Nope it hasn't. Feeling at all threatened by my comment qusp ?
Tangential - nothing wrong with using a single plane for both quiet (signal) grounds and noisy (power) grounds, just sounds crap.
apart from the zero distance decouple capacitor trick for the most demand diyer, the power supply to every pin can also be trick there are unlimited possibility, the fun is herethe bottom line is every version of our pcb is assembled and tested to work with the unmodified layout before released to the public. I think it's better to call it an improvement rather than an issue
The PCB is designed for the diyer easy to do trickmany people told i2s line should have series resistor, it do reduce overshoot but it will also reduce the rise time of the signal. There are some exception, the old chip tda1541 do need input series resistor due to the internal architecture. Some new chip has no effect or even causes bad result if added series resistor in the I2S line, diyer are welcome to try and everybody are happy to know the result
Last edited:
Nope it hasn't. Feeling at all threatened by my comment qusp ?
Tangential - nothing wrong with using a single plane for both quiet (signal) grounds and noisy (power) grounds, just sounds crap.
I will try to move the power supply connector to the middle of PCB to all new version pcb, it should provide better separating of left and right channel
Nope it hasn't. Feeling at all threatened by my comment qusp ?
not threatened, I just think star earthing to replace a ground plane is stupid
citation vs a properly designed PCB with single plane please?
Hi,
I'm sure it works great just as it is, but I want to make it better if I can, and this is easier to do before fitting components.
I have removed the power lines completely where they come close to signal lines near the DAC IC, so there is no floating copper traces, and shorted the lines to ground on the input points. I will feed in power.
I have also added extra copper for the use of through hole 620R resistors mounted on the ground plane side, and removed the smd pads and lines for these resistors, to help isolate the op amp power / signals from each other.
I cut and grounded the traces shorting C1-C4 for a continuous ground plane, and shorted the connections with wire on the other side.
I have also cut and grounded the fb1 trace so I can use a through hole ferrite, elevated above the ground plane and perpendicular to the mclk line.
It looks ugly...
I'll also look at using a different ground routing for the filter caps. Probably the easiest way is to mount the 1000pF directly across the 560R smd.
Thanks !
I'm sure it works great just as it is, but I want to make it better if I can, and this is easier to do before fitting components.
I have removed the power lines completely where they come close to signal lines near the DAC IC, so there is no floating copper traces, and shorted the lines to ground on the input points. I will feed in power.
I have also added extra copper for the use of through hole 620R resistors mounted on the ground plane side, and removed the smd pads and lines for these resistors, to help isolate the op amp power / signals from each other.
I cut and grounded the traces shorting C1-C4 for a continuous ground plane, and shorted the connections with wire on the other side.
I have also cut and grounded the fb1 trace so I can use a through hole ferrite, elevated above the ground plane and perpendicular to the mclk line.
It looks ugly...
I'll also look at using a different ground routing for the filter caps. Probably the easiest way is to mount the 1000pF directly across the 560R smd.
Thanks !
Last edited:
not threatened, I just think star earthing to replace a ground plane is stupid
Perhaps you haven't fully digested my suggestion. I'm not suggesting it replaces the groundplane, I'm saying to add star earthing and substitute this for groundplane connections only for the signal grounds. Groundplanes are useful in giving low inductance connections for power, the signal connections OTOH don't need to be low inductance because they're not low impedance. Given that low inductance isn't a necessity for these then we can provide cleaner (lower noise) 0V by star earthing.
You lost me. You want a citation that signal grounds and power grounds are best kept separate in audio?
Henry Ott on mixed signal PCB layout and split ground planes has this to say:
Grounding of Mixes Signal Systems
Grounding of Mixes Signal Systems
Henry Ott is great but he's not someone (as far as I know) who has listened to high quality digital audio circuit layouts. For those who think I'm just blowing smoke here, try the experiment for yourselves and report back what you hear. I've done that so I know what the result is
Not saying that it will be *bad* to connect it how you're saying, it may well be very good, are you able to say that it is categorically the *ultimate* method of connecting the system? Which is why we're having this conversation I suppose.
He has presented at AES conferences a few times, so I take it that he is at least familiar with whatever special nuances audio electronics design needs over other branches of electronics demanding low noise.
Here is an article about one of those occasions - link.
A quote from that articles:
I have only just last week received my copy of Ott's book but when I get some time I will be reading the sections about this topic with interest.
Chris
No, I'm not going to claim that, it would be way too arrogant, even for me
I'm just putting it out as a suggestion for improvement on what's currently being used. It doesn't cost any more to get the layout better so its a win-win all round as far as I can see.Well when I see him talking about not just digital ground vs analog ground but also signal ground vs power ground then I'll know he's up to speed. AES doesn't have a particularly good history of being concerned about SQ in electronics, though they're better in respect of acoustics. They have published a few good papers though, I recall one by Paul Frindle (an ex-colleague of mine) being excellent though this is going back over a decade and probably was only a convention paper, not in their journal.
It is interesting that he disagrees with the semiconductor companies' recommendations. counter_culture, are you reading this?
When doing analogue instrumentation a ground plane is best for numerous reasons. The star refers to a star point where you join power grounds to low level grounds, this has morphed into the horrible spider leg routing that some use today, not the best solution. And for digital layouts a ground plane is a must.
Star gnd for this ? It's not easy to do a star ground for this design - only one ground plane with shared analogue ( 5V and +/-15V ) and digital ( also 5V ). I'm thinking about creating a dual 5V supply with a shared gnd - so two 5V supplies share the same gnd. This is the closest I can get ?
I have put the 1000pF across the 560R - so they have the same gnd point.
Actually, I might change the output stage and go for LME49713HA CFB because they can drive high imp cans directly.
I have put the 1000pF across the 560R - so they have the same gnd point.
Actually, I might change the output stage and go for LME49713HA CFB because they can drive high imp cans directly.
Last edited:
Its not reported but it will be ones that manufacture ADCs and advise separating ground planes I surmise, so potentially TI, ADI and perhaps Cirrus.
I'm comfortable with my unique status
Actually they often recomend one ground plane, though some earlier notes do reccomend seperat ones, as voltages have dropped they often go for one ground. I have numerous papers by Ti National and others I'll dig out some links, there interesting reading, especially in the Ti analogue news letter.
I'll rephrase that out of all the engineers and technical people I have ever met...
I know you have your own whacky views
For the older stuff that worked on decent voltages the high impedance that having a star point creates wasn't a problem, but as I said lower voltages mean that anything that causes one ground to change potential over another can cause problems
KlipschKid, the digital realy does require a ground plane, and for complex digital boards you may have several ground planes to provide return paths for several signal layers, plus power plane pairs (VCC's and GND).
Its an emotive subject is grounding and even in the same R&D lab you'll often get engineers with oposing views! which can make life fun. I treat every layout as a seperate entity, as I do mainly (well almost exclusivly) multi layer and often with SMPS's I use various techniques to keep noise isolated, and minimise current loops and thus limit noise from one section to another. One of the biggest causes of noise is not SMPS's but actual gates switching when you have FPGA's MCU DDR etc, the instantaneous current requirements are often in the 100A range and there is very little you can do to stop that kind of load causing ground bounce.
Have Fun
Marc
I'll rephrase that out of all the engineers and technical people I have ever met...
I know you have your own whacky views
For the older stuff that worked on decent voltages the high impedance that having a star point creates wasn't a problem, but as I said lower voltages mean that anything that causes one ground to change potential over another can cause problems
KlipschKid, the digital realy does require a ground plane, and for complex digital boards you may have several ground planes to provide return paths for several signal layers, plus power plane pairs (VCC's and GND).
Its an emotive subject is grounding and even in the same R&D lab you'll often get engineers with oposing views! which can make life fun. I treat every layout as a seperate entity, as I do mainly (well almost exclusivly) multi layer and often with SMPS's I use various techniques to keep noise isolated, and minimise current loops and thus limit noise from one section to another. One of the biggest causes of noise is not SMPS's but actual gates switching when you have FPGA's MCU DDR etc, the instantaneous current requirements are often in the 100A range and there is very little you can do to stop that kind of load causing ground bounce.
Have Fun
Marc
Henry Ott is great but he's not someone (as far as I know) who has listened to high quality digital audio circuit layouts. For those who think I'm just blowing smoke here, try the experiment for yourselves and report back what you hear. I've done that so I know what the result is
Why is your high frequency return current flowing in the analog portion of your ground plane? It only wants to return via the lowest impedance (inductance) path which is going to be right under the trace. So if you do have a HF problem, your layout is not optimal.
A single ground plane does not work very well if you have a lot of large, low frequency return currents though, since it may be harder to predict the path they will take and their "spread" through the plane will be greater.
I would say your assertion that good design for EMC and instrumentation is different than for audio is a dubious one.
I agree on the low frequency currents, as they get nearer DC they want to follow the path of least resistance, and one reason why system ground loops can be a problem. The problem with audio it covers a range of frequencies that cover the crossover from path of least resistance to least inductance (at 20kHz the current density is not as steep as under a 1MHz+ signal); low frequecies can find the path of least resistance is often the protective earth loops. Providing the lowest impedance and resistive 0V return path helps.
Makes it fun
Makes it fun
- Status
- Not open for further replies.