Deep Memory Mode (DMM)


Functionality overview

Deep Memory Mode commands enable extended data acquisition and generation by streaming data directly to/from Red Pitaya’s DDR3 RAM at full sampling speed. This overcomes the 16384 sample limitation of standard acquisition, allowing captures of millions of samples limited only by available RAM.

Deep Memory Acquisition (DMA) uses AXI DMA protocol to transfer data between FPGA and DDR memory at high speed, while Deep Memory Generation allows playback of extended arbitrary waveforms.

Important notes

  • Available from Red Pitaya OS version 2.00-23 onwards.

  • Requires sufficient DDR3 RAM allocation (recommend leaving 100 MB for Linux OS).

  • Transfer time to computer increases with buffer size.

  • Use binary format (ACQ:DATA:FORMAT BIN) for faster SCPI data transfer.

  • For detailed information, see Deep Memory Acquisition overview.

Code examples

Here are some examples of how to use Deep Memory Mode:

Deep Memory Acquisition (DMA)

DMA settings

Parameter options:

  • <n> = {1,2} (set channel IN1 or IN2)

  • <byte> = {0...} in bytes

  • <decimation> = {1, 2, 4, 8, 16, 17, 18, 19, ..., 65534, 65535, 65536} Default: 1

  • <decimated_data_num> = {value in samples} Default: 0

  • <pos> = {position inside circular buffer in samples}

  • <enable> = {ON, OFF} Default: OFF

  • <address> = {byte} Address of reserved memory

  • <size> = {byte} Size of buffer in bytes. Default: 2 MB

  • <samples> = {sample} Size of the acquisition buffer in samples. Default: 2 MB

  • <units> = {RAW, VOLTS} Default: VOLTS

STEMlab 125-14 4-Input only (additional):

  • <n> = {3,4} (set channel IN3, or IN4)

Available Jupyter and API macros:

  • Fast analog channels - RP_CH_1, RP_CH_2

STEMlab 125-14 4-Input only (additional):

  • Fast analog channels - RP_CH_3, RP_CH_4

SCPI

API, Jupyter

DESCRIPTION

ECOSYSTEM
ACQ:AXI:START? > <byte>
Example:
ACQ:AXI:START? > 16777216
C++: rp_AcqAxiGetMemoryRegion(uint32_t *_start, uint32_t *_size)

Python: rp_AcqAxiGetMemoryRegion()
Returns the start address of the Deep Memory region.
API: Also returns the size of the memory region.

2.00-18 and up
ACQ:AXI:SIZE? > <byte>
Example:
ACQ:AXI:SIZE? > 2097152
C++: rp_AcqAxiGetMemoryRegion(uint32_t *_start, uint32_t *_size)

Python: rp_AcqAxiGetMemoryRegion()
Get size of reserved memory for Deep Memory mode.
API: Also returns the start address of the memory region.

2.00-18 and up
ACQ:AXI:SOUR<n>:ENable <enable>
Example:
ACQ:AXI:SOUR1:ENable ON
C++: rp_AcqAxiEnable(rp_channel_t channel, bool enable)

Python: rp_AcqAxiEnable(<channel>, <enable>)

Enables the specified acquisition channel in the Deep Memory Acquisition.

2.00-18 and up
ACQ:AXI:DEC <decimation>
Example:
ACQ:AXI:DEC 4
C++: rp_AcqAxiSetDecimationFactor(uint32_t decimation)

Python: rp_AcqAxiSetDecimationFactor(<decimation>)

Sets the decimation used for acquiring signals in the Deep Memory Mode.

2.00-18 and up
ACQ:AXI:DEC? > <decimation>
Example:
ACQ:AXI:DEC? > 1
C++: rp_AcqAxiGetDecimationFactor(uint32_t *decimation)

Python: rp_AcqAxiGetDecimationFactor()

Returns the decimation used for acquiring signals in the Deep Memory Mode.

2.00-18 and up
ACQ:AXI:DEC:CH<n> <decimation>
Example:
ACQ:AXI:DEC:CH1 4
C++: rp_AcqAxiSetDecimationFactorCh(rp_channel_t channel, uint32_t decimation)

Python: rp_AcqAxiSetDecimationFactorCh(<channel>, <decimation>)
Sets the decimation used for acquiring signals in the Deep Memory Mode.
Used only in split trigger mode

2.05-37 and up
ACQ:AXI:DEC:CH<n>? > <decimation>
Example:
ACQ:AXI:DEC:CH1? > 1
C++: rp_AcqAxiGetDecimationFactorCh(rp_channel_t channel, uint32_t *decimation)

Python: rp_AcqAxiGetDecimationFactorCh(<channel>)
Returns the decimation used for acquiring signals in the Deep Memory Mode.
Used only in split trigger mode

2.05-37 and up
ACQ:AXI:SOUR<n>:Trig:Dly <decimated_data_num>
Example:
ACQ:AXI:SOUR1:Trig:Dly 2314
C++: rp_AcqAxiSetTriggerDelay(rp_channel_t channel, int32_t decimated_data_num)

Python: rp_AcqAxiSetTriggerDelay(<channel>, <decimated_data_num>)
Sets the number of decimated data after the trigger is
written into memory.

2.00-18 and up
ACQ:AXI:SOUR<n>:Trig:Dly? > <decimated_data_num>
Example:
ACQ:AXI:SOUR1:Trig:Dly? > 2314
C++: rp_AcqAxiGetTriggerDelay(rp_channel_t channel, int32_t *decimated_data_num)

Python: rp_AcqAxiGetTriggerDelay(<channel>)
Returns the number of decimated data after the trigger is
written into memory.

2.00-18 and up
ACQ:AXI:SOUR<n>:SET:Buffer <address>,<size>
Example:
ACQ:AXI:SOUR<n>:SET:Buffer 16777216,512
C++: rp_AcqAxiSetBufferSamples(rp_channel_t channel, uint32_t address, uint32_t samples)
rp_AcqAxiSetBufferBytes(rp_channel_t channel, uint32_t address, uint32_t size)
Python: rp_AcqAxiSetBufferSamples(<channel>, <address>, <samples>)
rp_AcqAxiSetBufferBytes(<channel>, <address>, <size>)
Sets the Deep Memory buffer address and size in samples.
Buffer size must be a multiple of 2.

2.00-18 and up
ACQ:AXI:DATA:UNITS <units>
Example:
ACQ:AXI:DATA:UNITS RAW
C++: - (see rp_AcqAxiGetDataV and rp_AcqAxiGetDataRaw)

Python: - (see rp_AcqAxiGetDataV and rp_AcqAxiGetDataRaw)
Select units in which the acquired data will be returned.
For API commands the units are selected with the get data function.

2.00-18 and up
ACQ:AXI:DATA:UNITS? > <units>
Example:
ACQ:AXI:DATA:UNITS? > RAW
C++: - (see rp_AcqAxiGetDataV and rp_AcqAxiGetDataRaw)

Python: - (see rp_AcqAxiGetDataV and rp_AcqAxiGetDataRaw)
Get units in which the acquired data will be returned.
For API commands the units are selected with the get data function.

2.00-18 and up
- (NA)
C++: - (look for malloc function online)

Python: rp_createBuffer(<maxChannels>, <length>, <initInt16>, <initDouble>, <initFloat>)
Performs memory allocation and returns the requested buffer.
- <maxChannels> - how many channels will be acquired
- <enght> - length of the buffer in samples (max 16384)
- <initInt16>, <initDouble>, <initFloat> - buffer sample type, set one
to true, others are false.
For Python API specifically. Replaced by functions returning NumPy buffers.

2.00-18 - 2.04-35
- (NA)
C++: - (look for free function online)

Python: rp_deleteBuffer(<buffer>)
Free the allocated resources.
- <buffer> - buffer to be released/freed
For Python API specifically.
Replaced by functions returning NumPy buffers.

2.00-18 - 2.04-35

DMA data read

Parameter options:

  • <n> = {1,2} (set channel IN1 or IN2)

  • <count> = {value in samples} Default: 0

  • <pos> = {samples} Position inside circular buffer in samples

  • <size> = {samples} Size of acquired data in samples

  • <buffer> Array to store the data into. For Python API use rp_createBuffer and for C++ API use malloc.

STEMlab 125-14 4-Input only (additional):

  • <n> = {3,4} (set channel IN3, or IN4)

Available Jupyter and API macros:

  • Fast analog channels - RP_CH_1, RP_CH_2

STEMlab 125-14 4-Input only (additional):

  • Fast analog channels - RP_CH_3, RP_CH_4

SCPI

API, Jupyter

DESCRIPTION

ECOSYSTEM
ACQ:AXI:SOUR<n>:TRIG:FILL?
Example:
ACQ:AXI:SOUR1:TRIG:FILL? > 1
C++: rp_AcqAxiGetBufferFillState(rp_channel_t channel, bool* state)

Python: rp_AcqAxiGetBufferFillState(<channel>)

Indicates whether the Deep Memory Acquisition buffer is full of data.

2.00-18 and up
ACQ:AXI:SOUR<n>:Write:Pos? > <pos>
Example:
ACQ:AXI:SOUR1:Write:Pos? > 1024
C++: rp_AcqAxiGetWritePointer(rp_channel_t channel, uint32_t* pos)

Python: rp_AcqAxiGetWritePointer(<channel>)
Returns current position of the Deep Memory Acquisition write pointer.

2.00-18 and up
ACQ:AXI:SOUR<n>:Trig:Pos? > <pos>
Example:
ACQ:AXI:SOUR1:Trig:Pos? > 512
C++: rp_AcqAxiGetWritePointerAtTrig(rp_channel_t channel, uint32_t* pos)

Python: rp_AcqAxiGetWritePointerAtTrig(<channel>)
Returns position of Deep Memory Acquisition write pointer at time when
the trigger arrived.

2.00-18 and up
ACQ:AXI:SOUR<n>:DATA:Start:N? <pos>,<size>
Example:
ACQ:AXI:SOUR1:DATA:Start:N? 20,3 >
{1.2,3.2,-1.2}
C++: rp_AcqAxiGetDataV(rp_channel_t channel, uint32_t pos, uint32_t* size, float* buffer)
rp_AcqAxiGetDataRaw(rp_channel_t channel,  uint32_t pos, uint32_t* size, int16_t* buffer)
Python: rp_AcqAxiGetDataV(<channel>, <pos>, <size>, <buffer>)
rp_AcqAxiGetDataRaw(<channel>, <pos>, <size>, <buffer>)
Read count samples from the pos position onwards.
SCPI: Returns the value as a text array of values or a byte array.
Depending on the ACQ:AXI:DATA:UNITS setting.
API: Returns the Deep Memory buffer in specified units from specified
position and desired size.

2.00-18 and up
‘’
Python: rp_AcqAxiGetDataVNP(<channel>, <pos>, <np_buffer>) (Numpy buffer dtype=np.float32)
rp_AcqAxiGetDataRawNP(<channel>, <pos>, <np_buffer>) (Numpy buffer dtype=np.int16)
Copies the captured DMA buffer data into the passed NumPy buffer from pos
onwards. The length of the copied data matches the np_buffer length.
Numpy buffer must have the specified dtype format.
Faster than the Python functions above.

2.05-37 and up
‘’
Python: rp_AcqAxiGetDataRawDirect(<channel>, <pos>, <size>)
Returns a memory region without copying data (the fastest method).
Use frombuffer function to extract the data.

2.07-43 and up

Deep Memory Generation (DMG)

DMG settings

Parameter options:

  • <n> = {1,2} (set channel IN1 or IN2)

  • <byte> = {0...} in bytes

  • <start> = {0...} in bytes

  • <end> = {0...} in bytes

  • <state> = {ON, OFF} Default: OFF

  • <decimation> = {1, 2, 4, 8, 16, 17, 18, 19, ..., 65534, 65535, 65536} Default: 1

  • <array> = {value1, ...} Floats in the range -1 to 1

  • <enable> = {True, False} Default: False

  • <start>, <end> = {byte} Start and end address of reserved memory

Available Jupyter and API macros:

  • Fast analog channels - RP_CH_1, RP_CH_2

SCPI

API, Jupyter

DESCRIPTION

ECOSYSTEM
GEN:AXI:START? > <byte>
Example:
GEN:AXI:START? > 16777216
C++: rp_GenAxiGetMemoryRegion(uint32_t *_start, uint32_t *_size)

Python: rp_GenAxiGetMemoryRegion()
Returns the start address of the Deep Memory region.
API: Also returns the size of the memory region.

2.07-48 and up
GEN:AXI:SIZE? > <byte>
Example:
GEN:AXI:SIZE? > 2097152
C++: rp_GenAxiGetMemoryRegion(uint32_t *_start, uint32_t *_size)

Python: rp_GenAxiGetMemoryRegion()
Get size of reserved memory for Deep Memory mode.
API: Also returns the start address of the memory region.

2.07-48 and up
SOUR#:AXI:RESERVE > <start>,<end>
Example:
SOUR1:AXI:RESERVE 2097152,2197152
C++: rp_GenAxiReserveMemory(rp_channel_t channel, uint32_t start, uint32_t end)

Python: rp_GenAxiReserveMemory(<channel>, <start>, <end>)
Reserve an allocated area within the Deep Memory region for generation
(number of Bytes).

2.07-48 and up
SOUR#:AXI:RELEASE
Example:
SOUR#:AXI:RELEASE
C++: rp_GenAxiReleaseMemory(rp_channel_t channel)

Python: rp_GenAxiReleaseMemory(<channel>)
Release the reserved generation memory in the Deep Memory region.

2.07-48 and up
SOUR#:AXI:ENable <state>
Example:
SOUR1:AXI:ENable ON
C++: rp_GenAxiSetEnable(rp_channel_t channel, bool enable)

Python: rp_GenAxiSetEnable(<channel>, <enable>)
Enables the Deep Memory Generation for the specified channel.
Memory region must be reserved beforehand.

2.07-48 and up
SOUR#:AXI:ENable? > <state>
Example:
SOUR1:AXI:ENable? > ON
C++: rp_GenAxiGetEnable(rp_channel_t channel, bool* enable)

Python: rp_GenAxiGetEnable(<channel>)
Retrieves the enable state of the Deep Memory Generation for the specified
channel.

2.07-48 and up
SOUR#:AXI:DEC <decimation>
Example:
SOUR1:AXI:DEC 1
C++: rp_GenAxiSetDecimationFactor(rp_channel_t channel, uint32_t decimation)

Python: rp_GenAxiSetDecimationFactor(<channel>, <decimation>)
Sets the Data Delay Mode value (generation decimation) for Deep Memory
Generation. Each sample remains on the DAC for <decimation> clock cycles,
effectively reducing the output sample rate.

2.07-48 and up
SOUR#:AXI:DEC? > <decimation>
Example:
SOUR1:AXI:DEC? > 1
C++: rp_GenAxiGetDecimationFactor(rp_channel_t channel, uint32_t* decimation)

Python: rp_GenAxiGetDecimationFactor(<channel>)
Retrieves the Data Delay Mode value (generation decimation) for Deep Memory
Generation.

2.07-48 and up
-
C++: rp_GenAxiWriteWaveform(rp_channel_t channel, float* buffer, uint32_t length)

Python: rp_GenAxiWriteWaveform(<channel>, <np_buffer>)
Copies data from the NumPy buffer into the Deep Memory Generation region. The
data must be in float format and the length of the buffer must be half the
reserved DMG region (each float is converted into two int16). Values should
be within full scale output range (±1 V).

2.07-48 and up
SOUR#:AXI:OFFSET#:DATA# <array>
Example:
SOUR1:AXI:OFFSET0:DATA256 1,2,1,..
C++: rp_GenAxiWriteWaveformOffset(rp_channel_t channel, uint32_t offset, float* buffer, uint32_t length)

Python: rp_GenAxiWriteWaveformOffset(<channel>, <offset>, <np_buffer>)
Copies data from the NumPy buffer into the Deep Memory Generation region. The
data must be in float format and the length of the buffer must be half the
reserved DMG region (each float is converted into two int16). Values should
be within full scale output range (±1 V).
Unlike the previous function, this one adds an offset in the separators to
allow writing data in portions.

2.07-48 and up
SOUR#:AXI:SET:CALIB
Example:
SOUR1:AXI:SET:CALIB
C++: rp_GenSetAmplitudeAndOffsetOrigin(rp_channel_t channel)

Python: rp_GenSetAmplitudeAndOffsetOrigin(<channel>)
Applies the DAC calibration values to the DMG waveform.

2.07-48 and up