This IP serves as storage interfacing element for external memories like
HBM or DDR4 which have AXI3 or AXI4 data interfaces.
The core leverages the axi_dmac as building blocks by merging an array of
simplex DMA channels into duplex AXI channels. The core will split the
incoming data from the source AXIS interface to multiple AXI channels,
and in the read phase will merge the multiple AXI channels into a single
AXIS destination interface.
The number of duplex channels is set by syntheses parameter and must be
set with the ratio of AXIS and AXI3/4 interface.
Underflow or Overflow conditions are reported back to the data offload
through the control/status interface.
In case multiple AXI channels are used the source and destination AXIS
interfaces widths must match.
This allows the external synchronization input to be driven from
asynchronous sources like a 1 PPS signal or just signals from different
clock domains in general.
Signed-off-by: David Winter <david.winter@analog.com>
up_xfer_done should signalize when a previous control set is
transferred to the other clock domain and the current control set is latched.
If a bit from the up_data_cntrl changes, it should stay in that state until
the up_xfer_done asserts.
When using a CLK_RATIO > 1 the first n samples(n=CLK_RATIO) after sync, are
noisy. This is because the phase accumulator data is passed to the phase to
amplitude converter, during the phase synchronization step.
This patch fixes an issue where the pulse width is only updated two
periods after the current one.
Signed-off-by: David Winter <david.winter@analog.com>
Current implementation is correct only for datapath width of 8.
The buswidth of latency measurement inside a beat has a fixed width (3 bits)
for each lane that must be taken in account when computing the total latency.
Signed-off-by: Laszlo Nagy <laszlo.nagy@analog.com>
Data to DMA/system memory must be presented in widths of multiple of 8 bits,
however this padding is not optimal if is done in the transport layer
since this will affect the DAC/ADC FIFO or offload storage.
This utility block adds or removes padding from sample stream in case the
sample with is not multiple of 8 bits, and can be placed between the DMA
and FIFO/Offload blocks.
On large projects with multiple channels the databus on the FIFO/AXI
stream interface can get wider that 1024 bits.
This commit allows a wider range for all the interfaces,
in case for the memory mapped interfaces where the range is 32-1024 the
user selects a bus width out of range that will be handled by the IPI.
The 4 parameters are added to facilitate transmiting project
related information to the software. They act as read-only
memory which is written in Vivado when the project builds.
Set 31 to SDI FIFO's almost full threshold
In case of multiple SDI (MISO) lanes, the samples arrives in a parallel
fashion. For example in case of 4 MISO line, at the first latching clock
edge 4 bits of a sample will be saved, one bit into each shift register.
The data reorder module reconstruct the incoming samples from the AXI
stream of the offload module.
Some IPs like JESD link layer were marked as not supported on Versal devices by
the current flow while other not (e.g. TPL).
The auto family support seems to workaround this issue.
This commit fixes an issue in situations where we provide an oversized
transaction to the data offload in TX mode. Previously, the data offload
would stop accepting new data (wr_ready <= 0) after filling up the
internal storage, and get stuck waiting for the input transaction to
end, thus locking up the device.
This commit addresses that issue by allowing the data offload to consume
the full input transaction, even if the tail of the buffer will be
truncated in the output.
Signed-off-by: David Winter <david.winter@analog.com>
This commit adds a new synthesis option to the design, that controls
whether an internal clock domain crossing will be generated. Disabling
this option allows you to use a synchronization signal that is
synchronized to the write clock domain externally, and possibly shared
between multiple devices.
The default value retains the old behavior.
Signed-off-by: David Winter <david.winter@analog.com>
Add CMOS support for the interface for the following symbol modes on Xilinx devices:
A B C D E F G H
CSSI__1-lane 1 16/8 80(SDR)/160(DDR) 80 - SDR/DDR SDR/DDR->4/2(C=16), 2/1(C=8)
Columns description:
A - SSI Modes
B - Data Lanes Per Channel
C - Serialization factor Per data lane
D - Max data lane rate(MHz)
E - Max Clock rate (MHz)
F - Max Sample Rate for I/Q (MHz)
G - Data Type
H - DDS Rate
CSSI - CMOS Source Synchronous Interface
This commit changes the transfer length register to work in increments of
64 bytes and without offset. The true transfer length can now be
determined by multiplying the value of the transfer_length register with
64.
A value of zero is interpreted as a request for all available storage.
Additionally, this commit fixes an off by one issue that was discovered
during testing of the RX path.
Signed-off-by: David Winter <david.winter@analog.com>
Update vivado version to 2020.2:
- update default vivado version from 2020.1 to 2020.2
- add conditions to apply specific contraints only in Out Of Context mode.
- update DDR controler parameters for vcu118 and kcu105 dev boards
The width of the parameter `device_cfg_octets_per_multiframe` doesn't match the width in the submodules and corresponding slave module jesd204_tx, resulting in a warning generated during validation in Vivado. This patch increases the width of this parameter in axi_jesd204_tx.
Fix offset for pwms with different periods.
The previous version was using an offset scheme based on pwm counter_0.
By using a separate offset counter the user will not be constrained by
pwm_0 period in regards with the offset of other pulses. In this version
offset 0 is used to delay pwm 0 in regards to the offset counter.
The offset counter will start after the load_config signal is asserted
and all active pwm counters finish the previous cycle or by a software
reset.
The software reset should also be used when using external_sync.
* fixes DRC warning that the clocking configuration may result in data errors
* fixes ioserdes reset issue with synchronous de-assert in data clock domain
This commit adds a standalone TDD IP core based on the
existing up_tdd_cntrl module and the up_axi pcore <-> axi bridge.
Signed-off-by: David Winter <david.winter@analog.com>
If all channels are selected for read the values and ready signals from every
transceiver are combined. Each element merges his signals with the previous.
The first element of the chain must assume the previous channel is always ready.
This reverts commit 829e4155ca.
The first element of the read chain must assume there is no valid element
in front of it. For each element the ready signal of the transceiver should be
routed if the channel is selected either by channel number or broadcast.
When the current element is not selected it should forward the ready signal from
the previous element, however this is not the case for the first one.
Having a constant 1'b1 connected to the ready input of the first element
corrupts the first read of the first channel after a channel switch.
This change will break broadcast reads.
Adds a magic identification value of 0x54444443 at word address 0x3.
It is derived from the ASCII String "TDDC" interpreted as a big-endian
32-bit unsigned integer.
Signed-off-by: David Winter <david.winter@analog.com>
For GTH3/4 64b66b mode add a second clock that drives CLKUSR with a clock
that is 2x of the CLKUSR2 (lane rate/66),
CLKUSR = 2 x CLKUSR2
CLKUSR = lane rate / 33
This can be driven from the GT reference clock or second out clock div2.
This commit also:
- fix eyescan scale on GTY
- remove irrelevant parameters
If R1 mode is first syncronized to the dac clock domain will prevent its
usage if the dac clock is missing. In such case the synchronization will not
propagate.
Depending on FPGA technology the physical layer uses different
deserialization factors and corresponding clock division factors to
divide the source synchronous interface clock. This must be
exposed to software so it can act on it while setting the DDS rate.
Xilinx CMOS clock ratio - 4
Xilinx LVDS clock ratio - 4
Intel CMOS clock ratio - 1
Laszlo Nagy
David Winter
Iulia Moldovan
Adrian Costina
AndrDragomir
PopPaul2021
Nick Pillitteri
alin724
alin724
AndreiGrozav
LIacob106
Filip Gherman
sergiu arpadi
Nicola Corna
Dan Hotoleanu
stefan.raus
Istvan Csomortani
Laszlo Nagy
Mihaita Nagy
Robin Getz
alin724
Isaac T
Josh Blum
Alin-Tudor Sferle