With lower buswidth, if all 4 channels are captured some samples are lost
With fifo size of 64, there are timing violations in the DMAC
With this configuration, 65536 samples could be captured from all 4 channels with no sample lost
Because of the DMAC destination bus is 256, the number of samples to be captured must be a multiple of 16, otherwise the system will freeze. This will be corrected in software
- removed the delay controller from the top file and added it inside the util_gmii_to_rgmii core
- removed delay related xdc constraints as they are not needed
Since we are just doing a loopback all the logic is contained within the
IO bank. By using a BUFIO instead of a BUFG we avoid having to route the
clock signal from the IO bank to the middle of the FPGA and back to the IO
bank. This reduces the skew between clock and the data signals and makes
sure that the we can use the same design over a range of different
resolutions without having to calibrate the delay.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The ADV7511 samples on the rising edge. Update the data on the falling
edge, this gives us a larger margin and improved signal stability.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Those were added by mistake. It does not seem to be legal Verilog, but for
some reason Vivado accepts it.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
ffff and 0000 are always reserved control characters when using embedded
syncs. So make sure that we never have them in the pixel data, even when
running in full-range mode.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Add a control bit to the register map that allows to bypass the chroma
sub-sampler in the axi_hdmi_tx core. This is primarily interned to be used
to send the test-pattern directly to the HDMI transmitter without modifying
it.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Connect the raw HDMI data as generated by the ADV7604 to the ILA. For
debugging it is quite useful to be able to compare the data before and
after conversion pipeline.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Currently the hdmi_de_int signal is asserted one clock cycle too early in
packed 422 mode. As a result the EAV sequence ends up in the first pixel
and every other pixel is off by one.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Check for both full range and limited range test-pattern sequences and only
if both don't match assert the tpm_oos signal.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Move the test pattern matcher to its own module. This makes it easier to
use it in other configurations as well.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The TPG is no longer part of the RX core and the corresponding bit in the
register map isn't hooked up to anything. So drop it.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
The TPM OOS status flag is in bit 1. Make sure writing to bit 1 rather than
bit 0 clears the TPM OOS.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Add .gitattributes file which sets up the eol encoding handling. This will
make sure that we get a uniform eol encoding across different operating
systems.
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
- connected reset pin, as vivado reports the reset pin erroneously
- configured XADC in simultaneous sampling mode from XAUX0 and XAUX8
- connected XADC interrupt
- because in the project constraints some base pin constraints are overwritten, the project constraints are processed late
- GPI pins were assigned instead of the XADC GIO0 and GIO1, which were assigned to the XADC for external mux mode
- removed commented code
- removed up_drp_control, up_delay_control dependencies where not needed
- added axi_jesd_gt core in the library makefile
- fixed timing tcl dependency for altera projects
Some cases, when software changed the image formats, the packed formats (24bit/pixel) lost alignment.
(the first 32 bit after sof got lost) This commit fix that issue.
Adrian Costina
Istvan Csomortani
Lars-Peter Clausen