Back when I was involved with a team building boards using Virtex, Zynq, and similar... we had a system architect, a couple dedicated schematic guy, three people on PCB/layout and an RF guy who had to bless trace length/impedances/stackup and so on. Built about two or three boards a year max. Not to mention the FPGA designers themselves doing the VHDL/Verilog, plus the half dozen offshore guys doing simulation and verification. Just so you get a sense of how cool it is to see one guy hacking this up in basement (or wherever it is).
ive been working through the phils lab videos related to zynq too, what he covers he seems to cover well im playing around with an old spartan 3e atm, hoping to move onto something more modern soon not sure if ill skip straight to something like the zynq or build up one of the ice40's first edit: i also love TI for the documentation and support. theyre really helpful easy to deal with as a company i find.
Get used to QFNs - as soon as you go beyond say 2 Amps, you will have to deal with parts that have exposed pad(s) for sinking heat into the pcb, otherwise heat will have nowhere to go but to overheat the part. That's the reason all medium and high current converters are in QFN-like packages.
The pace of this series of videos is staggering! Btw, I was doing something with an old Spartan3 board and was thinking of making a custom board once everything is working. I'm sure that I checked pricing and availability on Digikey not long ago and was pleased to see it was in stock and inexpensive. But I just checked again and AMD just obsoleted these. No big surprise there but the price for available stock just skyrocketed to $150 or so each!
2.54 mm contacts are typically rated for 2 Amps, and so are jumpers. I often use higher input voltage (typically 12 V when using a power brick, or 15 or 20 V when using Usb-c power) to reduce current and reduce losses.
With USB, you are allowed to draw 0.5 A from usb 2.0 port, or 0.8 A from usb 3.0 port. If you use USB-C port without PD, you can draw up to 1.5 A. Now, I have to warn you that a lot of PCs don't have any current protection whatsoever, so you can easily fry a port if you take too much current. Same can be said about cheap usb-c wall warts. So I highly recommend adding a polyfuse to your board to have overcurrent protection.
I believe that the voltages need to come up in a particular order. You can also take a look at the schematic for the EBAZ4205. It uses TLV62569 for power. It is a good example of using only a four layer board with a ZYNQ-70x0, however it only has 16bit DDR-RAM. You need six layers for 32bit (see Phil's video). The problem with the ZYNQ is that if you want to use USB then you have to supply a USB PHY and therefore it consumes many IO. It would be nice if you were only using 16bit DDR the upper 16bits for the 32bit data could be repurposed as additional IO but the pins can only be used for DDR.
@rehsd it's not the best idea to keep all regulators together as they all heat up which can lead (in extreme cases) to local overheating and even bending of pcb. It's best to spread them around, placing each one as close as possible to circuitry it powers for best transient response.
@@asmi06 I'll experiment with that as I get more of the board components added. It seems like I see many PCBs with the regulators and inductors all along the same edge, but what you're saying makes sense. Thanks!
@rehsd I typically place converters to the side of FPGA which corresponds to IO bank that is to be powered by it, this naturally fits with power planelets on a power layer.
Back when I was involved with a team building boards using Virtex, Zynq, and similar... we had a system architect, a couple dedicated schematic guy, three people on PCB/layout and an RF guy who had to bless trace length/impedances/stackup and so on. Built about two or three boards a year max. Not to mention the FPGA designers themselves doing the VHDL/Verilog, plus the half dozen offshore guys doing simulation and verification. Just so you get a sense of how cool it is to see one guy hacking this up in basement (or wherever it is).
Basement, yes. 🙂 So... you're telling me this could be a multi-year project. 😊
ive been working through the phils lab videos related to zynq too, what he covers he seems to cover well
im playing around with an old spartan 3e atm, hoping to move onto something more modern soon
not sure if ill skip straight to something like the zynq or build up one of the ice40's first
edit: i also love TI for the documentation and support. theyre really helpful easy to deal with as a company i find.
Get used to QFNs - as soon as you go beyond say 2 Amps, you will have to deal with parts that have exposed pad(s) for sinking heat into the pcb, otherwise heat will have nowhere to go but to overheat the part. That's the reason all medium and high current converters are in QFN-like packages.
The pace of this series of videos is staggering! Btw, I was doing something with an old Spartan3 board and was thinking of making a custom board once everything is working. I'm sure that I checked pricing and availability on Digikey not long ago and was pleased to see it was in stock and inexpensive. But I just checked again and AMD just obsoleted these. No big surprise there but the price for available stock just skyrocketed to $150 or so each!
$150... ouch... Time for a Zynq-7000! :)
2.54 mm contacts are typically rated for 2 Amps, and so are jumpers. I often use higher input voltage (typically 12 V when using a power brick, or 15 or 20 V when using Usb-c power) to reduce current and reduce losses.
With USB, you are allowed to draw 0.5 A from usb 2.0 port, or 0.8 A from usb 3.0 port. If you use USB-C port without PD, you can draw up to 1.5 A. Now, I have to warn you that a lot of PCs don't have any current protection whatsoever, so you can easily fry a port if you take too much current. Same can be said about cheap usb-c wall warts. So I highly recommend adding a polyfuse to your board to have overcurrent protection.
A resettable fuse is on my list. Thanks, @asmi06!
You won't need 1.2V - that one is only required for devices with multi gigabit transceivers - like xc7z015.
Love this series already!
I believe that the voltages need to come up in a particular order.
You can also take a look at the schematic for the EBAZ4205. It uses TLV62569 for power. It is a good example of using only a four layer board with a ZYNQ-70x0, however it only has 16bit DDR-RAM. You need six layers for 32bit (see Phil's video).
The problem with the ZYNQ is that if you want to use USB then you have to supply a USB PHY and therefore it consumes many IO.
It would be nice if you were only using 16bit DDR the upper 16bits for the 32bit data could be repurposed as additional IO but the pins can only be used for DDR.
Thank you for all the info, Stephen! The EBAZ4205 schematic will be handy for reference. I will work on power sequencing as I move forward.
Do one part per sheet, aim for Letter sheet size so that schematics can be easily exported into pdf and will remain readable.
Thanks for the suggestion! I'm going to shoot for 11x17 (B) sheets (close to A3), which I find easy to print and still read on a decent monitor.
@@rehsd I've emailed you an example schematics just to give you some ideas of how such things are typically done. Hopefully it will prove useful.
спасибо, хорошая работа
Lol, trust me, you aren't gonna be "FINE" with power taking up a lot of space on a 8 or 10 layer PCB once you find out how much that space costs 😄
😆 Ya', I know it gets expensive quickly. Here's how it's laying out so far: imgur.com/a/FwkhoTS.
@rehsd it's not the best idea to keep all regulators together as they all heat up which can lead (in extreme cases) to local overheating and even bending of pcb. It's best to spread them around, placing each one as close as possible to circuitry it powers for best transient response.
@@asmi06 I'll experiment with that as I get more of the board components added. It seems like I see many PCBs with the regulators and inductors all along the same edge, but what you're saying makes sense. Thanks!
22uF 0603 capacitor is going to be pricey, so I recommend going for 0805 or even larger if higher voltage rating is required.
@rehsd I typically place converters to the side of FPGA which corresponds to IO bank that is to be powered by it, this naturally fits with power planelets on a power layer.
Have you tried to build an Risk-v processor to drive an LCD
I have not.