Hi everybody!
Some of you already knew I was drawing boards for the ESS Sabre. And some of you wanted some. So here is all about this project!!!!
- DIY use?
In fact, I'm a sound engineer that studied electronics for 5 years before switching to sound... I'm not really experienced to say the truth. But I'm doing my best!
I'm creating a full system for how most sound engineers work now (at least in France): oscillating between live sound reinforcement, live recording, studio recording, field recording... So it's a platform of high performance converters / interconnections, that can also transmit over long distances, and where everything can be chained. Its center network part is the Dice from TC.
I took too long to release that DAC board... I'm going to work harder now. Had not much time unfortunately.
What does that mean?
What I'm making right now are prototypes. I'm not going to create a "diy" company. That means I won't have much boards to spare...
The other point being that, before selling these boards to you, I'm going to build them and first check if they work, and if they perform adequately. But to save on PCB cost, some of these boards don't have mounting holes... They weren't designed to be sold. They just are designed to work, and to be integrated in larger boards.
These designs are not free. I'm keeping them to make my full system, even if there's no rocket science in them.
Last but not least, everything is surface mounted. So I'll ship most of the boards ready soldered, unless you can do it yourself. Forget about soldering these boards with a soldering iron: it sometimes is a really tight fit, and I use some 0603 parts (and even 0402 jumpers but these are just to be solder bumps to connect or not connect some lines).
Concerning the price: I still have no real idea. Expensive because I'm making few boards and only use top grade parts... I'll let you know.
That was the disclaimer/annoying stuff. Now, the fun stuff!
How is the project going on?
Boards have been sent to be made to the factory! I chose a very long delivery time to save... But well. Here it is, we're near testing!
What's inside?
Huh. The most important point, of course.
I have no photos of course, but here are the 3d views and descriptions!!!
DAC board
On this board, there is only the DAC, and decoupling using solid polymer capacitors, big Fair Rite beads, NP0/C0G local bypass, and headers. And the clock of course. The clock is the CCHD 950 from crystek, the most expensive part.
It can be used in 2ch or 8ch mode! Your choice.
Left header: Half of the outputs.
Right header: The other 4 outputs.
Top header: Lines for all the signal imputs (so yes, you can do 8ch DSD with this board).
Bottom header: Power supplies, 7 independent lines for everything (AVCCL, AVCCR, AVDDL, AVDDR, Clock, Digital VCC, Digital VDD) with both remote sensing and remote ground sensing pins. Usefull if you want to try some super regs with remote sensing.
Bottom left header: i2c. To control the DAC.
Bottom left header, bis: Pullups. You need to have some on the i2c line at some point, so if you want to have them here, just put usual jumper on the header.
Bottom right: Control. Pins to set the i2c address, to reset, to get the "locked" and "muted" infos.
I/V Board
Well, everybody knows the Twisted Pear's IVY. It's basically the same thing, some slightly different stuff. I have no idea what the difference in practice will be with the IVY. I'll check. Probably should be the same thing.
- I only use big Silver Mica caps and instrumentation precision resistors in the signal path. For who cares. The footprints for the caps are compatible with ( I don't remember the ref) matching NP0 caps to minimize costs, and resistors footprints are standard 1206.
-Filtering is slightly different in that you can have a PI filter on the input, another on the output, and a resistor on input for half voltage/half current mode (see IVY about that).
-Always the same C0G/Oscon/FairRite decoupling stuff on each supply rail.
-You can shunt every part of the schematic you don't use and there are a lot of test points.
-You can plug in my gain switcher (see below)
Positive Voltage Regulator
It's made to use LT1763 or LT1963. The 1763 is 20µV noise, the 1963 is higher current/RF regulation but 40µV noise. These parts are amongst the best integrated regulators I know. And I must confess I love LT.
These can be used in adjustable mode (resistor ratio on the back) or fixed mode.
In fixed mode, you can use remote sensing or local sensing. In adj mode, local sensing only.
Negative Voltage Regulator
Guess what's in...
The LT1964, lowest noise of integrated negative regulators. Adjustable mode or fixed voltage mode. Bypass. I think that's all.
Pad
I don't know why I make this board as I probably won't use it. So I'll probably have a lot to spare.
This is a double T pad (balanced), with true bypass (disconnecting both input and output). To serve wherever you want to put it.
Feedback resistor switcher for I/V
This is a strange part. There's a relay. You can plug it in my I/V board to change the resistor in the signal path. It could switch between two output levels, or switch between 2ch or 8ch operation. The bad is that it makes a longer feedback path, and that never is good news. Test purpose, mostly.
Balanced receiver to single ended
Near everything is in the title. It's based on the best integrated circuit right now: the THAT 1200. Hey, they are cheap on mouser! So why use something not as good?
You can use it after the I/V board to drive unbalanced loads. But as it is made to be driven from not perfectly balanced load etc... it would make a great input stage for where you need it.
All these boards have been sent to factory. They are ALL made from 4 layers pcb! with solid ground plane (of course).
They all have a separate Via/hole to connect a single ground wire for star grounding of boards between them. Care has been taken that there is single non shared path to ground for everything inside each board.
Now, I'm waiting for the boards to come... Solder them... And start to cry for the mistakes I'll have made, of course!
Thank you!
Nicolas
Some of you already knew I was drawing boards for the ESS Sabre. And some of you wanted some. So here is all about this project!!!!
- DIY use?
In fact, I'm a sound engineer that studied electronics for 5 years before switching to sound... I'm not really experienced to say the truth. But I'm doing my best!
I'm creating a full system for how most sound engineers work now (at least in France): oscillating between live sound reinforcement, live recording, studio recording, field recording... So it's a platform of high performance converters / interconnections, that can also transmit over long distances, and where everything can be chained. Its center network part is the Dice from TC.
I took too long to release that DAC board... I'm going to work harder now. Had not much time unfortunately.
What does that mean?
What I'm making right now are prototypes. I'm not going to create a "diy" company. That means I won't have much boards to spare...
The other point being that, before selling these boards to you, I'm going to build them and first check if they work, and if they perform adequately. But to save on PCB cost, some of these boards don't have mounting holes... They weren't designed to be sold. They just are designed to work, and to be integrated in larger boards.
These designs are not free. I'm keeping them to make my full system, even if there's no rocket science in them.
Last but not least, everything is surface mounted. So I'll ship most of the boards ready soldered, unless you can do it yourself. Forget about soldering these boards with a soldering iron: it sometimes is a really tight fit, and I use some 0603 parts (and even 0402 jumpers but these are just to be solder bumps to connect or not connect some lines).
Concerning the price: I still have no real idea. Expensive because I'm making few boards and only use top grade parts... I'll let you know.
That was the disclaimer/annoying stuff. Now, the fun stuff!
How is the project going on?
Boards have been sent to be made to the factory! I chose a very long delivery time to save... But well. Here it is, we're near testing!
What's inside?
Huh. The most important point, of course.
I have no photos of course, but here are the 3d views and descriptions!!!
DAC board
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
On this board, there is only the DAC, and decoupling using solid polymer capacitors, big Fair Rite beads, NP0/C0G local bypass, and headers. And the clock of course. The clock is the CCHD 950 from crystek, the most expensive part.
It can be used in 2ch or 8ch mode! Your choice.
Left header: Half of the outputs.
Right header: The other 4 outputs.
Top header: Lines for all the signal imputs (so yes, you can do 8ch DSD with this board).
Bottom header: Power supplies, 7 independent lines for everything (AVCCL, AVCCR, AVDDL, AVDDR, Clock, Digital VCC, Digital VDD) with both remote sensing and remote ground sensing pins. Usefull if you want to try some super regs with remote sensing.
Bottom left header: i2c. To control the DAC.
Bottom left header, bis: Pullups. You need to have some on the i2c line at some point, so if you want to have them here, just put usual jumper on the header.
Bottom right: Control. Pins to set the i2c address, to reset, to get the "locked" and "muted" infos.
I/V Board
An externally hosted image should be here but it was not working when we last tested it.
Well, everybody knows the Twisted Pear's IVY. It's basically the same thing, some slightly different stuff. I have no idea what the difference in practice will be with the IVY. I'll check. Probably should be the same thing.
- I only use big Silver Mica caps and instrumentation precision resistors in the signal path. For who cares. The footprints for the caps are compatible with ( I don't remember the ref) matching NP0 caps to minimize costs, and resistors footprints are standard 1206.
-Filtering is slightly different in that you can have a PI filter on the input, another on the output, and a resistor on input for half voltage/half current mode (see IVY about that).
-Always the same C0G/Oscon/FairRite decoupling stuff on each supply rail.
-You can shunt every part of the schematic you don't use and there are a lot of test points.
-You can plug in my gain switcher (see below)
Positive Voltage Regulator
An externally hosted image should be here but it was not working when we last tested it.
It's made to use LT1763 or LT1963. The 1763 is 20µV noise, the 1963 is higher current/RF regulation but 40µV noise. These parts are amongst the best integrated regulators I know. And I must confess I love LT.
These can be used in adjustable mode (resistor ratio on the back) or fixed mode.
In fixed mode, you can use remote sensing or local sensing. In adj mode, local sensing only.
Negative Voltage Regulator
An externally hosted image should be here but it was not working when we last tested it.
Guess what's in...
The LT1964, lowest noise of integrated negative regulators. Adjustable mode or fixed voltage mode. Bypass. I think that's all.
Pad
An externally hosted image should be here but it was not working when we last tested it.
I don't know why I make this board as I probably won't use it. So I'll probably have a lot to spare.
This is a double T pad (balanced), with true bypass (disconnecting both input and output). To serve wherever you want to put it.
Feedback resistor switcher for I/V
An externally hosted image should be here but it was not working when we last tested it.
This is a strange part. There's a relay. You can plug it in my I/V board to change the resistor in the signal path. It could switch between two output levels, or switch between 2ch or 8ch operation. The bad is that it makes a longer feedback path, and that never is good news. Test purpose, mostly.
Balanced receiver to single ended
An externally hosted image should be here but it was not working when we last tested it.
Near everything is in the title. It's based on the best integrated circuit right now: the THAT 1200. Hey, they are cheap on mouser! So why use something not as good?
You can use it after the I/V board to drive unbalanced loads. But as it is made to be driven from not perfectly balanced load etc... it would make a great input stage for where you need it.
All these boards have been sent to factory. They are ALL made from 4 layers pcb! with solid ground plane (of course).
They all have a separate Via/hole to connect a single ground wire for star grounding of boards between them. Care has been taken that there is single non shared path to ground for everything inside each board.
Now, I'm waiting for the boards to come... Solder them... And start to cry for the mistakes I'll have made, of course!
Thank you!
Nicolas
Hi
Thank you all.
More informations to come tomorrow. First: I'm going to sleep, it's 1 AM here.
Quickly:
I don't sell full sets. I will sell individual boards.
There's no group buy. There only are spares that I won't use.
And I first have to test them!!!!
I will tell here how many I do have left, considering I have some people that already are on the waiting list - they are helping me since I started this project and I thank them again.
For board outlines, dimensions are on the pictures.
For schematics, I will post them of course.
Ah, and please apologize for schematics faults that I may not have updated and silkscreen flaws. Of course I'll tell you what is wrong lol.
Stuffed or empty boards, your choice.
Thanks,
Nicolas
Thank you all.
More informations to come tomorrow. First: I'm going to sleep, it's 1 AM here.
Quickly:
I don't sell full sets. I will sell individual boards.
There's no group buy. There only are spares that I won't use.
And I first have to test them!!!!
I will tell here how many I do have left, considering I have some people that already are on the waiting list - they are helping me since I started this project and I thank them again.
For board outlines, dimensions are on the pictures.
For schematics, I will post them of course.
Ah, and please apologize for schematics faults that I may not have updated and silkscreen flaws. Of course I'll tell you what is wrong lol.
Stuffed or empty boards, your choice.
Thanks,
Nicolas
Ok so far.
I had sent the files to the factory when I submit this thread. There were some minor flaws in the generated files so I decided to let the factory do the panelizing in the end and handle this stuff. It added further delays.
Boards should be right there in say 2 weeks maximum, given the time the post will take (they will come from eastern europe).
Sorry for the long wait and not having kept you up to date.
I had sent the files to the factory when I submit this thread. There were some minor flaws in the generated files so I decided to let the factory do the panelizing in the end and handle this stuff. It added further delays.
Boards should be right there in say 2 weeks maximum, given the time the post will take (they will come from eastern europe).
Sorry for the long wait and not having kept you up to date.
Hi,
Boards arrived yesterday!
They look sooo goood
the factory seems to have done an excellent job. Invoice was 540€ final just for the boards and solder stencil, but it looks incredibly neat work. Each were individually optically and electrically tested, something I won't have to deal with too. Hehe.
Ok ok will post you a handful of photos taken with my phone tomorrow - got no digital camera.
Components are currently shipping - ES9018s included, yay.
Caps and so on still are waiting on Mouser's stock to ship to France.
So in 1 week will start the great mess: check if everything's ok, and how to improve the performance up to the desired level.
I let you imagine the number of options that can be tried, from voltage regulator, type of caps, I/V filtering, topology, etc etc... If I get mad, don't worry.
I already see some small mistakes I made... but these are very small ones that really don't matter at this prototyping stage. Well I can tell you: clearance between some of the pins of the ESS chip is a bit small and they will be a pain to solder (but I have 0.1mm precise stencil to counter that).
And, stupid me, I put the sot23 diodes in the bad position on my relay commands. Stupid, no? Well, I'll use good old through hole parts soldered on the connector, that's not that bad.
You'll soon know what works, how it works, how much it would cost, and how much you can buy them from me!
Cheers
Nicolas
Boards arrived yesterday!
They look sooo goood
the factory seems to have done an excellent job. Invoice was 540€ final just for the boards and solder stencil, but it looks incredibly neat work. Each were individually optically and electrically tested, something I won't have to deal with too. Hehe.Ok ok will post you a handful of photos taken with my phone tomorrow - got no digital camera.
Components are currently shipping - ES9018s included, yay.
Caps and so on still are waiting on Mouser's stock to ship to France.
So in 1 week will start the great mess: check if everything's ok, and how to improve the performance up to the desired level.
I let you imagine the number of options that can be tried, from voltage regulator, type of caps, I/V filtering, topology, etc etc... If I get mad, don't worry.
I already see some small mistakes I made... but these are very small ones that really don't matter at this prototyping stage. Well I can tell you: clearance between some of the pins of the ESS chip is a bit small and they will be a pain to solder (but I have 0.1mm precise stencil to counter that).
And, stupid me, I put the sot23 diodes in the bad position on my relay commands. Stupid, no? Well, I'll use good old through hole parts soldered on the connector, that's not that bad.
You'll soon know what works, how it works, how much it would cost, and how much you can buy them from me!
Cheers
Nicolas
Are you going to use the (dfm audio made) diceII prom module as your dice center board solution?
I am very interested in that diceII prom module. With a group buy we could easily lower its price.
Also I would be even more interested in your module boards if we were using both the same hardware.
I am very interested in that diceII prom module. With a group buy we could easily lower its price.
Also I would be even more interested in your module boards if we were using both the same hardware.
Hi,
Well I'm not going to use the DFM for several reasons:
- It can only output 12 channels of i2s at full speed
- Dice II is going to be discontinued (I hope they will output an incredible replacement!)
- It's not exactly what I want, nor is it cheap.
Hell, my final boards would cost me around 100-150$ to do exactly what I want!
But as there is a design stage involved, I need to buy that $2000 EVM to save hassle. I'll try to ask for some help from the TC guys, if they could lend me one or something like that.
I don't want to go through an ocean of unknowns as I don't know enough about high frequency electronics. I studied them but I was just too bad at it... To be honest, the only parts that I fear are: the connection to memory (impedance matched?), the impedance matching for the FireWire part, the clock distribution, the bootloader.
Apart from that, it's just laying i2s tracks with LVDS drivers in the end - a board that I could lay in a few days, wouldn't be these "impedance controlled tracks". Stuff like microstrip etc...
I'll try to contact one of my old professors and see if any of them wants to give me hf pcb layout lessons.
Well I'm not going to use the DFM for several reasons:
- It can only output 12 channels of i2s at full speed
- Dice II is going to be discontinued (I hope they will output an incredible replacement!)
- It's not exactly what I want, nor is it cheap.
Hell, my final boards would cost me around 100-150$ to do exactly what I want!
But as there is a design stage involved, I need to buy that $2000 EVM to save hassle. I'll try to ask for some help from the TC guys, if they could lend me one or something like that.
I don't want to go through an ocean of unknowns as I don't know enough about high frequency electronics. I studied them but I was just too bad at it... To be honest, the only parts that I fear are: the connection to memory (impedance matched?), the impedance matching for the FireWire part, the clock distribution, the bootloader.
Apart from that, it's just laying i2s tracks with LVDS drivers in the end - a board that I could lay in a few days, wouldn't be these "impedance controlled tracks". Stuff like microstrip etc...
I'll try to contact one of my old professors and see if any of them wants to give me hf pcb layout lessons.
An externally hosted image should be here but it was not working when we last tested it.
Price for the stuffed board should be 250$
You'll need a few extras to use it:
- A µC. I'm going to work with Seeduino and will release the code. At $20 this µC board, you can't go wrong - and you can easily control everything that would go around your dac!
-Power supplies regulators, at least two (3.3V and 1,2V), more (up to 7) as quality is required. I'll have regulators boards to spare, between 10 to 15$ each board. I'll let you pick your mains to DC adapter - you could use TP ones with excellent results.
- Input circuitry. Make this on veroboard for spdif, you may also use the SPDIF to i2s modules from Twisted Pear too if you don't want to design or solder anything.
- Optionnal output circuitry. With just resistors, the Sabre will work in Voltage Out mode. With I/V, it will work in current output mode with far better performance.
TP's IVY is, again, a reference. I may have a few I/V channels based on the THS4131 to spare too. Too soon to give you any more info.
Thanks
A few considerations, for what they may be useful...
Considering an I/V filter based around a THS4131.
With the gain setting resistor in the FB loop.
With a cap in parallel with this resistor, and no damping resistor with it (seems unnecessary).
Results from sim, not direct calc, results should be about the same.
A 10n cap will provide a filter with Fc(-3db)=96kHz.
@20kHz, amplitude is 0.24dB under nominal
@20kHz, phase shift is -14°
A 6.8n cap will provide a filter with Fc(-3db)=130kHz.
@20kHz, amplitude is 0.12dB under nominal
@20kHz, phase shift is -9°
Without this FB cap, what if we filter via post-filtering like on ESS's 9008 datasheet? (just too lazy to copy the filter schematic... A symetrical resitor-cap-resistor-cap filter).
With 1k resistors and 470p as in the datasheet,
Fc= 106kHz, dA=0,18dB, dp=-14°
With 22 ohms resistors and 18nF,
Fc= 117kHz, dA=0,127dB, dp=-8°
These are values I chose and computed today to choose between filtering schemes. I may associate them too. Things to try! I give them here if it can save hassle to anybody.
Also note that I provided space for a pi input filter. On spice, they create a really MAD peak near 1.5Mhz. Max L that can be fitted in this space is a 120nH. To start getting LP filtering effect you have to use caps in the µF range (without adding resistor to the DAC's output though). And you still have an insane peak just before the slope (don't know the english word for that). Definately useless. I though small 2nF caps could help remove very HF from the chip, thus avoiding any demodulation effect, but it seems to prove worse.
A few considerations, for what they may be useful...
Considering an I/V filter based around a THS4131.
With the gain setting resistor in the FB loop.
With a cap in parallel with this resistor, and no damping resistor with it (seems unnecessary).
Results from sim, not direct calc, results should be about the same.
A 10n cap will provide a filter with Fc(-3db)=96kHz.
@20kHz, amplitude is 0.24dB under nominal
@20kHz, phase shift is -14°
A 6.8n cap will provide a filter with Fc(-3db)=130kHz.
@20kHz, amplitude is 0.12dB under nominal
@20kHz, phase shift is -9°
Without this FB cap, what if we filter via post-filtering like on ESS's 9008 datasheet? (just too lazy to copy the filter schematic... A symetrical resitor-cap-resistor-cap filter).
With 1k resistors and 470p as in the datasheet,
Fc= 106kHz, dA=0,18dB, dp=-14°
With 22 ohms resistors and 18nF,
Fc= 117kHz, dA=0,127dB, dp=-8°
These are values I chose and computed today to choose between filtering schemes. I may associate them too. Things to try! I give them here if it can save hassle to anybody.
Also note that I provided space for a pi input filter. On spice, they create a really MAD peak near 1.5Mhz. Max L that can be fitted in this space is a 120nH. To start getting LP filtering effect you have to use caps in the µF range (without adding resistor to the DAC's output though). And you still have an insane peak just before the slope (don't know the english word for that). Definately useless. I though small 2nF caps could help remove very HF from the chip, thus avoiding any demodulation effect, but it seems to prove worse.
Yeah you're right.
Regarding the schematics, I'll post one later on imageshack. I wanted to post the exact BOM but realised it wouldn't look nice nor be fully usefull. Of course I'll post details, but do you really need the reference of each resistor? I'll post details. If you want the BOM I'll post it.
So what is it exactly?
=> Globally
An ESS 9018 or 9018 Sabre Reference universal boards.
ALL sabre's input are tied to a connector with a series and a resistor to ground so you can input ANYTHING the Sabre can accept and have the correct input impedance and current adequation.
ALL sabre's outputs are tied to a high quality connector (2.54 latching header gold-on-copper low impedance from 3m).
ALL sabre's commands are tied to connectors: i2c, connector to engage or disengage pull-ups on i2c lines, reset/addr inputs, mute/locked outputs.
ALL power supplies (AVCCR, AVCCL, AVDDR, AVDDL, VCC, VDD, Clock) tied to a 30 pins header with remote SENSE and Ground sense lines. Sense lines can be cut (remove 0402 shunt) near the load to avoid picking up interference if you don't use sensing.
Ultra low jitter clock: Crystek's lowest jitter noise part CCHD-950, 80Mhz, 25ppm (lowest drift version available). -162 dBc phase noise.
=> Grounding and bypass questions
- Full uninterrupted ground plane on all the board.
- Full star grounding with no shared via (with a very few exceptions for some non-critical shared vias where I couldn't stuff an additionnal one).
- 100nF NP0/C0G on EACH supply pin of the chip. Caps chosen to be NON MICROPHONIC.
- Solid Polymer Capacitor, 100µF, 15mohms ESR on each supply rail. NON MICROPHONIC too. Chosen to filter out HF.
- Big (very big) bead from Fair Rite on each supply rail. 600 ohms at 100Mhz. Insanely low DC resistance bead : 6,2mOhms.
- Star power distribution on dedicated layer and via quite large planes.
Note that the 100nF bypass caps are under the chip, so current flow in one direction on top layer, then on the opposite on bottom layer ==> ultra small loops that cancel RF emissions.
=> PCB things
- Gilded 4-layers, 35µ copper (double than standard), 0.15mm trace/space, 0.3mm vias. 2 silkscreens. That's why the bare PCB is expensive!
- Well partitionned layout: no digital things near analog lines and vice versa (of course).
- Perpendicular crossing of lines on seperate layers to avoid interference.
- Shields (screens) on each layer, that can be tied to local ground plane or remote earth.
- Separate small ground plane on top layer for clock with a single big via to ground plane layer to avoid clock interference leaking into nearby components. Ultra short and straight path from clock to chip.
- Shield between each power supply pin, strip, plane... on every layer.
- No shield between each side of differential signals (if one signals couples on one line, you want it on the other to cancel, not going to ground!).
Shields between each channel, and virtually about any line going to the chip.
Differential pairs length and impedance matched.
- Half of the connectors pins dedicated to shielding. Locally tied to ground on outputs only, so you can avoid ground loops by directly grounding through heavy gauge wire each board. On outputs because the you want to send the noise where it is less critical - outputs, what you want is clean signal at the next board's input.
- Everything SMD to lower series inductance.
- Quality components, 100ppm resistors even on non-critical lines.
There probably are some other details (I spent a very long time designing this board!) but I just don't remember them right now.
Please note that I keep less than 10% of the price for me - considering I do all the soldering! It all goes to pretty expensive components, boards, and 0.1mm copper stencils. Would I sell 100 boards it would not even pay my lab power supplies!
These are prototypes of a much more complicated setup I plan for future commercial use - if everything works as hoped.
So the classic legal stuff: boards are protected by copyright with a registered copy of the design and restricted to DIY use only.
Cheers!
Nicolas
Regarding the schematics, I'll post one later on imageshack. I wanted to post the exact BOM but realised it wouldn't look nice nor be fully usefull. Of course I'll post details, but do you really need the reference of each resistor? I'll post details. If you want the BOM I'll post it.
So what is it exactly?
=> Globally
An ESS 9018 or 9018 Sabre Reference universal boards.
ALL sabre's input are tied to a connector with a series and a resistor to ground so you can input ANYTHING the Sabre can accept and have the correct input impedance and current adequation.
ALL sabre's outputs are tied to a high quality connector (2.54 latching header gold-on-copper low impedance from 3m).
ALL sabre's commands are tied to connectors: i2c, connector to engage or disengage pull-ups on i2c lines, reset/addr inputs, mute/locked outputs.
ALL power supplies (AVCCR, AVCCL, AVDDR, AVDDL, VCC, VDD, Clock) tied to a 30 pins header with remote SENSE and Ground sense lines. Sense lines can be cut (remove 0402 shunt) near the load to avoid picking up interference if you don't use sensing.
Ultra low jitter clock: Crystek's lowest jitter noise part CCHD-950, 80Mhz, 25ppm (lowest drift version available). -162 dBc phase noise.
=> Grounding and bypass questions
- Full uninterrupted ground plane on all the board.
- Full star grounding with no shared via (with a very few exceptions for some non-critical shared vias where I couldn't stuff an additionnal one).
- 100nF NP0/C0G on EACH supply pin of the chip. Caps chosen to be NON MICROPHONIC.
- Solid Polymer Capacitor, 100µF, 15mohms ESR on each supply rail. NON MICROPHONIC too. Chosen to filter out HF.
- Big (very big) bead from Fair Rite on each supply rail. 600 ohms at 100Mhz. Insanely low DC resistance bead : 6,2mOhms.
- Star power distribution on dedicated layer and via quite large planes.
Note that the 100nF bypass caps are under the chip, so current flow in one direction on top layer, then on the opposite on bottom layer ==> ultra small loops that cancel RF emissions.
=> PCB things
- Gilded 4-layers, 35µ copper (double than standard), 0.15mm trace/space, 0.3mm vias. 2 silkscreens. That's why the bare PCB is expensive!
- Well partitionned layout: no digital things near analog lines and vice versa (of course).
- Perpendicular crossing of lines on seperate layers to avoid interference.
- Shields (screens) on each layer, that can be tied to local ground plane or remote earth.
- Separate small ground plane on top layer for clock with a single big via to ground plane layer to avoid clock interference leaking into nearby components. Ultra short and straight path from clock to chip.
- Shield between each power supply pin, strip, plane... on every layer.
- No shield between each side of differential signals (if one signals couples on one line, you want it on the other to cancel, not going to ground!).
Shields between each channel, and virtually about any line going to the chip.
Differential pairs length and impedance matched.
- Half of the connectors pins dedicated to shielding. Locally tied to ground on outputs only, so you can avoid ground loops by directly grounding through heavy gauge wire each board. On outputs because the you want to send the noise where it is less critical - outputs, what you want is clean signal at the next board's input.
- Everything SMD to lower series inductance.
- Quality components, 100ppm resistors even on non-critical lines.
There probably are some other details (I spent a very long time designing this board!) but I just don't remember them right now.
Please note that I keep less than 10% of the price for me - considering I do all the soldering! It all goes to pretty expensive components, boards, and 0.1mm copper stencils. Would I sell 100 boards it would not even pay my lab power supplies!
These are prototypes of a much more complicated setup I plan for future commercial use - if everything works as hoped.
So the classic legal stuff: boards are protected by copyright with a registered copy of the design and restricted to DIY use only.
Cheers!
Nicolas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.