SD card emulation

[Sri G]

Emulating an SD card in FPGA - with “data” stored over USB on host PC.

I’m testing single board computers and wearing out SD card by constantly flashing them with different builds, not to mention it tiring to keep swapping cards in & out. So I’d like to emulate and SD card and “store” the data on a host pc connected via USB.

Diagram:

microSD interface <— FPGA —> USB <——> Host computer with SSD storage

I will be using the SD 4-bit interface and would like to hit transfer speeds > 50MB/s, the clock should be ~100Mhz.

Is it possible to do this with the Arty-7 kit? Or Zynq-7010? I have those right now, something more purpose specific and cheaper would be helpful to know as well.

 

[Theo Markettos]

rickman <gnuarm@gmail.com> wrote:

On 11/30/2016 4:49 PM, Sri G wrote:
Emulating an SD card in FPGA - with “data” stored over USB on host PC.

I’m testing single board computers and wearing out SD card by constantly
flashing them with different builds, not to mention it tiring to keep
swapping cards in & out. So I’d like to emulate and SD card and “store”
the data on a host pc connected via USB.

Is it possible to do this with the Arty-7 kit? Or Zynq-7010? I have
those right now, something more purpose specific and cheaper would be
helpful to know as well.

I did a little search and found a hackaday page on this by antti.lukats.
I have seen him around here before talking about his interesting
projects. The hackaday page doesn't seem to include files or even any
detail, just a photo. If you can get in touch with him, I expect he
could share his HDL and maybe some of his thoughts.

I suspect it's just SPI mode, but you never know.

Interesting project. Is an SD card dimensions compatible with a
standard 0.062 inch thick PCB? I can find that the SD card is 0.083
inches thick (2.1 mm) I believe the pins are recessed with raised
sections between them making the contact area a different thickness.

It's possible to make something - you may have to pad the topside of a 1.6mm
PCB with some material, or just fab a 2mm PCB. MicroSD is 1mm which is a
common PCB thickness.

> On the other hand, SD is not really an open spec.

It is, as long as you aren't interested in the 'Secure' part of it that
nobody uses.
https://www.sdcard.org/downloads/pls/

Emulating an SD card is doable, but tedious. The interface is designed to
be driven at both ends by software, so there's lots of state machines you
have to implement.

Also if you want 50MB/s then that's 50MHz DDR - that might be in the part
that needs switching signalling down to 1.8v (I can't remember the cutoff).
The higher speeds are more complex to work with.

Can you use USB flash memory instead? That should be a lot easier to
interface to.

Something I've scoped out but not actually built is a double-ended SD card.
One end goes in the system under test, the other end goes in a USB SD
reader. On the card is a microSD socket with a real card in it and some
analogue switches. Flip the switches one way and the card is 'plugged' into
the test system, flip them the other way and it connects to the PC. Combine
this with some GPIOs for control and some power switching and you can do
automated testing - eg flash disc image to SD card with PC, 'plug' it into
test system, boot SBC from SD and run tests, turn off SBC and 'plug' into
PC, read off results. Rinse and repeat for every build of the software.

TI has some nice SD mux chips, for instance:
http://www.ti.com/lit/ds/symlink/ts3a27518e-q1.pdf
I can't remember the one I've used most recently, but it was in that family.

Ah, looks like somebody did it already:
http://hackaday.com/2014/06/08/the-in-circuit-sd-card-switch/

I'm sure there are easier ways to do this than SD emulation - the same idea
with USB for instance. But some SBCs like the Raspberry Pi (until recently
at least) like booting from SD.

Theo

(I have form on evil SD interfaces - my current hackery is a DIMM to SD
card interface...)

 

[rickman]

On 11/30/2016 4:49 PM, Sri G wrote:

Emulating an SD card in FPGA - with “data” stored over USB on host PC.

I’m testing single board computers and wearing out SD card by constantly flashing them with different builds, not to mention it tiring to keep swapping cards in & out. So I’d like to emulate and SD card and “store” the data on a host pc connected via USB.

Diagram:

microSD interface <— FPGA —> USB <——> Host computer with SSD storage

I will be using the SD 4-bit interface and would like to hit transfer speeds > 50MB/s, the clock should be ~100Mhz.

Is it possible to do this with the Arty-7 kit? Or Zynq-7010? I have those right now, something more purpose specific and cheaper would be helpful to know as well.

I did a little search and found a hackaday page on this by antti.lukats.
I have seen him around here before talking about his interesting
projects. The hackaday page doesn't seem to include files or even any
detail, just a photo. If you can get in touch with him, I expect he
could share his HDL and maybe some of his thoughts.

Interesting project. Is an SD card dimensions compatible with a
standard 0.062 inch thick PCB? I can find that the SD card is 0.083
inches thick (2.1 mm) I believe the pins are recessed with raised
sections between them making the contact area a different thickness.

On the other hand, SD is not really an open spec. Can you use USB flash
memory instead? That should be a lot easier to interface to.

--

Rick C

 

[rickman]

On 11/30/2016 6:05 PM, rickman wrote:

On 11/30/2016 4:49 PM, Sri G wrote:
Emulating an SD card in FPGA - with “data” stored over USB on host PC.

I’m testing single board computers and wearing out SD card by
constantly flashing them with different builds, not to mention it
tiring to keep swapping cards in & out. So I’d like to emulate and SD
card and “store” the data on a host pc connected via USB.

Diagram:

microSD interface <— FPGA —> USB <——> Host computer with SSD storage

I will be using the SD 4-bit interface and would like to hit transfer
speeds > 50MB/s, the clock should be ~100Mhz.

Is it possible to do this with the Arty-7 kit? Or Zynq-7010? I have
those right now, something more purpose specific and cheaper would be
helpful to know as well.

I did a little search and found a hackaday page on this by antti.lukats.
I have seen him around here before talking about his interesting
projects. The hackaday page doesn't seem to include files or even any
detail, just a photo. If you can get in touch with him, I expect he
could share his HDL and maybe some of his thoughts.

Interesting project. Is an SD card dimensions compatible with a
standard 0.062 inch thick PCB? I can find that the SD card is 0.083
inches thick (2.1 mm) I believe the pins are recessed with raised
sections between them making the contact area a different thickness.

On the other hand, SD is not really an open spec. Can you use USB flash
memory instead? That should be a lot easier to interface to.

Forgot the link...

https://hackaday.io/project/6946-sd-card-emulator-lattice-xp

--

Rick C

 

I stumbled on that project but its almost 10 years old! Nevertheless, I reached out to Antti Luktas, hopefully he's not too busy to get back to me

Making the physical connection isn't too difficult: https://www.amazon.com/48cm-Sd-micro-Card-Extender-Cable/dp/B007WYRO7O

Even jumper cables are suitable at 350 Mhz..

The specification for the SD protocol is nearly completely available in the Simple SD Specification freely available on sdcard.org - the state machines I'd need are bootup (3.3v), negotiating to high speed (which switches to 1.8v) and data transfer.

The SD-MUX option wears out, I've run through a couple SD cards already which is why I'm looking into the FPGA approach - not to mention faster in speed.

I'm uncertain about FPGA sizing and FPGA's in general which is why I posted here - there doesn't seem to be any relatable data between different FPGA manufacturers and I'm trying to keep costs low if I need to make a lot of boards.

 

[Sri G]

I stumbled on that project but its almost 10 years old! Nevertheless, I reached out to Antti Luktas, hopefully he's not too busy to get back to me

Making the physical connection isn't too difficult: https://www.amazon.com/48cm-Sd-micro-Card-Extender-Cable/dp/B007WYRO7O

Even jumper cables are suitable at 350 Mhz..

The specification for the SD protocol is nearly completely available in the Simple SD Specification freely available on sdcard.org - the state machines I'd need are bootup (3.3v), negotiating to high speed (which switches to 1.8v) and data transfer.

The SD-MUX option wears out, I've run through a couple SD cards already which is why I'm looking into the FPGA approach - not to mention faster in speed.

I'm uncertain about FPGA sizing and FPGA's in general which is why I posted here - there doesn't seem to be any relatable data between different FPGA manufacturers and I'm trying to keep costs low if I need to make a lot of boards.

 

[Theo Markettos]

rickman <gnuarm@gmail.com> wrote:

I saw the cables when looking for a mail SD card connector. I don't
know how well they work. I didn't know SD cards ran at 350 MHz, but at
that speed you need to pay careful attention to the details of
connections. Signal integrity (SI) can be a significant issue.

UHS-I goes up to 104MHz (104MB/s SDR), and then UHS-II adds a differential
interface (up to 312MB/s at present).

Even a small FPGA will do the job of providing the SD card interface.
The USB interface requires a lot of code (relatively speaking) but is
well supported by vendors and various other sources with example code
for MCUs.

A small FPGA will handle the physical interface, but implementing the card
end of the protocol is more complex (though not as bad as USB). I'd suggest
you want to use an FPGA to interface to a CPU (or a soft CPU inside a larger
FPGA) and running software to implement the protocol - doing it all in
hardware will be pretty painful.

Another option might be to take an existing SD card chip and use an FPGA to
emulate the NAND flash it's expecting to talk to. The NAND protocol is a
lot cleaner than SD. However you'd have to force the SD chip to disabling
wear levelling somehow, otherwise it would shuffle your blocks around.

It sounds like a somewhat finnicky project, particularly if you're expecting
to get decent speed out of it...

Theo

 

[rickman]

On 12/2/2016 2:04 AM, Sri G wrote:

I stumbled on that project but its almost 10 years old! Nevertheless,
I reached out to Antti Luktas, hopefully he's not too busy to get
back to me

Making the physical connection isn't too difficult:
https://www.amazon.com/48cm-Sd-micro-Card-Extender-Cable/dp/B007WYRO7O


Even jumper cables are suitable at 350 Mhz..

The specification for the SD protocol is nearly completely available
in the Simple SD Specification freely available on sdcard.org - the
state machines I'd need are bootup (3.3v), negotiating to high speed
(which switches to 1.8v) and data transfer.

The SD-MUX option wears out, I've run through a couple SD cards
already which is why I'm looking into the FPGA approach - not to
mention faster in speed.

I'm uncertain about FPGA sizing and FPGA's in general which is why I
posted here - there doesn't seem to be any relatable data between
different FPGA manufacturers and I'm trying to keep costs low if I
need to make a lot of boards.

People say the specs are available, but I looked for the mechanical
details and never found them.

My understanding is you can use an SD card in SPI mode (one bit data
width) but the 4 bit mode is not specified. Maybe that has changed and
my info is out of date.

I saw the cables when looking for a mail SD card connector. I don't
know how well they work. I didn't know SD cards ran at 350 MHz, but at
that speed you need to pay careful attention to the details of
connections. Signal integrity (SI) can be a significant issue.

Even a small FPGA will do the job of providing the SD card interface.
The USB interface requires a lot of code (relatively speaking) but is
well supported by vendors and various other sources with example code
for MCUs.

Lattice has a number of low end FPGAs that use internal flash to load
the config RAM so an external chip is not needed. That is why Antti
used an XP device. They are now obsolete, but Lattice has the XP2, XO2
and XO3 lines in addition to the iCE40 family. The iCE40 line uses one
time programmable NVM to load the RAM though. If you want the chip to
be reconfigurable in the field you need an external flash chip.

--

Rick C