It was while I was thinking about hacking the op amp power runs, that I decided I wouldn't be able to do it and this gnd plane improvement. Just too difficult for me. Maybe GLT can try ? So I did this because my experience with other DACs suggests the ground plane is more important for the noise floor.
I'm gonna start with thick film res on the underside and the TL1963 reg, and basically average ho hum components. Find out how bad it can be. Then do upgrades to see how they change things. I'm especially interested in the effect of fitting RN55 resistors over the top of new bit of copper ground vs the thick film on pads on the other side.
But I'm STILL waiting for parts. Wait and see. Here's a pic of a ferrite I fitted to route the xo power in a loop around the trace for the xo signal.
I'm gonna start with thick film res on the underside and the TL1963 reg, and basically average ho hum components. Find out how bad it can be. Then do upgrades to see how they change things. I'm especially interested in the effect of fitting RN55 resistors over the top of new bit of copper ground vs the thick film on pads on the other side.
But I'm STILL waiting for parts. Wait and see. Here's a pic of a ferrite I fitted to route the xo power in a loop around the trace for the xo signal.
An externally hosted image should be here but it was not working when we last tested it.
hmm it would have meant just cutting the trace on either side of each opamp, as it is now i'm afraid you are in a very similar position as before, the power is still routing under the output and effecting the impedance, but it should still be improved somewhat, hard to say how much though. gutsy move though, looks like the same copper strap I use...goertz?
forget thick film resistors, they have terrible VCR (Voltage Coefficient of resistance) and will add noise. oh I see you are making it as bad as you can? are you sure the TL is up for that much current with that..erm.. mounting? I suppose though given your low XO speed it should have less demand, I wouldnt expect it would be happy running with a 100MHz + XO though.
what type ceramics are they? x7r?
forget thick film resistors, they have terrible VCR (Voltage Coefficient of resistance) and will add noise. oh I see you are making it as bad as you can? are you sure the TL is up for that much current with that..erm.. mounting? I suppose though given your low XO speed it should have less demand, I wouldnt expect it would be happy running with a 100MHz + XO though.
what type ceramics are they? x7r?
Yep, that would be doable - six twisted pairs fed back to the regulators(s) - but a bit of a spaghetti nightmare for me. I'm not good at that. Another option is a catenary system with drop-downs at each op amp.
What I have done creates a ground layer over the power lines so the signal path can go over the ground, and hopefully that will be enough to mitigate the layout problem.
I only use COG and X7R but to be honest I don't remember which these were. Probably X7R.
What I have done creates a ground layer over the power lines so the signal path can go over the ground, and hopefully that will be enough to mitigate the layout problem.
I only use COG and X7R but to be honest I don't remember which these were. Probably X7R.
I like your ferrite mod for the clock power. Very nice!
Regarding GND, I need to first figure out where are the return paths. 1/2 of the opmap and components are already on the "right" side...
Regarding GND, I need to first figure out where are the return paths. 1/2 of the opmap and components are already on the "right" side...
...
I only use COG and X7R but to be honest I don't remember which these were. Probably X7R.
COG are way too expensive.
Are you gonna use thin film or through hole ? Through hole metal film might have an advantage. If you look at the smd pads for the 100R and 680R, it seems to me that I could cut the trace immediately after the through hole, disconnecting the pad. If I did this for both pads, then I could remove the surface screen surrounding them and solder the area so the pads are part of the ground plane on that side.
Cut the red lines, and then scrape and solder so the blue area is contiguous ground.
Cut the red lines, and then scrape and solder so the blue area is contiguous ground.
Last edited:
I think that would work, but I don't think it is that important.
I read up about "ground return currents" from no other than Walt Jung...
The return currents will flow under the signal lines (meaning if a bunch of electrons move one way in the signal lines, an equal amount of electrons would move in the opposite direction right under the signal lines. It is important to have an uninterrupted ground plane, otherwise the return current would have to around the break increasing impedance and susceptibility to external disturbances.
In this board, here are 4 signal lines from the DAC to the opamps but they are interrupted by the power lines. The return current will flow right under the resistors if they could. So your mod is a good mod but if you move the GND bridges right under the resistors, then that would be optimal.
In fact, you may not need a band, just some wires under the resistors as this is the shortest path for the return ground current to go through.
On the top, if using a wire for the return current, it may be more advantageous to have the two resistors on top of each other.
Elsewhere, there is uninterrupted ground under the signal lines, so the return currents flow right under the signal lines.
I read up about "ground return currents" from no other than Walt Jung...
The return currents will flow under the signal lines (meaning if a bunch of electrons move one way in the signal lines, an equal amount of electrons would move in the opposite direction right under the signal lines. It is important to have an uninterrupted ground plane, otherwise the return current would have to around the break increasing impedance and susceptibility to external disturbances.
In this board, here are 4 signal lines from the DAC to the opamps but they are interrupted by the power lines. The return current will flow right under the resistors if they could. So your mod is a good mod but if you move the GND bridges right under the resistors, then that would be optimal.
In fact, you may not need a band, just some wires under the resistors as this is the shortest path for the return ground current to go through.
On the top, if using a wire for the return current, it may be more advantageous to have the two resistors on top of each other.
Elsewhere, there is uninterrupted ground under the signal lines, so the return currents flow right under the signal lines.
Last edited:
Hi,
Thanks - that's a nice clear explanation. I understand ground returns - route of least resistance - and ideally, no resistance/impedance/inductance/capacitance right ?
I wanted to put the copper right under the resistor pins but they're very close to the power lines and I'd risk shorting something, so I felt I had to angle them slightly. Because of that, I think that the copper strip is most likely going to help the power lines, not the signal returns.
So, following your explanation, I'm going to do both ground changes because I think the second one is most likely to help the signal returns.
I'd still like to see more vias between the ground planes on this board. I might drill a couple of 1mm holes and do it that way, but perhaps the two mods will be enough.
How's your build coming along ? My parts are due to arrive later today, but I've got to teach this evening, so I'll be stuffing the board tomorrow.
Thanks again for posting the explanation.
Tom
Thanks - that's a nice clear explanation. I understand ground returns - route of least resistance - and ideally, no resistance/impedance/inductance/capacitance right ?
I wanted to put the copper right under the resistor pins but they're very close to the power lines and I'd risk shorting something, so I felt I had to angle them slightly. Because of that, I think that the copper strip is most likely going to help the power lines, not the signal returns.
So, following your explanation, I'm going to do both ground changes because I think the second one is most likely to help the signal returns.
I'd still like to see more vias between the ground planes on this board. I might drill a couple of 1mm holes and do it that way, but perhaps the two mods will be enough.
How's your build coming along ? My parts are due to arrive later today, but I've got to teach this evening, so I'll be stuffing the board tomorrow.
Thanks again for posting the explanation.
Tom
I am learning as I go. Luckily there are books and articles from experts one can read.
The best practice basic approach is to have an uninterrupted ground plane "no matter what". Doesn't mean you cannot route on the ground plane, but Not break it under a signal line.
The ESS board routes through the Gnd plane, so the current would just flow along the signal lines on the GND plane on both sides of the signal wire.
The ESS DAC has a requirement that anywhere under the signal lines, there should be no interruption of the GND plane. This seems to be just "best practice" in the industry.
There are more "tricks" that I've read from people in the know:
Some people likes narrow power lines rather than "power planes" because it has more inductance and with the bypass capacitors they form a filter that further cleans up the power. Others prefer power planes because the added capacitance in the pf range provides additional bypass that is good for GHz operation.
So in this case, narrow power lines seems the better solution...
I am going slow. Not much time left after wife, kids, life and work. What you see in my blog is what I have...
But my controller is all ready to go
The best practice basic approach is to have an uninterrupted ground plane "no matter what". Doesn't mean you cannot route on the ground plane, but Not break it under a signal line.
The ESS board routes through the Gnd plane, so the current would just flow along the signal lines on the GND plane on both sides of the signal wire.
The ESS DAC has a requirement that anywhere under the signal lines, there should be no interruption of the GND plane. This seems to be just "best practice" in the industry.
There are more "tricks" that I've read from people in the know:
Some people likes narrow power lines rather than "power planes" because it has more inductance and with the bypass capacitors they form a filter that further cleans up the power. Others prefer power planes because the added capacitance in the pf range provides additional bypass that is good for GHz operation.
So in this case, narrow power lines seems the better solution...
I am going slow. Not much time left after wife, kids, life and work. What you see in my blog is what I have...
But my controller is all ready to go
I'm learning as I go too. There's a fantastic set of resources here :
http://www.diyaudio.com/forums/digi...ignal-path-line-resistors-5.html#post3017750\
Do post some info here about the controller, or links to the relevant hifiduino stuff. I'd love to learn about that.
Family first eh ? Good for you.
Cheers,
Tom
http://www.diyaudio.com/forums/digi...ignal-path-line-resistors-5.html#post3017750\
Do post some info here about the controller, or links to the relevant hifiduino stuff. I'd love to learn about that.
Family first eh ? Good for you.
Cheers,
Tom
then of course we have the fact that depending on the frequency a resistor looks like an inductor, parasitic effects etc. its not so neat with this board due to the many holes punched in it. even with your ground strap, the lowest impedance ground might be somewhere you dont expect.
drilling a heap of vias may make matters worse, you need to think about where the loop area for each signal is.
actually saying ground flows one way and signal the other is not really correct, that is how it used to be viewed. electrons flow and electron holes flow, it is not as simplified as them flowing in opposite directions. conventional signal flow is not really the most accurate model it turns out, I think your explanation is a bit too neat.
but maybe best to think of it that way for now, just dont expect that just giving a return current a wire to travel on will mean thats the way it goes.
speaking of capacitance though, your ground strap and tape is making a nice little cap =)
I explained all of this in detail in a deleted post on hifiduino a couple of weeks ago.... even offering some mods /solutions. I was pretty negative towards the design, but isnt the fact that people have not even finished one before modding it heavily a sign that I was on the right track? I did try to offset it by offering ways to fix it for those who already had the board.
glt, almost nothing on this board could be called a very high frequency signal, its only barely RF on the clock lines. planes are the way to go vs striplines
the place that will be most critical will be the outputs to the IV resistor and the clock lines, as well as taking care where the i2c lines are.
the dac outputs will still have some high frequency components riding in the signal from the modulator, the ESS is pretty clean compared to some ∑∆ dacs, but still will have some. this is what the LPF is for in the opamp stage, skimming that off and its why a high slewrate opamp is preferred
drilling a heap of vias may make matters worse, you need to think about where the loop area for each signal is.
actually saying ground flows one way and signal the other is not really correct, that is how it used to be viewed. electrons flow and electron holes flow, it is not as simplified as them flowing in opposite directions. conventional signal flow is not really the most accurate model it turns out, I think your explanation is a bit too neat.
but maybe best to think of it that way for now, just dont expect that just giving a return current a wire to travel on will mean thats the way it goes.
speaking of capacitance though, your ground strap and tape is making a nice little cap =)
I explained all of this in detail in a deleted post on hifiduino a couple of weeks ago.... even offering some mods /solutions. I was pretty negative towards the design, but isnt the fact that people have not even finished one before modding it heavily a sign that I was on the right track? I did try to offset it by offering ways to fix it for those who already had the board.
glt, almost nothing on this board could be called a very high frequency signal, its only barely RF on the clock lines. planes are the way to go vs striplines
the place that will be most critical will be the outputs to the IV resistor and the clock lines, as well as taking care where the i2c lines are.
the dac outputs will still have some high frequency components riding in the signal from the modulator, the ESS is pretty clean compared to some ∑∆ dacs, but still will have some. this is what the LPF is for in the opamp stage, skimming that off and its why a high slewrate opamp is preferred
Last edited:
Care to calculate the value of the added capacitance between the ground strap and the power lines ? The "dielectric" is one layer of common insulation tape, plus the pcb coating.
Gents I was reading this article earlier today and thought of you
return path discontinuities
It might help the understanding of return currents and choices made and also gives some suggestions for location of extra caps when the reference plane changes
I like your quip KK
return path discontinuities
It might help the understanding of return currents and choices made and also gives some suggestions for location of extra caps when the reference plane changes
I like your quip KK
Gents I was reading this article earlier today and thought of you
return path discontinuities
It might help the understanding of return currents and choices made and also gives some suggestions for location of extra caps when the reference plane changes
I like your quip KK
Good read, thanks. I hope they don't mind me quoting, "Note, that if the two reference planes are of the same type (either both power, or both ground) you can put a via (ground to ground or power to power)....... immediately adjacent to the signal via."
In other words, for this design with two ground planes, I should add vias near any signal path vias.
In this case, the through hole resistors for I/V are prime candidates for a via close by.
I like your quip KK
yeah, yet i'm sure he'll gladly accept help from us.... its only when its something they dont want to know that we become ********
dont be silly, it was a joke, though i'm sure there is a little bit there. layers of ground plane and power are often used in high frequency digital and RF design as planar capacitance designed right into the PCB, for high frequency transient storage, or decoupling VHF
Last edited:
Sorry, that was my misunderstanding. I'm used to jokes starting something like this - why are lesbians vegetarian ?
I think there's also a bit of a mis-communication generally. I'm not trying to build a perfect DAC. As the Irish joke goes, when you wanna get somewhere, you don't wanna start from here. I'm building this one as a budget DAC, say US$200, and wondering if it'll beat my dual AD1955 384Khz DAC.
I think there's also a bit of a mis-communication generally. I'm not trying to build a perfect DAC. As the Irish joke goes, when you wanna get somewhere, you don't wanna start from here. I'm building this one as a budget DAC, say US$200, and wondering if it'll beat my dual AD1955 384Khz DAC.
In this case, it does, otherwise it must travel around the power lines which is a mile away.
You even removed the screen in almost exact squares. 
- Status
- Not open for further replies.