I hope you don’t mind but I have re-drawn part of your schematic to give it a more readable format and to follow more industry standards in circuit diagram drafting. (There have been numerous attempts at standards over the years, the most current and most adhered to these days is the IPC-2612, an amalgamation of various ISO BS EJA IEE standards used over the years.)
1: Always use explicit voltages, prefixed with either a + or – to indicate polarity. This has numerous benefits, both during layout, where locating and routing supplies is a critical part of the layout process, if the power delivery is not done right the rest of the circuit will not work. And during fault finding/developing, having an explicit value makes probing easy, you know what voltage you want.
Signal naming becomes even more critical in digital design, to explicitly label a signal and thus its use, clock, differential pair etc. Just to add more information to the pool (or bore you stupid) Appendix A. below has some text extracted from a procedure regarding signal names I have written, mainly for users of high speed simulation software, but also to aid layout and development, and avoids ambiguities and mistakes.
2: Never have 4 connections join at one point, it can be confusing .
I would also change the symbol for the n-channel FET to a more standard one, which is also easier to discern in the diagram, the one you have used looks very similar to the connections.
Appendix A:
SCHEMATIC NET NAMING CONVENTIONS
INTRODUCTION
The following procedure details the net naming conventions that shall be used for all electronic design within Company Name. These naming conventions are to allow us to categorise and easily identify critical nets and reference voltages etc within any design. This is critical to allow the easy addition of constraints and attributes to the nets for control of layout parameters and signal integrity verification. The main object is to name critical nets in a logical fashion that will allow the easy selection of these nets while minimising ambiguities and reducing the prospect of missing a critical net due to naming issues.
There is a hierarchy to the naming convention, where there is a conflict in a signals naming convention, such as clocks that are part of a bus, then the bus name prefix shall take precedent.
Where signals that are generic across a system, the same name shall be used across all flexi and ridged PCB schematics, allowing these signals to be easily traced across multiple designs.
EXCEPTIONS
The following rules shall apply in all instances, unless there is an industry standard reference for a signal name or bus. In the instance that an industry standard reference name is available for a signal or bus that shall be used, if two or more standard signals or busses are present they shall be differentiated by an _TXT suffix. An example of this is the I2C bus, where the standard names SDA and SCL are used for data and clock respectively. Another example is the JTAG clock TCK, where this signal is buffered on a board the format TCK_CPLD, TCK_DSP1 etc shall be used.
BASIC.
SUPPLY VOLTAGES
All supply voltages except 0V (GND) shall follow the format:
+1V0, +3V3, -3V3 primary supply voltages.
+1V0_TX1, +3V3_TX1, -3V3_TX1 secondary supply voltages
+1V0_TX1_TX2, +3V3_TX1_TX2, -3V3_TX1_TX2 tertiary supply voltages
SIV attribute
The attribute ref_voltage should be set to the nominal voltage value as +3.3V, -1.2V, this implicitly tells the SIV and HS tools that this net is to be treated as a reference voltage for all calculations.
BUSSES
All nets within a bus shall have the same prefix, exclusive to that bus. Data and address busses shall have numerical increments after the prefix, control busses or miscellaneous busses:
ADDR_00 ... ADDR_32
McBSP_CLK, McBSP_
DDR_ADDR00 ... DDR_ADDR32
DDR_DATA0 ... DDR_DATA8
CLOCKS
All general clock signals shall be preferably pre-fixed by, or where this conflicts with other rules that take precedent it shall be a suffix:
CLK_
ACTIVE LOW SIGNALS
Generally reset signals, these are signals that are effective when either in a low state or on the falling edge of the signal. Historically when schematics were hand drawn, and in some CAD systems this can be indicated by a line over the signal name, as this is not always possible, example a text lists of signal names you get a conflict of format, as the bar cannot be produced, the following shall be used.
The _n format is used to differentiate the negative indicator from the actual signal name that is always capitalised.
RESET_n
LVDS (DIFFERENTIAL PAIRS)
DIFF_SIG_p
DIFF_SIG_n
Marc
1: Always use explicit voltages, prefixed with either a + or – to indicate polarity. This has numerous benefits, both during layout, where locating and routing supplies is a critical part of the layout process, if the power delivery is not done right the rest of the circuit will not work. And during fault finding/developing, having an explicit value makes probing easy, you know what voltage you want.
Signal naming becomes even more critical in digital design, to explicitly label a signal and thus its use, clock, differential pair etc. Just to add more information to the pool (or bore you stupid) Appendix A. below has some text extracted from a procedure regarding signal names I have written, mainly for users of high speed simulation software, but also to aid layout and development, and avoids ambiguities and mistakes.
2: Never have 4 connections join at one point, it can be confusing .
I would also change the symbol for the n-channel FET to a more standard one, which is also easier to discern in the diagram, the one you have used looks very similar to the connections.
Appendix A:
SCHEMATIC NET NAMING CONVENTIONS
INTRODUCTION
The following procedure details the net naming conventions that shall be used for all electronic design within Company Name. These naming conventions are to allow us to categorise and easily identify critical nets and reference voltages etc within any design. This is critical to allow the easy addition of constraints and attributes to the nets for control of layout parameters and signal integrity verification. The main object is to name critical nets in a logical fashion that will allow the easy selection of these nets while minimising ambiguities and reducing the prospect of missing a critical net due to naming issues.
There is a hierarchy to the naming convention, where there is a conflict in a signals naming convention, such as clocks that are part of a bus, then the bus name prefix shall take precedent.
Where signals that are generic across a system, the same name shall be used across all flexi and ridged PCB schematics, allowing these signals to be easily traced across multiple designs.
EXCEPTIONS
The following rules shall apply in all instances, unless there is an industry standard reference for a signal name or bus. In the instance that an industry standard reference name is available for a signal or bus that shall be used, if two or more standard signals or busses are present they shall be differentiated by an _TXT suffix. An example of this is the I2C bus, where the standard names SDA and SCL are used for data and clock respectively. Another example is the JTAG clock TCK, where this signal is buffered on a board the format TCK_CPLD, TCK_DSP1 etc shall be used.
BASIC.
SUPPLY VOLTAGES
All supply voltages except 0V (GND) shall follow the format:
+1V0, +3V3, -3V3 primary supply voltages.
+1V0_TX1, +3V3_TX1, -3V3_TX1 secondary supply voltages
+1V0_TX1_TX2, +3V3_TX1_TX2, -3V3_TX1_TX2 tertiary supply voltages
SIV attribute
The attribute ref_voltage should be set to the nominal voltage value as +3.3V, -1.2V, this implicitly tells the SIV and HS tools that this net is to be treated as a reference voltage for all calculations.
BUSSES
All nets within a bus shall have the same prefix, exclusive to that bus. Data and address busses shall have numerical increments after the prefix, control busses or miscellaneous busses:
ADDR_00 ... ADDR_32
McBSP_CLK, McBSP_
DDR_ADDR00 ... DDR_ADDR32
DDR_DATA0 ... DDR_DATA8
CLOCKS
All general clock signals shall be preferably pre-fixed by, or where this conflicts with other rules that take precedent it shall be a suffix:
CLK_
ACTIVE LOW SIGNALS
Generally reset signals, these are signals that are effective when either in a low state or on the falling edge of the signal. Historically when schematics were hand drawn, and in some CAD systems this can be indicated by a line over the signal name, as this is not always possible, example a text lists of signal names you get a conflict of format, as the bar cannot be produced, the following shall be used.
The _n format is used to differentiate the negative indicator from the actual signal name that is always capitalised.
RESET_n
LVDS (DIFFERENTIAL PAIRS)
DIFF_SIG_p
DIFF_SIG_n
Marc
No Problems Marce
...Marce, please do point out best practice to me! Here to learn, so being guided as to the correct methods is more than welcome. I'll prehaps defer the additional reading for now, but I will come back later.
...Gentlevoice, more than happy to edit the first post as we go to signpost the major points along the way. Should this thread turn into a great resource for others then it'll be quite handy for others. Shall we agree on milestones as we go? And I'll edit accordingly. I expect the first task might be cataloging the various document links in one spot! Documents and so on perhaps to a Google docs repository of some form? I'll have a think - it makes the version control comments very relevant in an immediate sense!
My thoughts on the various suggestions so far and maybe we can agree on a direction for us all to proceed in.
Design choice - A DAC sounds fine to me and would appear to be of most use in a learning context. Have you a specific schematic you'd like us to approach?
Passing of layouts - PDF's sounds good, shall we lock this away?
Number of layers - I'm thinking a double sided board, so as to make the designs generated achievable for most hobbyist folk - what are your thoughts?
Level of content - there's been a couple of requests to keep it simple in the thread so far, and to balance I imagine that the more skilled hands will enjoy a masterclass in design. What level do you guys want to hit? I'm definitely in the beginner camp (some of the links have been pretty deep reading for me so far - If I've a toe in the water, then they're written for the Olympic swimmer!)
Anything else I may have missed?
...Marce, please do point out best practice to me! Here to learn, so being guided as to the correct methods is more than welcome. I'll prehaps defer the additional reading for now, but I will come back later.
...Gentlevoice, more than happy to edit the first post as we go to signpost the major points along the way. Should this thread turn into a great resource for others then it'll be quite handy for others. Shall we agree on milestones as we go? And I'll edit accordingly. I expect the first task might be cataloging the various document links in one spot! Documents and so on perhaps to a Google docs repository of some form? I'll have a think - it makes the version control comments very relevant in an immediate sense!
My thoughts on the various suggestions so far and maybe we can agree on a direction for us all to proceed in.
Design choice - A DAC sounds fine to me and would appear to be of most use in a learning context. Have you a specific schematic you'd like us to approach?
Passing of layouts - PDF's sounds good, shall we lock this away?
Number of layers - I'm thinking a double sided board, so as to make the designs generated achievable for most hobbyist folk - what are your thoughts?
Level of content - there's been a couple of requests to keep it simple in the thread so far, and to balance I imagine that the more skilled hands will enjoy a masterclass in design. What level do you guys want to hit? I'm definitely in the beginner camp (some of the links have been pretty deep reading for me so far - If I've a toe in the water, then they're written for the Olympic swimmer!)
Anything else I may have missed?
Last edited:
aspringv,
Great start. The editable (by you) first post could become a powerful resource. Note too that we also have a wiki that could be linked to, if someone created pages there.
One variation I can imagine being helpful is possibly getting advice at multiple levels, so as to not overly limit what the experts can impart to us. For example, while I totally agree that a two-sided solution is the best that the vast majority of us can make at home or by hand, I am also afraid that limiting the discussion to that will constrain the experts too much and we would miss too much important information.
We don't HAVE to also capture all of the advanced techniques that would necessarily involve using multilayer boards, but keep in mind that when someone here comes up with a popular design, there is often great interest in generating a group buy, in which case the pooled resources that can be brought to bear are more than sufficient to have the pcbs fabricated professionally, with multiple layers.
I am not sure what the best alternative format would be but I can imagine a sort of multi-tiered discussion, along the lines of "here's the ideal way and here's the best you can do with two-sided" and maybe also "and using two-sided instead of multilayer will be worse by xxx extent). Another alternative might be to cover the different construction types sequentially. But that seems like it might be more difficult within the threaded structure that we have to work in.
Great start. The editable (by you) first post could become a powerful resource. Note too that we also have a wiki that could be linked to, if someone created pages there.
One variation I can imagine being helpful is possibly getting advice at multiple levels, so as to not overly limit what the experts can impart to us. For example, while I totally agree that a two-sided solution is the best that the vast majority of us can make at home or by hand, I am also afraid that limiting the discussion to that will constrain the experts too much and we would miss too much important information.
We don't HAVE to also capture all of the advanced techniques that would necessarily involve using multilayer boards, but keep in mind that when someone here comes up with a popular design, there is often great interest in generating a group buy, in which case the pooled resources that can be brought to bear are more than sufficient to have the pcbs fabricated professionally, with multiple layers.
I am not sure what the best alternative format would be but I can imagine a sort of multi-tiered discussion, along the lines of "here's the ideal way and here's the best you can do with two-sided" and maybe also "and using two-sided instead of multilayer will be worse by xxx extent). Another alternative might be to cover the different construction types sequentially. But that seems like it might be more difficult within the threaded structure that we have to work in.
This topic is of great interest to someone like myself - a tenacious DIY'er who knows full well he/she is in over their head on a lot of the stuff but are unwilling to admit defeat. The vast majority of us are not electrical engineers nor anything even close but we all want to do things as well as possible.
So that being said, IMO, what the 'average' person on this forum would benefit most from is PCB design tips that are within our basement lab capabilities i.e. double sided, photo resist or even toner transfer derived trace's, stitched ground planes (no plated via's), wire pads, combination through hole/SMD components in hand soldered sizes (SOIC, 0805) etc.
Most of us learn very quickly that all PCB's are a compromise to achieve the best results with the minimum in parts and complexity so attention to the most glaring Do's and Don'ts would be terrific. For example, being restricted to double sided boards I have no choice but to stitch a ground plane with wire via's even though I know every stitch creates capacitance (or a loop) and I get an awful feeling every time I place one in a design. What should be avoided at all costs when doing something I have no choice but to do? Another prime example would be the power rails. In my own designs it's not unusual to have 3 different power rails derived from the same source and only two copper planes to work with. What are 'best principles' in the layout of multiple rails with only two copper planes?
Of course no instruction will cover it all but there must be some 'best bang for the buck' things we can learn from the experts. Personally I can only read so much theory and still learn. Ultimately someone needs to demonstrate a few things.
You have my full attention in this thread.
So that being said, IMO, what the 'average' person on this forum would benefit most from is PCB design tips that are within our basement lab capabilities i.e. double sided, photo resist or even toner transfer derived trace's, stitched ground planes (no plated via's), wire pads, combination through hole/SMD components in hand soldered sizes (SOIC, 0805) etc.
Most of us learn very quickly that all PCB's are a compromise to achieve the best results with the minimum in parts and complexity so attention to the most glaring Do's and Don'ts would be terrific. For example, being restricted to double sided boards I have no choice but to stitch a ground plane with wire via's even though I know every stitch creates capacitance (or a loop) and I get an awful feeling every time I place one in a design. What should be avoided at all costs when doing something I have no choice but to do? Another prime example would be the power rails. In my own designs it's not unusual to have 3 different power rails derived from the same source and only two copper planes to work with. What are 'best principles' in the layout of multiple rails with only two copper planes?
Of course no instruction will cover it all but there must be some 'best bang for the buck' things we can learn from the experts. Personally I can only read so much theory and still learn. Ultimately someone needs to demonstrate a few things.
You have my full attention in this thread.
Hi All,
... intriguing to see this thread get going
@marce: Looks interesting what you have posted, however, I have so many things to read and relate to in the days to come so I most like will not have the energy to read through the literature. I will read your actual posts though and assimilate what's in them. About your altered schematic it looks very straightforward and immediately accessible ;-)
@aspringv:
Actually, the ES9102 and PCM4202 are ADCs but I realize that DACs might be more interesting to people here. To that end I would suggest either the ES9018 which from reading around appears to be the de facto "best" (whatever that is...) DAC there is, or the DSD1794 from TI. There might be a practical challenge with the ES9018 though since I have heard that the datasheet is not publicly available - someone else now about this?
The DSD1794 on the other hand to me poses an interesting challenge because it's got many power supply pins distanced from eachother, separate DSD inputs apparently capable of DSD256, I2C programming pins, and superb data. Datasheet can be found here:
http://www.ti.com/lit/ds/symlink/dsd1794a.pdf
Combined with amanero's Combo384 to me it's a very interesting USB DAC combination - and I'd actually like to learn what would be the best way of connecting the digital output signals from the Combo384 to the DSD1794 because they necessarily must cross eachother (and I'd rather keep the connection wires/tracks as short as possible):
http://www.diyaudio.com/forums/vend...-i2s-384khz-dsd-converter-37.html#post3173603
I guess this is not that feasible for this kind of signal ...?
However, I'm of course open to other DAC suggestions (the chance of learning is what to me is interesting here) and also to your circuitry which I perceive to be interesting in terms of analog layout ;-)
Fine with me.
I haven't much experience with this but reckon that two criteria are that it is easily accessible (do Google docs e.g. require an account and learning how to use it?) and ads to what diyaudio already offers in terms of immediate usability. Can I ask what your thoughts are?
@gootee & copper dog:
These would be my suggestions as well. Maybe go for the best possible solution and then those interested in "cutting corners" can ask/suggest how it can be fitted e.g. to two layer boards?
And then, although this may be a bit off topic I personally would also appreciate comments on sustainability issues as things progress. Maybe about different PCB types pollution imprint, manufacturers' sustainability measures etc. This matters to me ...
How to progress?
- I suggest we choose a suitable DAC (with a range of learning challenges), select/make a schematic and then take it from there. How is that for you - all - (not least marce)?
- I'll send aspringv a PM with a short text for the two books (but maybe not today).
Just FYI, for me the next couple of days I most likely will not be around diyaudio but look forward to seeing how things have progressed after that.
'Best for your day
Jesper
... intriguing to see this thread get going
@marce: Looks interesting what you have posted, however, I have so many things to read and relate to in the days to come so I most like will not have the energy to read through the literature. I will read your actual posts though and assimilate what's in them. About your altered schematic it looks very straightforward and immediately accessible ;-)
@aspringv:
... Fine! I'd be pleased to add a few comments to the sources I've (also) mentioned - Henry Ott's & Howard Johnson's books - so that other readers here may know a bit about what these book's topics are.
Actually, the ES9102 and PCM4202 are ADCs but I realize that DACs might be more interesting to people here. To that end I would suggest either the ES9018 which from reading around appears to be the de facto "best" (whatever that is...) DAC there is, or the DSD1794 from TI. There might be a practical challenge with the ES9018 though since I have heard that the datasheet is not publicly available - someone else now about this?
The DSD1794 on the other hand to me poses an interesting challenge because it's got many power supply pins distanced from eachother, separate DSD inputs apparently capable of DSD256, I2C programming pins, and superb data. Datasheet can be found here:
http://www.ti.com/lit/ds/symlink/dsd1794a.pdf
Combined with amanero's Combo384 to me it's a very interesting USB DAC combination - and I'd actually like to learn what would be the best way of connecting the digital output signals from the Combo384 to the DSD1794 because they necessarily must cross eachother (and I'd rather keep the connection wires/tracks as short as possible):
http://www.diyaudio.com/forums/vend...-i2s-384khz-dsd-converter-37.html#post3173603
I guess this is not that feasible for this kind of signal ...?
However, I'm of course open to other DAC suggestions (the chance of learning is what to me is interesting here) and also to your circuitry which I perceive to be interesting in terms of analog layout ;-)
Fine with me.
I haven't much experience with this but reckon that two criteria are that it is easily accessible (do Google docs e.g. require an account and learning how to use it?) and ads to what diyaudio already offers in terms of immediate usability. Can I ask what your thoughts are?
@gootee & copper dog:
These would be my suggestions as well. Maybe go for the best possible solution and then those interested in "cutting corners" can ask/suggest how it can be fitted e.g. to two layer boards?
And then, although this may be a bit off topic I personally would also appreciate comments on sustainability issues as things progress. Maybe about different PCB types pollution imprint, manufacturers' sustainability measures etc. This matters to me ...
How to progress?
- I suggest we choose a suitable DAC (with a range of learning challenges), select/make a schematic and then take it from there. How is that for you - all - (not least marce)?
- I'll send aspringv a PM with a short text for the two books (but maybe not today).
Just FYI, for me the next couple of days I most likely will not be around diyaudio but look forward to seeing how things have progressed after that.
'Best for your day
Jesper
This is not a problem, and does not create the loops you are worried about, this is standard practice and has been for many many years. There are only certain specialised cases where you have to worry about loops on a PCB and those are mainly when doing SMPS,s and other high DI/Dt switching, the rest of the time it is nothing to worry about, because your ground plane has already numerous loops on it.
And for two layer digital designs it is the best way of getting as near to a ground plane as you can, to provide the best possible return current path for such a restricted design format. A ground plane is only such if it is a contiguous unbroken plane, or as near as you can get...two layer designs with signals top and bottom have ground copper pours, where you have minimal signals on the bottom layer you can get a decent enough ground plane but only with minimal signals and minimal breakup of the plane into separate areas.
The majority of boards I do these days are multi-layer, it is rare to see a double sided board apart from really simple stuff, even then it is probably 6 months since I did one; this applies to all genres of design analogue, digital, power supplies and RF.
Sorry but DIYers are going to have to move with the times, and there are numerous designs these days (digital mainly) that require 4 or more layers, there is just no way round it, and to restrict yourself to two layers these days is counter productive.
How do currents flow, really flow, ie they are not trains of electrons shunting round a PCB; understanding current flow is critical to designing PCB's especially digital, and for high speed this knowledge is a must. (and high speed is not determined by the clock frequency, but by the rise time of the signals on the board).
This is also critical for audio and understanding some of the "ground loop" problems people get.
Audio frequencies cover the range where the return path is changing from resistive (the lower frequencies) to inductive (the higher frequencies), thus for some set ups the lower frequencies may actually return through the power cords earth lead instead of the interconnect, whilst the high frequencies will travel down the interconnect return.
This was discussed in detail on an interconnect thread, that IF I can find the link I will post.
This excellent link is worth studying, even if you only read this link, it is fundamental to understanding current flow especially on a PCB.
This link is especially relevant to a DAC design that we are going to play with.
http://www.x2y.com/filters/TechDay0...log_Designs_Demand_GoodPCBLayouts _JohnWu.pdf
The links I have added are there for reference, even numerous full time PCB designers I know don't go into it to the depth I do, but then I get to work on the fun stuff and do consultancy
Ok then - away we go!
Excellent, I guess we're good to go. Just to reiterate (I intend to repost this goal list to the first post, but would like to confirm before I do so)
1) pdf's being the prefered file format for sharing
2) Dac to be designed from schematic (TBC)
3) 4 layer pcb design, with potential to boil it down to 2 layers later in the piece (call this a stretch goal?)
4) build a pcb design resource from this thread, featuring best practices, do's and don'ts. Vehicle for this TBC (wiki? Dropbox? Google Docs? Yahoo Groups?)
I did a very quick bit of googlage and turned up the attached if it's any use?
Excellent, I guess we're good to go. Just to reiterate (I intend to repost this goal list to the first post, but would like to confirm before I do so)
1) pdf's being the prefered file format for sharing
2) Dac to be designed from schematic (TBC)
3) 4 layer pcb design, with potential to boil it down to 2 layers later in the piece (call this a stretch goal?)
4) build a pcb design resource from this thread, featuring best practices, do's and don'ts. Vehicle for this TBC (wiki? Dropbox? Google Docs? Yahoo Groups?)
I did a very quick bit of googlage and turned up the attached if it's any use?
Marce, Thanks for the layout reference. That's the best description/depiction of ground planes and the effect of via's I've seen yet. Already this thread has taught me a basic yet very valuable lesson regarding the use of via's in ground planes and why not to use them in by-passing... which unfortunately I did on my last project.
Could I suggest the occasional reference be provided to documentation such as this that describes the science behind the DAC layout. If an expert likes the way a particular technique is described that's good enough for me.
Another thought, would it be feasible to, in conjunction with the PCB layout tutorial being planned here, to have a periodic parallel explanation of best construction tips for the project? This board does eventually need to be populated, turned on and controlled. How will it be connected to it's various controls? What are best practices for soldering wire directly to a board if that's what will be used? Would construction/soldering (heat settings, flux selection,board washing)/connector selection tips etc. be something worth considering?
Everything we learn from this project will be carried forward to the next and I'm very excited to see this progress. I only wish I had something useful to contribute.
Could I suggest the occasional reference be provided to documentation such as this that describes the science behind the DAC layout. If an expert likes the way a particular technique is described that's good enough for me.
Another thought, would it be feasible to, in conjunction with the PCB layout tutorial being planned here, to have a periodic parallel explanation of best construction tips for the project? This board does eventually need to be populated, turned on and controlled. How will it be connected to it's various controls? What are best practices for soldering wire directly to a board if that's what will be used? Would construction/soldering (heat settings, flux selection,board washing)/connector selection tips etc. be something worth considering?
Everything we learn from this project will be carried forward to the next and I'm very excited to see this progress. I only wish I had something useful to contribute.
Last edited:
Hi all,
@ marce & Copper Dog:
Very informative, indeed
Regarding sources for the first post, I've written up these short suggestions for texts to accompany a couple of already suggested books:
High-Speed Digital Design - a Handbook of Black Magic by Howard W. Johnson
A book with both theory and many practical examples on the effects of inductances, capacitances, vias, sending & terminating resistors, ground bouncing, etc. IMO modestly technical - oriented towards practical use and implications of different design approaches.
Electromagnetic Compatibility Engineering by Henry W Ott. Again a book that combines theory, yet IMO with a very practical orientation. Discusses e.g. the effects of jitter, multilayer boards, detailed on split/non-split ground planes, noise effects of distances between signal and ground layers, what is needed to ensure high resolution digital layouts, etc.
Any comments?
I suggest that marce's latest link is included as well ...
@aspringv:
a) I agree with your latest post but personally I have no time to join in on building something external to diyaudio (point 4). But please feel free to do so ...
b) Your Sabre link is interesting, yet I wonder if it's too "comprehensive"? This also is a question to marce as I assume you will somehow be involved in laying it out .. ?
For a trial layout I was thinking the DAC itself (DSD1794A/combo384 combination .. or Esstech ES9018/combo384 combination) with a few simple power supplies and input/output I/V converters + input/output connections. I assume this would allow for understanding the intricacies involved in laying out the board around the DAC - knowledge I guess could be used in other contexts as well. Comments on this?
The challenge with the EES9018 is that AFAIK their datasheets are NDA protected and thus not publicly available, so I'm underways with contacting them to hear if we can share at least the pin layout publicly.
However, personally I already have a DSD1794A/combo384 system - which plays well - but poses some challenges (low level non-periodic ~2 MHz square waves on the outputs for instance
)) and if you would agree to use this as an example design that would be superb for me ... I could write up a schematic in LTSpice (although it won't simulate due to the DSD1794A being a TI product) as a starting point ...
How is that for you (all)?
Greetings,
Jesper
@ marce & Copper Dog:
Very informative, indeed
Regarding sources for the first post, I've written up these short suggestions for texts to accompany a couple of already suggested books:
High-Speed Digital Design - a Handbook of Black Magic by Howard W. Johnson
A book with both theory and many practical examples on the effects of inductances, capacitances, vias, sending & terminating resistors, ground bouncing, etc. IMO modestly technical - oriented towards practical use and implications of different design approaches.
Electromagnetic Compatibility Engineering by Henry W Ott. Again a book that combines theory, yet IMO with a very practical orientation. Discusses e.g. the effects of jitter, multilayer boards, detailed on split/non-split ground planes, noise effects of distances between signal and ground layers, what is needed to ensure high resolution digital layouts, etc.
Any comments?
I suggest that marce's latest link is included as well ...
@aspringv:
a) I agree with your latest post but personally I have no time to join in on building something external to diyaudio (point 4). But please feel free to do so ...
b) Your Sabre link is interesting, yet I wonder if it's too "comprehensive"? This also is a question to marce as I assume you will somehow be involved in laying it out .. ?
For a trial layout I was thinking the DAC itself (DSD1794A/combo384 combination .. or Esstech ES9018/combo384 combination) with a few simple power supplies and input/output I/V converters + input/output connections. I assume this would allow for understanding the intricacies involved in laying out the board around the DAC - knowledge I guess could be used in other contexts as well. Comments on this?
The challenge with the EES9018 is that AFAIK their datasheets are NDA protected and thus not publicly available, so I'm underways with contacting them to hear if we can share at least the pin layout publicly.
However, personally I already have a DSD1794A/combo384 system - which plays well - but poses some challenges (low level non-periodic ~2 MHz square waves on the outputs for instance
)) and if you would agree to use this as an example design that would be superb for me ... I could write up a schematic in LTSpice (although it won't simulate due to the DSD1794A being a TI product) as a starting point ... How is that for you (all)?
Greetings,
Jesper
Last edited:
I will do a couple of layouts in parallel with others, depending on the complexity, I will certainly do a 4 layer, and if a more challenging design is chosen, I will also do a multilayer "ultimate" layout, where signal integrity etc will be paramount and layer count secondary.
if I can also get the data (IBIS files) I would like to have a play with the signal integrity side of things. The latter depends on time, though, as I am going down to London for a few weeks of on site design, as the data is not allowed off site so will have to rely on Hotel Internet Connections.
RE: the books mentioned above (both of which I have) post #9 has links to the gentlemen's internet sites.
if I can also get the data (IBIS files) I would like to have a play with the signal integrity side of things. The latter depends on time, though, as I am going down to London for a few weeks of on site design, as the data is not allowed off site so will have to rely on Hotel Internet Connections.
RE: the books mentioned above (both of which I have) post #9 has links to the gentlemen's internet sites.
Hi all,
Hmmm... I have a recurring thought saying that I may somehow have altered the original direction of this thread and as this seems plausible reading the thread I have decided to step back a bit ...
I may then open a different thread in some days dealing with ADC or/and DAC layout challenges - feel free to join in should you be so inclined
Best wishes for your weekend,
Jesper
Hmmm... I have a recurring thought saying that I may somehow have altered the original direction of this thread and as this seems plausible reading the thread I have decided to step back a bit ...
I may then open a different thread in some days dealing with ADC or/and DAC layout challenges - feel free to join in should you be so inclined
Best wishes for your weekend,
Jesper
Seems we're running out of steam here
Hi all,
@gentlevoice - yes, we have indeed drifted from the original topic a little but the way forward we've discussed seems like more than enough to answer my original question. Your participation is likely essential for us to move forward! Primarily as all I have left to offer is building a resource/documentation and learning - 4 layers rules me out as I have no software able to design on that level and a DAC design is well past my previous knowledge.
@marce - I would enjoy building a 2 layer design in conjunction with your 4 layer layout if you're keen, but I would require a fair bit of direction - which I'm happy to conceed might be more than you'd be happy to devote time to.
Gents, if we choose to move forward it seems we need a schematic and at least one pcb designer to strut their stuff. I'll facilitate in whatever way I can, but the balls moved a fair way out of my court in terms of moving forward. Let me know what you all think!
Hi all,
@gentlevoice - yes, we have indeed drifted from the original topic a little but the way forward we've discussed seems like more than enough to answer my original question. Your participation is likely essential for us to move forward! Primarily as all I have left to offer is building a resource/documentation and learning - 4 layers rules me out as I have no software able to design on that level and a DAC design is well past my previous knowledge.
@marce - I would enjoy building a 2 layer design in conjunction with your 4 layer layout if you're keen, but I would require a fair bit of direction - which I'm happy to conceed might be more than you'd be happy to devote time to.
Gents, if we choose to move forward it seems we need a schematic and at least one pcb designer to strut their stuff. I'll facilitate in whatever way I can, but the balls moved a fair way out of my court in terms of moving forward. Let me know what you all think!
Bring it on
yes a schematic or two is what's needed now, then we can move on to the practicalities of layout.
If Gentlevoice still wants to be apart of this a more complex schematic and design would be useful, and give us more to look at and discuss.
I will reply more IF I can get on the internet in my Hotel (down London, well Harrow actually, working on site).
yes a schematic or two is what's needed now, then we can move on to the practicalities of layout.
If Gentlevoice still wants to be apart of this a more complex schematic and design would be useful, and give us more to look at and discuss.
I will reply more IF I can get on the internet in my Hotel (down London, well Harrow actually, working on site).
honestly, I think the 9018 is not the best choice here. it requires quite a bit of expense to do properly, even if sticking to objective goals only, you want to try and get as close to the datasheet spec in a design do you not? well thats not an easy or cheap task with the 9018, it just isnt and cutting corners will land you in a mess and there are examples of those messes scattered over this forum ... especially lately.
its also not widely available, you would need to GB the parts direct from the manufacturer at $50 a throw. you cannot stick within guidelines and share the datasheet openly, even though you dont need to sign an NDA anymore, its still not publicly available, you need to ask the distributor for it.
Everyone has an opinion when it comes to this dac and I bet it will land you in a world of pain. it also requires some sort of software/micro-controller to leverage even basic functions and certainly if you want to do a good job. of course there is glt's hifiduino libraries for it, but I think there is more to getting the best out of it, most top shelf designs are programming FPGA for it. it really NEEDS 4 layer unless you are going to make a nest of modules that all connect together and thats not what marce has in mind i'm sure.
dont want to be a wet blanket, I think this is a great initiative, but i'm just not sure how productive this choice is, its pretty limiting and I believe you will end up relying on marce too much. the datasheet is not even really that helpful tbh, its probably not what marce would be looking for, especially if he wants to do any sims, since there is no model out there afaik. it runs at high speed as well. to do the best for this dac you would need to therefore do a few revisions and build/measure/build/measure. unless of course you just want to replicate what is already out there ... that seems a waste of this opportunity if you ask me
the new AKMs look interesting, there is a 4 channel part that seems to have supplanted the now discontinued AK4399 on the top of the roost. but I would also concur with the DSD1794 (1792 needs software control) as the datasheet and app note are excellent and there is a world of info on the ti forums, plus you might be able to get a basic model out of them.
its also not widely available, you would need to GB the parts direct from the manufacturer at $50 a throw. you cannot stick within guidelines and share the datasheet openly, even though you dont need to sign an NDA anymore, its still not publicly available, you need to ask the distributor for it.
Everyone has an opinion when it comes to this dac and I bet it will land you in a world of pain. it also requires some sort of software/micro-controller to leverage even basic functions and certainly if you want to do a good job. of course there is glt's hifiduino libraries for it, but I think there is more to getting the best out of it, most top shelf designs are programming FPGA for it. it really NEEDS 4 layer unless you are going to make a nest of modules that all connect together and thats not what marce has in mind i'm sure.
dont want to be a wet blanket, I think this is a great initiative, but i'm just not sure how productive this choice is, its pretty limiting and I believe you will end up relying on marce too much. the datasheet is not even really that helpful tbh, its probably not what marce would be looking for, especially if he wants to do any sims, since there is no model out there afaik. it runs at high speed as well. to do the best for this dac you would need to therefore do a few revisions and build/measure/build/measure. unless of course you just want to replicate what is already out there ... that seems a waste of this opportunity if you ask me
the new AKMs look interesting, there is a 4 channel part that seems to have supplanted the now discontinued AK4399 on the top of the roost. but I would also concur with the DSD1794 (1792 needs software control) as the datasheet and app note are excellent and there is a world of info on the ti forums, plus you might be able to get a basic model out of them.
Last edited:
Good morning (as it is here in Denmark) to you all,
First @qusp:
Hi qusp - thanks for your input .... I wouldn't be the person to evaluate this but personally the ES9018 would not be my first choice not least because it looks like a rather complex IC to design around and - to me - has a high price. And, as I've written above I already have a DSD1794 which poses some challenges (it plays but have some interesting noise signals here-and-there) that I somehow assume other's may also learn from ...
And then "second": There's an idea surfacing in me that I hope might fulfill many wishes with respect to pcb layout. It awaits a bit of feedback from another person but will post here when I get this.
Best wishes from me,
Jesper
First @qusp:
Hi qusp - thanks for your input .... I wouldn't be the person to evaluate this but personally the ES9018 would not be my first choice not least because it looks like a rather complex IC to design around and - to me - has a high price. And, as I've written above I already have a DSD1794 which poses some challenges (it plays but have some interesting noise signals here-and-there) that I somehow assume other's may also learn from ...
And then "second": There's an idea surfacing in me that I hope might fulfill many wishes with respect to pcb layout. It awaits a bit of feedback from another person but will post here when I get this.
Best wishes from me,
Jesper
The multilayer design I will do for fun, whether it is made or not, the idea (in my view) is to use all the tools I have at my disposal and many years of doing cutting edge designs to provide a reference designs, where points can be discussed as well as being illustrated in an actual design. In fact I would like to do two, one really cutting edge using HDI design principles:
The HDI Handbook
as I am doing more and more of these designs when I can persuade customers to go this way, as there are so many benefits.
So anyone with a good design can have it laid out professionally for nowt And the rest gain a design or two to discuss, talk about and learn from
And I get to give something practical to this site, which I haven't really done since I documented building the Curvy Chang, many moons ago.
The HDI Handbook
as I am doing more and more of these designs when I can persuade customers to go this way, as there are so many benefits.
So anyone with a good design can have it laid out professionally for nowt
And the rest gain a design or two to discuss, talk about and learn fromAnd I get to give something practical to this site, which I haven't really done since I documented building the Curvy Chang, many moons ago.