Questions re: digital grounding
I have been working for some time on my first DAC project (CS8420, SM5847 (if I can find one!), PCM1704), and am currently working on the PCB design. The PCB will be 2 layers, and am wondering what the proper method is for grounding the digital and analog sections; it seems the more I try to learn about this subject, the more different opinions I find, and the more confused I become...
I know that most mixed signal designs split the digital and analog ground planes to avoid noise from entering the analog section, connecting at only one point, often under the dac or adc chip. I have also read that the dac chip should be considered an analog part and should rest entirely on the analog ground plane.
An article I found on the Analog Devices web site:
suggests that the two ground planes should be "kept separate all the way back to a common system 'star' ground, generally located at the power supplies."
This is the same technique used by Erland Unruh in an article a couple years back in Audio Electronics magazine.
However, I have also been re-reading the article posted on Marc Heiligers DAC page:
by Guido Tent, an EMC engineer at Philips who suggests NOT splitting the ground plane, but keeping it closed:
He states that (with proper decoupling of all chips and the use of lossy ferrite beads on all power feeds) splitting the ground plane creates a huge loop which "leads to high emmision, and a higher chance of inter-system pollution."
I had also considered using isolators such as the Burr Brown ISO150's (as used by Parts Connection on the DAC3.1) to isolate separate digital and analog groundplanes. Would this help address the poblem, or would these devices just be another source for noise and jitter?
So what should I do? I am using lossy ferrites and SMD polymer electrolytics for local decoupling for each chip.
Any feedback, or links to articles on this subject would be greatly appreciated.
I'm afraid I don't have any answers for you, but I know exactly how you feel. The more research you do, the more confused you get, as lots of possible solutions and 'best techniques' seem to be conflicting and at complete odds with one another, and lead to impossible PCB layout problems.
In my last DAC, I eventually just bit the bullet and implemented a split groundplane, joined in the middle, with completely seperate power supplies (4 x 12v Sealed Lead Acid Batteries for the analog stage) and lots of ferrites and SMD decoupling caps and components. It sounded fine.
If I ever get around to building another DAC, I'll isolate the thing completly by using a transformer coupled digital input, ISO150's between the digital and DAC/Analog stage, and seperate isolated power supplies again. A bit more physical seperation between the two stages wouldn't go amiss either (Oh no, now you've got me thinking...)
I think the only way to be certain something works or doesn't work is to build a prototype, measure it and then modify and measure again, which is outside the abilities of most DIYers to do it accurately. We're not interested in obtaining international EMI certification, we only want to make sure it doesn't interfere with our music.
Here is a good place to start research. I highly recomend his book also. www.sigcon.com
Follow Analog Devices advice. The subtle problem with a ground palne is that it can act like an antenna. You need to bypass all wires in/out of the device better.
Don't split the ground planes too far apart: it will cause a larger loop for the currents to circulate.
Can't get the NPC chip still? Let me know.
Thanks for the feedback!
The sigcon site is great- most of it is way over my head, but I have already found some useful information.
I suppose I will go for the split ground plane like Adrian, bridging the two sections on the board. I'm thinking that using an entirely separate power supply for the analog section with separate power transformer might be a good idea, so that the only ground connection between analog and digital section would occour at the ground bridge on the board. Or would this be coutrerproductive? (I still want to think of digital signals like analog signals, although I know they behave quite differently) Could this Increase system noise? Should power supply grounds be tied together before they enter the board?
Or maybe the ISO150 isolators are the best solution. Anyone else have any experience with these?
Lastly, what kind of bypassing do you recommend Jocko? (and look for an email regarding that impossible to find 5847 chip)
If you place the receiver circuitry and maybe even the filter on the digital plane, include 100-300 R series resistors in the signal line and place all of the DAC on the analog plane, the separate plane approach might be OK. You will have some loss of noise margin on the digital signals if the two grounds are connected with too much inductance in between. This is probably not going to hurt signal transmission but may screw up the timing (jitter!) which is only critical if the clock is not generated in or right next to the DAC chip.
Do not by any means connect the digital ground and supply pins of the DAC to the digital network! Any noise between the two ground planes will end up inside the DAC substrate.
Have you looked at the price of those isolator thingies??? I suspect not. You are on the right track......keep going. Bypassing: 0.1 uF caps everywhere you can shove them, the big electrolytics don't seem to be that critical. I've built ones with nothing, and others with big caps everywhere. Personal preference works here.
I don't think the ISO150 is worth it. In fact,
it is probably worse than having a little noise
on the ground plane. You will need two
IS0150s for a stereo DAC: one for the clocks
and one for each data line. Unfortunately,
the worst-case interdevice time delay is
quite high, so you might wind up with more
time domain problems with your digital
signal than you would other wise have.
That is to say, your clocks and your data
might not arrive together.
Keep it simple: decouple each IC power pin
with a big cap and a fast cap, liberally
employ ferrite beads. You'll have as clean
power and ground planes as you can hope
for in a 2-layer PCB layout.
Stay away from the opto's. I got a board here with an opto coupled digi-analog split. Nothing but trouble concerning noise/hum. Both grounds are completely split, but what do you do with the cables/connectors going out? The timing problem with opto's can be solved by using a pll and VCXO on the analog side driving SH circuit. Only a single Fs then unless you switch VCXO's.
I would go with the single plane. Once your circuit is done, fill both sides with copper and lay via's where you can connecting both planes together. Make sure to bridge any important gap in the gnd-copper and try to have gnd under each Vcc trace.
Hopefully I can at least temporarily bring this thread back to life. Some very useful discussion here.
With my current design, I've been leaning towards the "moat" paradigm suggested by Jocko and endorsed by Dr. Johnson. I've got a couple of issues, though:
(1) As Havoc noted, in any DAC design we'll have a non-isolated connection for the audio output. Won't this open up the analog "island" to circulating currents due to the additional inter-ground connection(s)?
(2) In the case of multiple DAC ICs, it is no longer possible for the chips to straddle the "drawbridge".
-Next move is to lay the DACs on the analog ground and bring the signals across the drawbridge. Crosstalk aside, won't this create significant issues by increasing the return path of the digital input?
-Alternatively, create an "island" for each DAC, allowing preservation of the original layout and ground return paths. Is this a problem?
A single plane certainly seems like the easy way out of this dilemma, but I have grave doubts of its ability to provide the necessary isolation.
To stir things up even more, why isn't there any discussion of ground topologies other than ground fill? Although I am relatively ignorant, I was under the impression that planar fill was (often) not encouraged for some of these reasons, as well as radiation and RFI pickup.
|All times are GMT. The time now is 04:39 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2016 DragonByte Technologies Ltd.
Copyright ©1999-2016 diyAudio