Multi-Channel Skew Testing with the Femto-2000 TIA

Overview

The Femto 2000's ±30ps accuracy makes it an ideal instrument for making skew measurements across multiple single ended or differential channels. Applications include LVDS, DVI and USB clock-to-data, output-to-output, inter-pair, or intra-pair skew measurements. These devices require skew measurements of less than 100pS that are not easily attainable on conventional ATE equipment.

Test Requirements

In the example of clock to data skew (shown below) the skew measurement is taken from the clock which is the reference channel, to the Data Channels which are the target channels.

Differential and Single Ended Measurements signal measurement capability is required as well as the ability to measure small voltage levels of less than 100mV pk-pk.

The Femto 2000 Solutions Skew Between Channels of a Single Site The ± Time Interval Measurement Mode can be operated from the front panel of the Femto-2000. This two-edge measurement mode observes two input channel signals, and measures the time between transitions on each, with no assumption as to the time ordering of the two measurement events. Transition polarities are independently selectable for each signal; for single-ended signals, switching threshold can be independently set for each signal. Both edges must be armed in unison. The instrument then measures the time of the next measurement event on each input signal, then computes and reports the difference in times.

Where TAn are the time measurements for the input signal on input channel A and TBn are the time measurements for the input signal on input channel B.

Multi-Site Skew Measurements:

The same methodology can also be implemented across multiple channels. This is because the Femto-2000 provides time correlation to the same T0 for all channels and all sites on the machine. This makes it ideal for channel-to-channel skew applications. Currently this feature is only available through the gt_drvr.c command interface. To use it requires the implementation of the following gt_drvr.c command: gt_hl_setup_one_to_many_skew( ref_site_mask, ref_channel, target_chan_mask, ref_polarity, target_polarity, unsigned block_size , unsigned block_count, ref_block_arm_mode, target_block_arm_mode, ref_measurement_arm_mode, target_measurement_arm_mode, ref_arm_event_interval, target_arm_events_interval, ref_threshold_mode, target_threshold_mode, double ref_threshold, double target_threshold, ref_sensitivity_mode, target_sensitivity_mode, target_results); If successful, the function will return GT_SUCCESS otherwise it will return a GT_FAILURE if any error occurred. The valid command arguments are as follows: ref_site_mask - indicates which site has the reference channel. Only 1 site can be set. Any ONE of the following: GT_SITE0 , GT_SITE1 , GT_SITE2 , GT_SITE3. ref_channel - indicates which channel of the reference site is to used as the reference. Only ONE channel can be selected. Choose ONE of the following: GT_CHANNEL_A, GT_CHANNEL_B target_chan_mask - channel(s) to compare to reference. Any combination of : GT_RD_CHAN_0A, GT_RD_CHAN_0B, GT_RD_CHAN_1A, GT_RD_CHAN_1B GT_RD_CHAN_2A, GT_RD_CHAN_2B, GT_RD_CHAN_3A, GT_RD_CHAN_3B or GT_RD_ALL_CHAN If the value is GT_RD_ALL_CHAN, then all channels will be compared to the reference channel, except of course the reference channel won't be compared to itself. If the value is not GT_RD_ALL_CHAN, then including the reference channel, as a target in the target channel mask, will produce an error. ref_polarity - polarity of edges to use for measurement on the reference channel target_polarity - polarity of edges to use for measurement on the target channel GT_POL_POS, GT_POL_NEG - measure using positive/negative edges block_size - number of measurements to make in a block block_count - number of blocks of measurements to make . ref_block_arm_mode - Reference channel block arming mode.

Considerations for Frequencies above 250MHz:

For high-frequencies above 250MHz the user needs to take into account that this 3nS of arm latency is not a precise specification. It can vary across different measurement setups, Femto channels, test system, and cable setups. The cumulative effect of some or all of these may cause the different channels to arm at slightly different points in time. This can have various effects that are outlined below. The time at which the arm is active will determine if the skew will be POS, NEG, or if it will skip a period. Situation 1: Small Pos Skew

The above figure shows the desired measurement setup. The Arm edge occurs prior to the edge we want to measure. After the arming latency it becomes active BEFORE the positive edge of the signal on channel A. The measurement is a small positive skew from A-B. In this example we used 200pS so that it can be drawn easily in the diagram.

Situation 3: Large Positive Skew

The above diagram shows what happens when the arming is misplaced so that it becomes active in BETWEEN the two rising edges. The edge on channel B is missed and it takes the next positive edge on A as the start point. The skew is then measured from the positive edge of A to the positive edge of B which results in a large positive measurement. Situation 4: Large Negative Skew This is the same as Situation 3 with the order of the A and B channel edged reversed. External Arm Calibration: In some cases the user may be required to adjust the timing position of the external arm signal provided to the Femto. In the event that the arming signal is on the border between one measurement scenario or another an unusually high standard deviation in measurement results will occur. A close look at the gt_hl_setup_one_to_many_skew() command reveals that external arm channel setups can only be specified for reference and target channels as groups. To specify separate external arms for each individual channel a set of block and measurement arm commands that must be executed before the test setup is complete. Arming commands issued after gt_hl_setup_one_to_many_skew() will override the previous settings. In the example is shown below, the user has specified explicitly the external arming to be independent for each and every channel of the Femto.