826

Model 826 board

This is the technical wiki page for Sensoray's model 826, a versatile analog and digital I/O system on a PCI Express board. The board has 48 digital I/Os with edge detection, sixteen 16-bit analog inputs, eight 16-bit analog outputs, six 32-bit counter channels, a watchdog timer with fail-safe controller, and a flexible signal router.

Counters

Snapshot counts upon match

When a snapshot is caused by counts equal to a compare register, the snapshot counts will always equal the compare register value. Similarly, when a snapshot is caused by counts reaching zero, the snapshot counts will always be zero.

How to use interrupts

I want to use interrupts to perform actions after a time delay. I see that counters have an "Interrupt System (IRQ)" signal but have no idea how to access it.

An interrupt request (IRQ) is generated whenever a counter snapshot is captured. Every IRQ source on the 826 is associated with a blocking API function, which manages IRQs for you so that you need not be concerned with the complexities of interrupts. In the case of counters, the blocking API function is S826_CounterSnapshotRead(). Simply call this API function and counter IRQs will automatically be configured and handled.

A simple way to implement a delayed interrupt is to preload the counter with the desired time delay, configure it to count down, and have it capture a snapshot (and thus generate an IRQ) when it reaches zero counts. To wait for the interrupt, call S826_CounterSnapshotRead() with a non-zero tmax value (maximum wait time, which must be longer than the delay time). The function will return upon interrupt and you can then perform the desired actions. For example:

#include "826api.h"

// Wait 0.5 seconds while other threads are allowed to run ------------------

// Configure the delay timer:
S826_CounterModeWrite(0, 0, 0x01400020);      // Configure counter0: 1 MHz down counter, auto preload @startup.
S826_CounterPreloadWrite(0, 0, 0, 500000);    // Delay time in microseconds (0.5 seconds).
S826_CounterSnapshotConfigWrite(0, 0,         // Configure snapshots:
    S826_SSRMASK_ZERO                         //  capture snapshot when counts==0
    | (S826_SSRMASK_ZERO << 16),              //  disable subsequent snapshots when counts==0
    S826_BITSET);                             //  don't alter any other snapshot enables

// Now do the delay:
S826_CounterStateWrite(0, 0, 1);              // Start the delay timer running.
S826_CounterSnapshotRead(0, 0,                // Block while waiting for timer:
   NULL, NULL, NULL,                          //  ignore snapshot counts, timestamp and reason
   S826_WAIT_INFINITE);                       //  don't timeout

printf("Delay time has elapsed!");            // TODO: INSERT YOUR DESIRED ACTIONS HERE

Note that the calling thread cannot do anything else while it waits for S826_CounterSnapshotRead() to return, although other threads can still do productive work.

Periodic timer

How can I use a counter to call a function periodically?

Configure the counter so that it repeatedly counts down to zero and then preloads. The preload value determines the time period. A snapshot is captured every time zero counts is reached, which causes S826_CounterSnapshotRead() to return, whereupon you can call your periodic function.

First, configure the counter and start it running:

#include "826api.h"

// Configure counter so it can be used to call PeriodicFunction() 10 times per second

S826_CounterModeWrite(0, 0, 0x01C02020);      // Counter0: 1 MHz down counter; preload @startup and counts==0.
S826_CounterPreloadWrite(0, 0, 0, 100000);    // Period in microseconds (0.1 seconds).
S826_CounterSnapshotConfigWrite(0, 0,         // Configure snapshots:
    S826_SSRMASK_ZERO,                        //  capture snapshot when counts==0
    S826_BITWRITE);                           //  disable all other snapshot triggers
S826_CounterStateWrite(0, 0, 1);              // Start the periodic timer running.

Now execute the following code in its own thread. This code uses hardware interrupts, which allows other threads to run while this one waits for the next periodic counter interrupt.

// Repeat while no errors detected:
while (S826_CounterSnapshotRead(0, 0,         // Block until next period.
      NULL, NULL, NULL,                       //    (ignore snapshot values)
      S826_WAIT_INFINITE) = S826_ERR_OK)      //    (disable function timeout)
  PeriodicFunction();                         // Execute the periodic function.

Alternatively, if you need to perform other processing while waiting for the next period and cannot use another thread to do so, you can poll the counter by calling S826_CounterSnapshotRead() with a zero wait time:

while (1) { // Repeat forever:
  uint errcode = S826_CounterSnapshotRead(    // Poll to see if period has elapsed (but don't block).
      0, 0, NULL, NULL, NULL, 0);
  if (errcode == S826_ERR_OK)                 // If it's a new period
    PeriodicFunction();                       //  execute the periodic function.
  else if (errcode != S826_ERR_NOTREADY)      // Else if fatal error detected
    break;                                    //  exit the polling loop.
  else                                        // Else
    DoSomeOtherStuff();                       //  do other processing.
}

Incremental encoders

Encoder wiring

I want to connect two TTL/CMOS incremental encoders to counter channels 0 and 1 — what's the correct way to do this?

Connect the encoders as shown below. If the encoder has a "Z" (index) output, connect Z to the counter's +IX input.

Programming fundamentals

Which functions should I use for incremental encoders?

The flexible counter architecture allows for many options, but basic operation works as follows:

First configure and enable the counter channel:

#include "826api.h"

S826_CounterModeWrite(0, 0, 0x00000070);  // Configure counter 0 as incremental encoder interface.
S826_CounterStateWrite(0, 0, 1);          // Start tracking encoder counts.

To read the instantaneous encoder counts without invoking a snapshot:

uint counts;
S826_CounterRead(0, 0, &counts);          // Read current encoder counts.
printf("Encoder counts = %d\n", counts);  // Display encoder counts.

When reading instantaneous counts you may need to know when the counts were sampled. You could rely on your software and operating system to sample the counts at precise times, but there's an easier and more accurate way: trigger a snapshot (via software) and then read the counts and sample time — accurate to within one microsecond:

uint counts;      // encoder counts when the snapshot was captured
uint timestamp;   // time the snapshot was captured

S826_CounterSnapshot(0, 0);               // Trigger snapshot on counter 0.
S826_CounterSnapshotRead(0, 0,            // Read the snapshot:
    &counts, &timestamp, NULL,            //  receive the snapshot info here
    0);                                   //  no need to wait for snapshot; it's already been captured
printf("Counts = %d at time = %d\n", counts, timestamp);

For example, two snapshots allow you to measure speed:

uint counts0, counts1;      // encoder counts when snapshot was captured
uint tstamp0, tstamp1;      // timestamp when snapshot was captured

S826_CounterSnapshot(0, 0);               // Trigger first snapshot.

// TODO: WAIT AWHILE TO ALLOW ENCODER TO MOVE

S826_CounterSnapshot(0, 0);               // Trigger second snapshot.
S826_CounterSnapshotRead(0, 0,            // Read the first snapshot:
    &counts0, &tstamp0, NULL,             //  receive the snapshot info here
    0);                                   //  no need to wait for snapshot; it's already been captured
S826_CounterSnapshotRead(0, 0,            // Read the second snapshot:
    &counts1, &tstamp1, NULL,             //  receive the snapshot info here
    0);

printf("Speed (counts/second) = %d\n", 1000000 * (counts1 - counts0) / (tstamp1 / tstamp0));

Encoder counts can be changed to an arbitrary value at any time. This is typically done when the encoder is at a known reference position (e.g., at startup or whenever mechanical registration is required), but not at other times as it would disrupt position tracking. To change the counts, write the new counts value to the Preload0 register and then call S826_CounterPreload() to force a preload:

S826_CounterPreloadWrite(0, 0, 0, 12345); // Write desired counts value (12345) to Preload0 register.
S826_CounterPreload(0, 0, 0, 0);          // Jam Preload0 value into counter.

Using interrupts with encoders

This code snippet employs hardware interrupts to block the calling thread until the encoder counts equals a particular value. Other threads are allowed to run while the calling thread waits for the counter to reach the target value. A snapshot is captured when the target count is reached. The snapshot generates an interrupt request, which in turn causes S826_CounterSnapshotRead() to return. The example ignores the snapshot counts (which will always equal the target value as explained here), the timestamp, and the reason code (which will always indicate a Match0 event).

S826_CounterCompareWrite(0, 0, 0,         // Set Compare0 register to target value:
    5000);                                //  5000 counts (for this example)
S826_CounterSnapshotConfigWrite(0, 0,     // Enable snapshots:
    S826_SSRMASK_MATCH0,                  //  when counts==Compare0
    S826_BITWRITE);                       //  disable all other snapshot triggers
S826_CounterSnapshotRead(0, 0,            // Wait for counter to reach target counts:
    NULL, NULL, NULL,                     //  ignore snapshot counts, timestamp and reason
    S826_WAIT_INFINITE);                  //  disable function timeout
printf("Counter reached target counts");

In some cases you may want to wait for two different count values at the same time. The following example shows how to wait for the encoder counts to reach an upper or lower threshold (whichever occurs first). Furthermore, it will only wait for a limited amount of time. To set this up, the threshold values are programmed into Compare registers and then snapshots are enabled for matches to both Compare registers. To set a limit on how long to wait, a time limit value is specified by tmax when calling S826_CounterSnapshotRead.

Since snapshots can now be caused by different events, we must know what triggered a snapshot in order to decide how to handle it; this is indicated by the snapshot's reason flags. Alternatively, we could use the snapshot counts value, which will always equal the threshold value that triggered the snapshot.

uint counts;      // encoder counts when the snapshot was captured
uint timestamp;   // time the snapshot was captured
uint reason;      // event(s) that caused the snapshot
uint errcode;     // API error code

S826_CounterCompareWrite(0, 0, 0, 3000);  // Set Compare0 register to low limit (3000 counts)
S826_CounterCompareWrite(0, 0, 1, 4000);  // Set Compare1 register to high limit (4000 counts)
S826_CounterSnapshotConfigWrite(0, 0,     // Enable snapshots:
    S826_SSRMASK_MATCH0                   //  when counts==low limit
    | S826_SSRMASK_MATCH1,                //  or when counts==high limit
    S826_BITWRITE);                       //  disable all other snapshot triggers
errcode = S826_CounterSnapshotRead(0, 0,  // Wait for a snapshot:
    &counts, &timestamp, &reason,         //  receive the snapshot info here
    10000000);                            //  timeout if wait exceeds 10 seconds (10000000 us)

switch (errcode) {                        // Decode and handle the snapshot:
  case S826_ERR_NOTREADY:
    S826_CounterRead(0, 0, &counts);      //  snapshot not available, so read counts manually
    printf("Timeout -- counter didn't reach either threshold within 10 seconds; current counts = %d", counts);
    break;
  case S826_ERR_OK:
    if (reason & S826_SSRMASK_MATCH0)
      printf("Counter reached upper threshold at timestamp %d", timestamp);
    if (reason & S826_SSRMASK_MATCH1)
      printf("Counter reached lower threshold at timestamp %d", timestamp);
}

Measuring velocity

How can I use a counter to measure vehicle velocity at approximately 10 samples/second?

Velocity measurements depend on having the ability to determine the vehicle's displacement (distance traveled) at a known time. Counter snapshots are ideally suited for this task because they include both displacement and time information. To set this up, configure the counter to track displacement (by counting quadrature clocks without interruption), and then trigger 10 snapshots/second using one of these methods:

1. If the counter's IX input is not being used, select the internal tick generator (set to 10 Hz) as the IX signal source, and configure the counter to capture snapshots on IX rising edges.
2. If IX is already being used, create a thread that executes approximately 10 times per second (need not be exact). Each time the thread runs, have it call S826_CounterSnapshot to trigger a soft snapshot. In this case the sample rate will not be exactly 10 Hz, but velocity will still be measured precisely.

Create another thread to handle the counter shapshots. Each time this thread receives a new snapshot, it can use the counts/timestamp pairs from the new and previous snapshots to precisely compute the velocity (in counts per microsecond):

velocity = (counts_new – counts_prev) / (timestamp_new – timestamp_prev)

Output pulse every N encoder pulses

Can the 826 generate an output pulse every N shaft encoder pulses?

A short (20 ns) pulse can be generated with one counter ("counterA") and one general-purpose I/O (DIO). A longer output pulse can be generated with two counters ("counterA" and "counterB") and one DIO.

First, initialize the 826:

• Configure counterA as an incremental encoder interface.
• Configure counterA to capture snapshots upon Compare register matches.
• Configure counterA's ExtOut mode (OM=1) to output a pulse upon Compare register match.
• Case 1: Short output pulse:
  • Program the signal router to output counterA's ExtOut signal on the DIO. The pulse duration will be 20 ns and consequently an external pull-up must be added to the DIO to speed up the rising edge of the output pulse (see Using external pull-up resistors for details).
• Case 2: Output pulse with programmable duration:
  • Configure counterB as a pulse generator, using counterA's ExtOut as a preload trigger. Program the preload counts to the desired pulse width.
  • Program the signal router to output counterB's ExtOut signal on the DIO.

After initializing, program counterA's Compare register with the encoder counts that are to trigger the next output pulse, then wait for a snapshot. When the counts matches the Compare register, a snapshot will be captured and the output pulse will automatically be generated. Upon receiving the snapshot, the application must write into the Compare register the counts corresponding to the next output pulse, before the counter reaches that value.

Encoder FAULT output

My encoder has a FAULT output. Can the 826 monitor the FAULT signal and notify the processor when it changes state?

Some incremental encoders output a FAULT (or FLT) signal to indicate problems such as damaged bearings, code disc defects/contamination, or malfunctioning LEDs/detectors. The FLT signal is typically conveyed over an RS-422 differential pair, which is compatible with the 826's counter inputs. This hardware compatibility and the counter's flexible architecture allow it to monitor the FLT signal and automatically notify the processor when FLT changes state. Note that the general technique discussed here can be used to monitor the state of any RS-422 signal.

To implement this, connect FLT to the counter's IX input. If FLT is active-low, swap the differential pair between line driver and receiver as shown below (don't swap if active-high) so that a disconnected or unpowered encoder will be reported as a fault.

The following code initializes the counter, starts it running, and then monitors the FLT signal state. It should be run in a dedicated thread because it blocks while waiting for FLT state changes. The counter is configured to count up at 1 MHz. A preload is triggered whenever IX is high, and the first Compare0 match triggers a snapshot; these allow the FLT state to be determined when the counter is enabled. Snapshots are also triggered upon IX rising and falling edges; these are used to notify the processor that FLT has changed state.

int FLT_state = -1;    // Public FLT state indicator: -1=unknown, 0=normal, 1=fault

// FLT monitor thread: maintains FLT_state and notifies other threads when FLT_state changes.
int EncoderFaultMonitor(uint board, uint chan)
{
  uint errcode;
  S826_CounterModeWrite(board, chan, 0x01010020);       // Configure the counter channel.
  S826_CounterSnapshotConfigWrite(board, chan, 0x00010019, S826_BITWRITE)
  S826_CounterPreloadWrite(board, chan, 0, 0);
  S826_CounterCompareWrite(board, chan, 0, 1);
  S826_CounterStateWrite(board, chan, 1);               // Enable counter channel.
  errcode = ReadFLT(board, chan, 1);                    // Determine FLT initial state.
  while (errcode == S826_ERR_OK);                       // Monitor FLT state transitions.
    errcode = ReadFLT(board, chan, S826_WAIT_INFINITE);
  return errcode;
}

The following function is called by the above code to determine the state of FLT. Note that you must insert code that notifies other threads (e.g., via semaphore) when FLT_state has changed.

// Wait up to tmax for a FLT state change, then copy the current state to FLT_state.
static int ReadFLT(uint board, uint chan, uint tmax)
{
  uint reason;
  int errcode = S826_CounterSnapshotRead(board, chan, NULL, NULL, &reason, tmax);
  if (errcode == S826_ERR_NOTREADY) {     // If no snapshots captured within finite tmax then
    errcode = S826_ERR_OK;                //   cancel error;
    FLT_state = 1;                        //   counts holding at 0 because FLT asserted.
  } else if (errcode != S826_ERR_OK)      // Else if error detected then
    FLT_state = -1;                       //   abort and declare FLT state unknown.
  else                                    // Else
    FLT_state = ((reason & S826_SSRMASK_IXRISE) != 0);  //   FLT state determined by snapshot trig.
    
  // TODO: NOTIFY OTHER THREADS THAT FLT STATE CHANGED
    
  return errcode;
}

Unexpected snapshots

Why do I occasionally get two snapshots (upon counts match) from my incremental encoder when only one is expected?

Assuming your encoder has not changed direction, the unexpected snapshots are probably caused by noise or slow edges on the encoder clock signals. This is possible even when the snapshot timestamps are identical, because encoder clocks are sampled every 20 nanoseconds whereas timestamp counts are incremented only once per microsecond.

If this is what is happening, unexpected snapshots can be prevented by calling S826_CounterFilterWrite() to establish a clock filter. A small filter value is usually sufficient -- just enough to clean up clock edges, but not so long that valid encoder counts will be missed. For example, this will set the clock filter to 100 ns (index input will not be filtered):

#define FILT_NS    100                // Filter time in ns -- change as desired to multiple of 20.
#define FILT_RES   20                 // Filter resolution in nanoseconds.
#define FILT_CLK   (1 << 30)          // Bit flag to enable clock filtering.
S826_CounterFilterWrite(0, 0,         // Activate clock filter on counter 0.
    FILT_CLK + FILT_NS / FILT_RES);

Another way to handle this is to configure the Match snapshot trigger to become automatically disabled when it fires. Note that if you use this method, you will need to re-enable the Match trigger to capture snapshots of subsequent matches.

S826_CounterSnapshotConfigWrite(0, 5, // Configure snapshots on counter 5:
   S826_SSRMASK_MATCH0                //  enable snapshot upon Match0 (counts==Compare0 register)
   | (S826_SSRMASK_MATCH0 << 16),     //  disable subsequent Match0 snapshots upon Match0 
   S826_BITWRITE);                    //  disable all other snapshot triggers

Timing

Input filters

The ClkA, ClkB and Index inputs are acquired using an internal 50 MHz sampling clock (SysClk) with 3 ns minimum setup time. Consequently, an edge on any of these signals will will be detected 0.3 to 20.3 ns after it occurs. In the below timing diagram, the input edge is detected at SysClk 2 because it occurs after SysClk 1 and ≥ 0.3 ns before SysClk 2.

Each input has a digital filter with programmable time F. A filter's input signal will appear on its output when the input has remained stable for time F. The filter delays the detected edge by F+1 SysClk periods. This diagram show filter timing in the general case:

The filter delay can be minimized by setting F=0, which results in the following timing:

Counter functions

The filtered ClkA/B signals are decoded and used to control the counter. As shown in the following diagram, counts will change one SysClk period after a filter output change and then, one SysClk later, a snapshot is captured to the event FIFO and ExtOut goes active for one SysClk period. If ExtOut is routed to a DIO, it will be delayed one additional SysClk period by the DIO output sampler on its way to the DIO connector pin.

Summary of counter timing parameters:

Timing Parameter	Minimum	Maximum
ClkA/ClkB setup time	3 ns
ClkA/ClkB hold time	0 ns
ClkA/ClkB to counts change	40 ns	60 ns
ClkA/ClkB to counts change	40 ns	60 ns
ClkA/ClkB to snapshot	60 ns	80 ns
ClkA/ClkB to ExtOut (internal)	60 ns	80 ns
ClkA/ClkB to ExtOut (DIO pin)	80 ns	100 ns

Using external pull-up resistors

My PWM stops working when it outputs high frequencies. Why does this happen and how can I prevent it?

The PWM signal (from the counter's ExtOut) is output by a DIO channel. When the DIO driver transitions to the off state, its 10 KΩ source resistance (combined with circuit capacitance) stretches the DIO rise time and thereby delays its transition to logic '1' (see DIO rising edge in the above timing diagram). As the PWM off time decreases, the rise time (which is constant) becomes a higher percentage of the off time. When the off time is too short (i.e., off time < rise time), the PWM output will seem to "stop" because there is not enough time for the signal to reach logic '1'.

This situation can arise when the PWM is operating at high frequencies or generating short positive pulses, and can be avoided by speeding up the DIO rise time.

PWM operation

How to configure a counter for PWM operation

This example shows how to configure counter channel 0 to operate as a PWM generator with the output appearing on DIO channel 0. Note that you must call S826_SafeWrenWrite() before calling S826_DioOutputSourceWrite(); if you neglect to do this then the PWM signal will not appear on DIO 0.

#include "826api.h"

uint data[2]= {1, 0}; // DIO 0
S826_SafeWrenWrite(0, 2);                   // Enable writes to DIO signal router.
S826_DioOutputSourceWrite(0, data);         // Route counter0 output to DIO 0.
S826_CounterModeWrite(0, 0, 0x01682020);    // Configure counter0 for PWM, with auto-preload when starting.
S826_CounterPreloadWrite(0, 0, 0, 900);     // On time in us (0.9 ms).
S826_CounterPreloadWrite(0, 0, 1, 500);     // Off time in us (0.5 ms).
S826_CounterStateWrite(0, 0, 1);            // Start the PWM generator.

Fail-safe PWM generator

I'm using a PWM output to control a motor. Is there a way to automatically shut off the motor if my program crashes?

Yes, you can use the watchdog timer and fail-safe controller to force the PWM output to a constant state. To do this, configure the watchdog to activate safemode when it times out, as shown in this simplified block diagram:

Before enabling the PWM generator or watchdog, program the desired PWM failsafe level into the DIO channel's SafeData register; this specifies the signal that will be sent to your motor controller when your program crashes (which will shut off the motor). Note that the DIO output is active-low. The SafeEnable register is set to '1' by default, thus enabling fail-safe operation on the DIO channel. Next, program the watchdog interval and start the watchdog running. Finally, start the PWM running.

After enabling the watchdog, your program must periodically kick it to prevent it from timing out, by calling S826_WatchdogKick(). When your program is running normally, the PWM signal will appear on the DIO pin. If your program crashes (or fails to kick the watchdog in a timely manner), the watchdog will time-out and activate the fail-safe controller. This will switch the DIO pin to the level specified by the SafeData register, which in turn will halt the motor.

#include "826api.h"

#define CRASH_DET_SECONDS 0.5  // Halt motor if program fails to kick watchdog within this time.
#define MOTOR_HALT_LEVEL  0    // DIO pin level ('0'=5V, '1'=0V) that will halt motor.

uint wdtime[5] = {(uint)(50000000 * (CRASH_DET_SECONDS)), 1, 1, 0, 0}; // watchdog interval
uint dio_routing[2]= {1, 0};                                           // map counter0 to DIO 0
uint safe_data[2]= {MOTOR_HALT_LEVEL, 0};                              // fail-safe level

// Create a fail-safe PWM generator using counter0 and DIO 0.
S826_SafeWrenWrite(0, 2);                   // Enable writes to watchdog, router and SafeData.
S826_DioSafeWrite(0, safe_data, 2);         // Specify DIO state to use when program crashes.
S826_DioOutputSourceWrite(0, dio_routing);  // Route counter0 output to DIO 0.
S826_CounterModeWrite(0, 0, 0x01682020);    // Config counter0 for PWM; preload when starting.
S826_CounterPreloadWrite(0, 0, 0, 900);     // PWM on time in us (0.9 ms).
S826_CounterPreloadWrite(0, 0, 1, 500);     // PWM off time in us (0.5 ms).
S826_WatchdogConfigWrite(0, 0x10, wdtime);  // Set wdog interval; trig safemode upon timeout.

// Start the PWM generator running.
S826_WatchdogEnableWrite(0, 1);             // Start watchdog. PROGRAM MUST KICK IT FROM NOW ON!
S826_CounterStateWrite(0, 0, 1);            // Start the PWM generator.

Phase-locked PWM outputs

Some applications require multiple, phase-locked PWM outputs. Although the 826's counter channels do not directly support phase locking, it is possible to simulate phase-locked PWM outputs by using counters configured as hardware-triggered one-shots. This technique can be used in a variety of ways.

Example: quadrature generator

A quadrature generator can be implemented with three counter channels and two DIOs as shown below. Except for DIO load connections, no external wiring is required (counter and DIO interconnects are established by the board's programmable signal router).

The counter channels are configured as follows:

• CH0 - PWM with output on DIO1.
• CH1 - 1-shot with IndexSource=ExtOut0, preload upon Index. This delays Phase2 wrt Phase1.
• CH2 - 1-shot with IndexSource=ExtOut1, preload upon Index, output on DIO2. This generates the Phase2 output pulse.

Example: 3-phase controller

A 3-phase PWM controller (shown below) can be created by extending the above example. This is implemented with five counter channels and three DIOs (channel numbers are arbitrarily assigned). Except for DIO load connections, no external wiring is required (counter and DIO interconnects are established by the board's programmable signal router).

In the above example, counter channels are configured as follows:

• CH0 - PWM with output on DIO1.
• CH1 - 1-shot with IndexSource=ExtOut0, preload upon Index. This delays Phase2 wrt Phase1.
• CH3 - 1-shot with IndexSource=ExtOut1, preload upon Index. This delays Phase3 wrt Phase2.
• CH2 - 1-shot with IndexSource=ExtOut1, preload upon Index, output on DIO2. This generates the Phase2 output pulse.
• CH4 - 1-shot with IndexSource=ExtOut3, preload upon Index, output on DIO3. This generates the Phase3 output pulse.

Serial data capture

A counter channel can be used to capture serial data by leveraging its snapshot FIFO. When the counter has been appropriately configured and enabled, the computer allows the channel hardware to acquire snapshots while it asynchronously reads and processes snapshots from the FIFO.

Serial data capture (asynchronous)

A timestamp is included in every snapshot, which is especially useful for capturing data from asynchronous sources and irregularly-timed sources such as bar code wands. The basic idea is to apply the serial data signal to the counter's IX or ExtIn input and have the counter automatically capture snapshots at signal edges.

In each snapshot, only the timestamp and reason flags are of interest (counts are ignored). The reason flag indicates whether a snapshot was triggered by rising or falling edge and the timestamp indicates the time when the edge occurred. For example, in the serial data waveform shown below, the first rising edge (A) caused a snapshot to be captured when the timestamp generator value was 100, and the reason code indicates the snapshot was triggered by a rising edge.

Any two consecutive snapshots represent a matched pair of rising/falling or falling/rising edges, corresponding to an interval during which the serial data value was '1' or '0', respectively. For example, in the waveform shown above, snapshots A and B bracket a '1' interval. It is possible to determine the binary value of the serial data from the first reason code, and the duration of the data value from the difference between the timestamps.

To see how this works, consider the above serial data waveform. The counter automatically captures a snapshot for each of the edge events A, B, C and D. When the computer considers snapshots A and B, it determines that the serial data was '1' during interval A-B because A was triggered by a serial data rising edge. Furthermore, it knows that the serial data held at '1' for 300 microseconds (the difference between the A and B timestamps). Similary, it can determine that interval B-C was a logic '0' lasting 200 microseconds, and C-D was a 400 microsecond logic '1'.

For maximum efficiency, consider using a dedicated thread to read the FIFO. This eliminates polling and makes the application event-driven because the thread can wait in S826_CounterSnapshotRead() for the next snapshot without wasting CPU time. Also, this decouples the timing of serial data acquisition from other tasks, thereby greatly simplifying overall software development and maintenance. If fast processing of the serial data is required, raise the thread priority to an appropriately high level.

Serial data capture (synchronous)

To capture clocked serial data (e.g., SPI, I²C), apply the clock signal to the counter's ExtIn or IX input and configure the counter to capture snapshots at either the rising or falling edge of the clock signal (whichever edge indicates stable data). Apply the data signal to the counter's ClkA input and set the counter's clock mode K=6 (external quadrature clock, x2 multiplier); this will cause the counts to change (it will alternate between two values) upon each data edge.

A serial data bit can be obtained directly from the counts lsb (least significant bit) of each snapshot. Timestamps may be used to identify data packet boundaries, or ignored if frame markers are encoded in the data.

An open-source demo of this technique is available at synchronous serial receiver.

Mode register decoder utility

Is there an easy way to convert a mode register value to a human readable description of counter settings?

Yes: download Sensoray's counter mode decoder utility program (Windows compatible). Run the program and enter the mode register value to see an English language description of the counter mode settings.

ADC

Handling ADC interrupts

An interrupt request (IRQ) is generated whenever the ADC subsystem completes a conversion burst. The IRQ works in conjunction with the blocking API function S826_CounterSnapshotRead(). To handle ADC interrupts, simply call S826_CounterSnapshotRead() from any thread; the function will return upon ADC interrupt.

The following example shows how this works. In the example function, only slot 0 is of interest and timestamps are not used. Note that the ADC (and its trigger source) must have been previously configured and enabled.

void AdcHandler(void)
{
  uint errcode;
  int slotval[16];  // buffer must be sized for 16 slots
  while (1) {
    uint slotlist = 1;  // only slot 0 is of interest
    errcode = S826_AdcRead(0, adcdata, NULL, &slotlist, S826_WAIT_INFINITE); // wait for IRQ
    if (errcode != S826_ERR_OK)
      break;
    printf("Raw adc data = %d", slotval[0] & 0xFFFF);
  }
}

Periodic ADC conversions (self-paced)

Can the ADC periodically acquire samples without using a counter to pace it?

Yes: Configure the ADC for continuous triggering mode and use slot settling times to control the sampling rate. Note that the sampling rate will have a small amount of jitter (<= 1 microsecond) because ADC conversion time is 2-3 microseconds. This example shows how to acquire 20 samples per second from analog input channel 0:

#define SAMPLING_PERIOD 50000         // Sampling period in microseconds (50000 = 20 samples/s).
#define TSETTLE SAMPLING_PERIOD - 3;  // Compensate for nominal ADC conversion time.

// Configure the ADC subsystem and start it running
S826_AdcSlotConfigWrite(board, 0, 0, TSETTLE, S826_ADC_GAIN_1); // measuring channel 0 on slot 0
S826_AdcSlotlistWrite(board, 1, S826_BITWRITE);                 // enable slot 0
S826_AdcTrigModeWrite(board, 0);                                // trigger mode = continuous
S826_AdcEnableWrite(board, 1);                                  // enable conversions

AdcHandler();   // Handle periodic ADC interrupts (using code from earlier example)

Calibration errors caused by missing shunt

On 826 SDKs earlier than version 3.2.0, analog calibration values will not be applied without J6 (labeled "Calibration Enable") installed. A missing shunt is intended to protect against accidental overwriting of calibration values, but in these SDKs it also prevents the reading of those values. This is resolved in SDK version 3.2.0 and above; in these versions the shunt functions as intended and must be installed only when calibrating the board (though leaving it installed all the time is okay).

If board calibration is incorrect, make sure J6 is installed or upgrade to SDK version 3.2.0 or higher. The 826 SDK can be downloaded from the 826 product page.

Apparent nonlinearity

To prevent high CMV, connect isolated source to ADC ground.

ADC linearity can be adversely affected by high common-mode voltage (CMV). This can happen when the ADC is used to measure an isolated voltage source such as a battery, thermocouple, or isolated power supply. Since the source is isolated, the CMV may float up or down until it exceeds the maximum allowed CMV of the ADC's input circuitry.

When measuring an isolated source, be sure to connect one side of the source to the ADC power supply ground as shown in the diagram to the right. This will prevent high CMV that might othewise result in apparent non-linearity or calibration errors.

ADC accuracy specification

Resolution and no missing codes are both 16 bits minimum.

Parameter	Value			Units	Description
	Min	Typ	Max
Integral Nonlinearity Error		±0.75	±1.5	LSB	Deviation of ADC transfer function from best-fit line
Differential Nonlinearity Error		±0.5	±1.25	LSB	Deviation of ADC code width from ideal code width
Gain Error		±2	±40	LSB	Deviation of difference between actual level of last data transition and actual level of first transition from difference between ideal levels
Gain Error Temperature Drift		±0.3		ppm/°C
Zero Error			±0.8	mV	Difference between ideal midscale voltage and actual voltage producing the midscale output code
Zero Temperature Drift		±0.3		ppm/°C

Maximum input voltage

The analog inputs accept common mode voltages up to ±12V with no resulting input current or damage. CMV up to ±25V is tolerated continuously, though this will cause currents to flow in the analog inputs. CMV greater than 25V may be tolerated for brief intervals, but this can cause significant currents to flow in the analog inputs and is not specified nor guaranteed to be safe.

DAC

Specifying setpoint in Volts

Is there a way to program the analog outputs using Volts units?

The following function will set a DAC output to the specified voltage. Note that the function does not check for illegal voltage values.

int SetDacOutput(uint board, uint chan, double volts)
{
  uint setpoint;
  uint range;
  int errcode = S826_DacRead(board, chan, &range, &setpoint, 0);  // get DAC output range
  switch (range)
  {
    case S826_DAC_SPAN_0_5:   setpoint = (uint)(volts * 0xFFFF /  5); break;          // 0 to +5V
    case S826_DAC_SPAN_0_10:  setpoint = (uint)(volts * 0xFFFF / 10); break;          // 0 to +10V
    case S826_DAC_SPAN_5_5:   setpoint = (uint)(volts * 0xFFFF / 10) + 0x8000; break; // -5V to +5V
    case S826_DAC_SPAN_10_10: setpoint = (uint)(volts * 0xFFFF / 20) + 0x8000; break; // -10V to +10V
  }
  if (errcode == S826_ERR_OK)
    errcode = S826_DacDataWrite(board, chan, setpoint, 0);
  return errcode;
}

Linux Demo Sine Wave Generator counter error (-15)

The Linux sine wave generator demo may experience a timeout and exit with an error code -15 (S826_ERR_FIFOOVERFLOW). This occurs because the priority of the demo thread may be too low for the sample time. Linux is not a RTOS and the process (or interrupt) may be delayed and not complete the DAC output in the specified time.

Older version of the demo will exit when S826_ERR_FIFOOVERFLOW occurs. Later versions of the demo, however, will print an error code and continue outputting the sine wave.

In any case, if the DAC output sampling time requirements are very small and need to be precise, it is recommended to run the process at a higher priority. You may also consider using a low-latency or rt kernel.

To run the demo at a higher priority:

"nice -n 19 ./s826demo"

For Ubuntu low-latency kernel:

"sudo apt-get install linux-lowlatency linux-headers-lowlatency"

For Ubuntu rt kernel:

"sudo apt-get install linux-rt linux-headers-rt"

In extreme high performance cases, you may consider using the raw DAC write command (S826_DacRawWrite) instead of S826_DacDataWrite. You must make sure to understand the DAC ranges before doing so. This should normally not be necessary as S826_DacDataWrite is only marginally slower.

DAC accuracy specification

Resolution and monotonicity are both 16 bits minimum.

PARAMETER	CONDITIONS	VALUE			UNITS
		MIN	TYP	MAX
Integral Nonlinearity				±2	LSB
Differential Nonlinearity				±1	LSB
Gain Error			±4	±20	LSB
Gain Temperature Coefficient			±2		ppm/°C
Unipolar Zero-Scale Error	5V unipolar range, 25°C 10V unipolar range, 25°C 5V unipolar range 10V unipolar range		±80 ±100 ±140 ±150	±200 ±300 ±400 ±600	µV µV µV µV
V_offset Temperature Coefficient	All unipolar ranges		±2		µV/°C
Bipolar Zero Error	All bipolar ranges		±2	±12	LSB

DIOs

Controlling output rise-time with an external pull-up

Each DIO channel has an output driver that actively drives the signal to 0 V in the on state and is high-impedance in the off state. In the off state, the channel's internal 10 KΩ resistor passively pulls up the signal to +5 V. This 10 KΩ source resistance (combined with circuit capacitance) stretches the DIO rise time, which delays its transition to logic '1'. The rise time can be shortened by adding an external pull-up resistor (and reducing external capacitance) on the signal net.

The following table shows nominal rise times for unloaded DIO pins. Lower resistance values may be needed to compensate external circuit capacitance, but do not use an external pull-up resistor having less than 220 Ω (to prevent excessive DIO output current).

External pull-up	DIO rise time
None	200 ns
1.2 KΩ	20 ns
680 Ω	10 ns

Generating a burst of pulses

How can I make a DIO go active for 10ms and then inactive for another 10ms, and repeat this five times?

There are several ways to do this but the methods you can use depend on a number of factors. Here are two general strategies:

1. If high precision is not needed you can call S826_DioOutputWrite() multiple times, with 10ms software delays between the calls. The precision of this method may depend on your operating system and CPU load. The code for this method is simple and straightforward.

2. If you need precise timing then you could use two counter channels. For example, using counter channels 0 and 1:

• Counter 0: Configure as PWM generator with 10ms on/off times, with external enable. Connect its enable input to counter 1's output. The goal here is for counter 0 to enable counter 1 to output pulses; counter 0 will enable the PWM generator until 5 pulses have been generated.
• Counter 1: Configure as event down-counter, with preload (value=5) upon enable, with external clock, with output active when counts not zero. Connect its clock input (must be external connection) to counter 0's output. The goal here is for counter 1 to count PWM pulses until it counts from 5 down to 0; it will then set its output low, thus disabling the PWM generator.

Example application: I²C Emulator

Can I use DIOs to communicate with I²C devices?

Have a look at Sensoray's I²C emulator, which uses two DIOs to bit-bang an I²C bus. This open source software implements a full-featured I²C master emulator with bus arbitration and bus-hang resolver. All 826-specific code resides in a hardware abstraction layer (HAL) — an architectural feature that belongs in all production-quality software.

If you need to monitor an I²C bus or emulate a slave device, consider using a counter to capture synchronous serial data.

Interfacing RS-422 signals

I need to monitor the state of an RS-422 signal. Can I do this with a DIO channel?

DIO channels are not compatible with RS-422. However, it is possible to use a counter channel to monitor the state of an RS-422 signal, using the technique described here.

Software

Board IDs

The "BOARD NUM" switches (at top edge of board near mounting bracket) assign the board ID used by software. The ID is binary coded on the four switches and can be programmed to any value from 0 (default) to 15. A board's ID determines the corresponding bit that will be set to '1' in the value returned by S826_SystemOpen. If you have one 826 board, the returned value is (2^ID). If you have multiple boards, the returned value is the sum of (2^ID) for each board.

Examples

• You have one board with ID set to 0, so the value returned by S826_SystemOpen will be (2^0) = 1.
• You have two boards with IDs set to 1 and 4, so the value returned by S826_SystemOpen will be (2^1)+(2^4) = 2+16 = 18.

Custom installation and re-distribution (Windows)

Sensoray's installer uses the NSIS installation system. It is created from a .NSI script. The core API is installed as follows in NSI script code:

Section "Core API"
SectionIn RO
${If} ${RunningX64}
 SetOutPath "$WINDIR\system32";
 !insertmacro  DisableX64FSRedirection
 File "..\mid-826\code\Release64\s826.dll";
 !insertmacro  EnableX64FSRedirection
 SetOutPath "$WINDIR\SysWOW64";
 File "..\mid-826\code\Release\s826.dll";
${Else}
 SetOutPath "$WINDIR\system32";
 File "..\mid-826\code\Release\s826.dll";
${EndIf}
SectionEnd

The drivers are installed via dpinst.exe in the NSI script as follows:

Section "Drivers"
SectionIn RO
CreateDirectory "$INSTDIR\driver\x64";
SetOutPath "$INSTDIR\driver\x64";
File "..\cd\driver\x64\dpinst.exe";
File "..\cd\driver\x64\s826.cat";
File "..\cd\driver\x64\s826.inf";
File "..\cd\driver\x64\s826.sys";
File "..\cd\driver\x64\s826filter.cat";
File "..\cd\driver\x64\s826filter.inf";
File "..\cd\driver\x64\s826filter.sys";
File "..\cd\driver\x64\WdfCoInstaller01009.dll";
CreateDirectory "$INSTDIR\driver\x32";
SetOutPath "$INSTDIR\driver\x32";
File "..\cd\driver\x32\dpinst.exe";
File "..\cd\driver\x32\s826.cat";
File "..\cd\driver\x32\s826.inf";
File "..\cd\driver\x32\s826.sys";
File "..\cd\driver\x32\s826filter.cat";
File "..\cd\driver\x32\s826filter.inf";
File "..\cd\driver\x32\s826filter.sys";
File "..\cd\driver\x32\WdfCoInstaller01009.dll";
MessageBox MB_OK "Driver installation dialog will pop-up. Follow the prompts and click Finish when done"
${If} ${RunningX64}
 ExecWait '"$INSTDIR\driver\x64\dpinst.exe" /f'
${Else}
 ExecWait '"$INSTDIR\driver\x32\dpinst.exe" /f'
${EndIf}
NoInstallDriver:
SectionEnd

What other libraries does the installer install as part of the Core API?

The 826 is compiled with Microsoft Visual Studio C++ 2008. The re-distributables for C++ must be installed. The installer installs this library silently running the command:

"vcredist_x86.exe /q"

and the following additional command on 64-bit systems:

"vcredist_x64.exe /q"

These re-distributables are available from Microsoft for x64 and x86.

Are any other libraries required? I installed the libraries above, but the demo doesn't work with my custom installer?

The demo is written using .NET libraries (version 3.5). These are also available from Microsoft here. The executable can be silently installed using this command:

"dotnetfx35setup.exe /qb"

Could I obtain the full 826 NSI script as a template for creating my own installer?

Yes, the full script is available here.

Silent install (Windows)

I want to run the installer silently. Do you have any way to do this?

There are many options. For re-distribution, you may create your own installation package. Also, starting with version 3.3.9, there are additional command line options to quiet the setup.exe installer from command line or batch file. These options are described below.

What are the options for silent install?

The basic silent install is invoked by running the following command from command line or batch script:

"setup.exe /S"

Please note that the /S is case sensitive and must be upper case.

I've pre-installed the drivers and don't want to re-install them during the installation? Is there a command for that?

"setup.exe /S /no_driver=1"

Is there an additional command to not install the demo programs?

Yes, in version 3.3.9, the following command will install the required DLLs and system libraries, but no drivers or demo programs.

"setup.exe /S /no_driver=1 /no_demos=1"

I want to install the drivers silently, but there is always a pop-up to verify.

Unfortunately, there is no way around this. Windows requires confirmation from the user for driver install, even if the driver is signed.

I'm running the setup silently, but it pops up a dialog to confirm if I want to make changes to the PC (User Account Control). How do I prevent this?

Windows controls this through User Access Control. If running the setup from a standard windows console, the Windows User Account Control (UAC) will pop-up. This cannot be by-passed by Sensoray because the installer installs files to system directories.

One work-around is to launch the setup in an Windows Command Prompt Window started in administrator mode (right-click and select "Run As Administrator"). Another approach is to launch the setup as a user with administrator privileges. User access control may also be disabled, but we do not recommend this for security reasons.

Labview

Before running an 826 virtual instrument (VI) under Labview, make sure you install the latest versions of the 826 DLL (s826.dll) and device driver (both are contained in the 826 SDK, which you can obtain from the Downloads tab of the 826 product page). Each VI is a basically a wrapper for a DLL function and consequently the VIs are dependent on the DLL, which in turn depends on the driver. Board hardware and firmware version numbers will be automatically read from the 826 board by software when all dependencies are satisfied -- it is not necessary to manually enter any board selection information except the board number, which is specified by the board's switch settings (factory default is board number 0).

The VIs are not independently documented, but since each VI wraps a DLL function, the DLL documentation effectively explains the function of each associated VI. The DLL documentation can be found in the 826 product manual (download from the 826 product page Documentation tab).

Software updates

1. Windows 3.3.4

• C# demo application added to SDK. Error checking for invalid modes to S826_CounterModeWrite.

2. Linux 3.3.5

• C# GUI demo available, using Linux mono. To get required libraries on Ubuntu, type:

"sudo apt-get install mono-complete"

For a C# development environment, type:

"sudo apt-get install monodevelop"

Resources for custom driver development

I want to develop my own driver for the 826. Does Sensoray offer any resources for custom driver development?

Yes, we provide these resources free of charge:

• Linux Software Development Kit (SDK) - Includes source code for the 826 driver and middleware, comprising a complete 826 API for Linux. The middleware core is operating system independent and thread-safe, which makes this SDK a great starting point for porting to any operating system. The SDK can be found on the Downloads tab of the 826 product page. The SDK has been carefully designed for reliable operation in multi-threaded and multi-process applications, and consequently it can be easily ported to real-time operating systems.
• Model 826 Technical Manual - This comprehensive manual explains the API and 826 hardware in detail (download from the Documentation tab of the 826 product page).
• Register Map - A map of the board's hardware registers is available here. The registers are accessed through PCI BAR 2. Registers appear in both banked and flat address spaces. The banked space is only required for rev 0 boards; you should use the flat space exclusively if you have a later rev, as this will yield superior performance.

Linux versions

Do you recommend specific Linux distributions for use with the 826?

We no longer support the obsolete kernel 2.4, but otherwise have no specific recommendation as it depends on the application (e.g., it might be desirable to use a low-latency kernel). We normally test first on Ubuntu LTS, but have a script to test builds on kernel versions 2.6.x, 3.x, and 4.x.

VB.NET demo

To help you jump-start your VB.NET project, we created the VB.NET demo for model 826. This demo program provides a GUI for nearly every hardware resource on the board. All source files are provided, including a module that declares all functions, types, and constants for the 826 API.

Modprobe error

When I try "sudo make install" it reports "modprobe: ERROR: could not insert 's826': Required key not available". How can I install the driver?

It's likely that the 826 driver is installed and you are actually seeing a warning. You can confirm this by trying "sudo modprobe s826" and the 826 demo application.

Migrating from model 626

Using 626 cables with the 826

I have a 7505TDIN breakout board and 7501C1 (50-pin cable) for the 626. Can I use these with the 826?

The 7505TDIN and 7501C are both compatible with the 826. However, we recommend using an 826C2 cable instead of the 7501C because it has a low profile header at one end that results in a denser cable stackup. That said, the 7501C cable can be used if it doesn't cause mechanical interference in your system.

Connector pinout differences

Do the 826 and 626 have identical connector pinouts?

The digital and counter connector pinouts are identical. Analog connector pinouts differ slightly because the 826 has four additional analog outputs. The analog pinouts are identical except for pins 41, 43, 45 and 47, which convey DAC channel 4-7 outputs on the 826 (vs. remote sense inputs on the 626).

See also

• GPIO interfacing - design tips for DIO circuits