only that I am not entirely convinced. At least not yet. It would therefore be interesting to hear from anybody that might have done measurements in order to compare the two approaches. Anybody that knows about peer refereed publications dealing with this issue?
A power plane in close proximity to a ground plane will form a capacitor. This capacitor behaves as a capacitor to very high frequencies and might therefore make a second small bypass capacitor unnecessary. The capacitor formed by the two planes might of cause resonate with the inductance formed by the seris inductance of the ordinary bypass capacitor and/or with the IC's power pin and internal power supply wires.
However, if ICs are connected to the power plane with damping resistors the resonance can be curbed. A serial resistor does of cause result in an unvelcome voltage drop, so ferrite beads having a high loss at the resonance frequency might be a better choice.
I fully agree with Guido's use of local decoupling capacitors, but I am not entirely convinced by his dismissal of power planes. The risk of a resonance also excists when low ESR capacitors are employed and the cure seems to be the same in booth cases: ferrite beds.
I do not say that Guido is wrong, only that I am not entirely convinced. At least not yet. It would therefore be interesting to hear from anybody that might have done measurements in order to compare the two approaches. Anybody that knows about peer refereed publications dealing with this issue?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.