Optical tuning test system using parallel oven pipeline with parallel instrumentation channels and machine learning assistance

Abstract

A parallel oven pipeline with parallel instrumentation channels and a machine learning assisted optical tuning test system are used. A test system includes: a test and measurement instrument; ovens for accommodating a device under test (DUT), each oven having an oven switch selectively connectable to the DUT; a channel switch selectively connected to the oven switch and one channel of the instrument; one or more processors for selecting an oven and oven switches thereof, connecting the oven switches to a subset of the DUTs in the oven, connecting channel switches to the oven switches to receive signals from the subset of the DUTs, transmitting the signals to channels of the instrument to acquire waveforms from the subset of the DUTs in parallel, and transmitting the acquired waveforms to the oven switches. And repeating connecting the channel switch and the oven switch until an instrument has acquired a waveform from each DUT in the oven, tuning each DUT using machine learning, testing whether each DUT in the oven is optimally tuned, and repeating until all DUTs have been tuned and tested.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is a continuation-in-part of U.S. patent application No. 18 / 126,342, filed on March 24, 2023, entitled “OPTICAL TRANSCEIVER TUNING USING PARALLEL PIPELINE MACHINE LEARNING ASSISTANCE,” and claims the benefit of U.S. Provisional Application No. 63 / 325,373, filed on March 30, 2022, entitled “OPTICAL TRANSCEIVER TUNING USING PARALLEL PIPELINE MACHINE LEARNING ASSISTANCE,” and U.S. Provisional Application No. 63 / 513,331, filed on July 12, 2023, entitled “OPTICAL TUNING TEST SYSTEM USING PARALLEL OVENPIPELINES WITH PARALLEL INSTRUMENT CHANNELS AND MACHINE LEARNING ASSISTANCE,” the disclosures of all of the said U.S. provisional applications being incorporated herein by reference in their entirety. Technical Field

[0003] The present disclosure relates to test and measurement systems, and more particularly to systems and methods for tuning parameters of a device under test (DUT), such as an optical transceiver device. Background Art

[0004] Machine learning techniques can improve test times for tuning parameters of a device under test (DUT). U.S. Patent No. 11,923,895 and U.S. Patent No. 11,923,896 (each issued on March 5, 2024, the entire contents of which are hereby incorporated by reference into this disclosure) both disclose systems and methods for tuning parameters of a DUT (such as an optical transceiver). Those systems and methods can improve test times for tuning optical transceivers or other DUTs, for example, in a manufacturing environment. For optimal tuning parameter set prediction and verification, those systems and methods can reduce optical transceiver tuning parameter test times from a worst-case example of two hours per DUT to approximately 12 seconds per DUT per temperature. This represents a very significant speedup compared to a conventional tuning process in the worst case.

[0005] However, the DUT tuning process may include tuning at different temperatures. The time required to raise or lower the DUT to each desired tuning temperature causes significant delays. For example, in some tuning processes, each temperature rise time is 180 seconds. In addition, loading the DUT into and out of the temperature chamber for testing may take additional time. Therefore, what is needed is an overall system design and method that can reduce the temperature cycle time in the oven and the time to switch out the device to almost zero time.

[0006] U.S. patent application Ser. No. 18 / 126,342, filed on March 24, 2023 (hereinafter “the '342 application,” the entire contents of which are hereby incorporated by reference into this disclosure) describes a test system using parallel pipelined ovens in which the transmitter output from each oven is processed serially in one oscilloscope channel at a time. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 An embodiment of a test system using pipelining is shown.

[0008] Figure 2 A pipeline diagram of an embodiment of a testing system is shown.

[0009] Figure 3 Shown is how the four pipelines are aligned in timing.

[0010] Figure 4 Timing alignment of an embodiment of four pipelines having four channels of a test and measurement instrument is shown.

[0011] Figure 5 An embodiment of a test system using optical tuning and machine learning is shown.

[0012] Figure 6 An example of a single-pole double-throw optical switch is shown.

[0013] Figure 7 An example of a dual single-pole double-throw optical switch is shown.

[0014] Figure 8 An embodiment of parallel pipeline processing using two ovens (32 emitters per oven) and four channels is shown.

[0015] Fig. 9 An embodiment of a parallel pipeline with four-channel parallel test and measurement acquisition is shown. DETAILED DESCRIPTION

[0016] Embodiments of the present disclosure, in conjunction with a machine learning (ML) system, address the problem of reducing temperature cycling time when testing a DUT to speed up the total time to test the DUT. Embodiments generally provide systems and methods where the total tuning time depends only on one oscilloscope channel and the amount of time spent on machine learning for each transmitter. These systems and methods then produce the ability to output a fully tuned transmitter tuned at multiple temperatures. In one embodiment, a system using three temperatures can run at an uninterrupted rate of 36 seconds per transmitter. This represents a two hundred-fold speedup of the two hour tuning time of a worst case example of a conventional tuning process.

[0017] Embodiments of the present disclosure use novel techniques for pipeline processing of data, use novel machine learning elements, and use novel techniques for serial serialization of different instruments (such as oscilloscope ("scope") channels for different temperature chambers or ovens. Embodiments of the present disclosure do not typically parallelize oscilloscope channel acquisitions between channels to obtain a 200x speedup factor. Instead, embodiments of the present disclosure typically process channels serially one channel at a time. This avoids the very expensive and time-consuming process of redesigning oscilloscope hardware and software to accommodate parallel channel processing because parallel channel processing is not required. Another advantage of the overall test configuration is that the layers of optical or other switches required in the signal path between the DUT and the oscilloscope are minimized to only one layer. This provides the advantages of reduced cost and improved signal integrity compared to attempting to simultaneously process four waveforms input in parallel into four oscilloscope channels as some current manufacturing systems do.

[0018] Embodiments employ a parallel pipeline architecture. The number of pipelines is equal to the number of ovens or temperature chambers used. The number of pipelines also corresponds to the number of switches connected to the DUTs in each oven to cycle through the DUTs, the dimension of the switches that select between the switches connected to each oven, and the number of oscilloscope or test instrument channels. Another dimension is the number of DUTs per oven. Each of the oven DUT switches (those switches connected to the DUTs in the oven) will have a number of switches equal to the number of DUTs in the oven.

[0019] The following discussion uses a specific number of ovens, DUT switches, instrument switches of specific dimensions, and a specific number of channels on the test and measurement instrument. These numbers are for ease of discussion and understanding and are in no way intended to limit the number of any component in the system.

[0020] Figure 1The embodiment shows a pipeline system 10 under the control of a customer test automation application 12 and a machine learning system 14. These can reside on the same computing device 16, which can have one or more processors, or can be distributed between multiple computing devices, each having one or more processors. The customer test automation application 12 can control the switches through a programming interface (PI). Similarly, the test automation application can also interact with the instrumentation and machine learning system through a programming interface.

[0021] An example embodiment uses four ovens 20, 22, 24, and 26. This translates into four pipelines, four DUT switches 30, 32, 34, and 36, and a 4x1 instrument switch 50 that connects each DUT switch to an oscilloscope clock CLK. The Tx transmitter must be tuned for three or more temperatures and takes 180 seconds to ramp up to each temperature. In order to take advantage of the parallel architecture, the embodiments of the present disclosure overlap the ramp up time for the temperature changes in each oven. The end result is that the oven ramp up time becomes almost 0s. Similarly, the time to put the transmitter in and out of the oven also becomes almost 0s. Each pipeline also contains serial operation tasks.

[0022] In an embodiment, each oven contains eight emitters to be heated. The DUTs in this embodiment include optical emitters, but any type of tunable DUT (optical or electronic) can use the system. "Electronic" DUTs are those that are not optical devices. There are eight emitters per oven, resulting in each DUT switch 30, 32, 34 and 36 being an 8x1 switch. In the case of optical emitters, these are optical switches.

[0023] Even if eight emitters in an oven are heated in parallel, the emitters should be serially tuned into one instrument channel, one at a time. This avoids expensive redesign of the instrument to allow parallel processing of all channels.

[0024] Each DUT switch can be connected to a splitter that splits the signal between the instrument switch 50 and the channel on the instrument 52. Splitters 40, 42, 44, and 46 pick up some of the signal from each of the four DUT switch outputs to apply to the 4x1 instrument switch output into the CLK input of the sampling oscilloscope. This is for the purpose of clock recovery for the oscilloscope used to acquire the signal on each channel.

[0025] The instrument switch 50 selects the appropriate 8x1 DUT switch output to tune the emitter in the oven that has finished ramping up to the correct temperature. The selected 8x1 DUT switch will cycle through all emitters in the oven.

[0026] In operation, the test and measurement instrument acquires waveforms from the transmitter. One of the aspects of the system is that only one channel is processed at a time in a serial manner. Channels are not processed in parallel. This is because only one channel will be raised to the correct channel at a time. For example, when the oven 20 reaches temperature, then channel 1 will acquire waveforms from each of the 8 transmitters in series in the oven at that temperature. The user test automation application 12 will pass these waveforms to the Tektronix optical tuning application 14 for training the deep learning network and for predicting the optimal tuning parameters, as well as for calculating the TDECQ verification of the resulting tuning. This means that the time to tune the transmitter depends only on the acquisition time of the oscilloscope and the DSP processing time of the optical tuning application. The Tektronix optical tuning algorithm or machine learning system uses machine learning to provide tuning parameters for the optical transceiver based on the waveforms received from the DUT.

[0027] The client application 12 acts as the master controller for the entire system. It is responsible for the timing and sequencing of all tasks for all four parallel pipelines. These tasks include setting the temperature of the oven. The controller can pause when an operator or robot loads or unloads a transmitter into or from the oven, and can control the robot (if a robot is used). Tasks also include loading tuning parameters into the transmitter and controlling the oscilloscope to acquire waveforms from the transmitter. The controller also collects waveforms and parameters and sends them to the optical tuning application to train the deep learning network, receive reverse predictions of the optimal tuning parameters, and receive measurements on the waveform to verify the tuning. The control of the system is embodied in one or more processors configured to execute code to operate various aspects of the system. One or more processors can be located on a test automation application 12 running on a computing device separate from the instrument and machine learning system. Each of those can have its own processor, they can all be included in one system, or any mixture in between.

[0028] The controller also controls the 8x1 DUT switch to select which transmitter in the oven to collect the waveform from, and controls the 4x1 instrument switch to select the correct output of the 8x1 switch to apply to the clock recovery input CLK of the oscilloscope. The controller also controls the instrument to acquire the waveform from the correct channel of the instrument, depending on which channel has the oven currently raised to temperature for the tuning operation. Figure 2 It will be seen that there is an unused period of time between the increase in temperature of each oven and the tuning interval of the transmitters in that oven at that temperature, in this example 96 seconds.

[0029] As mentioned above, the tuning process employs a novel machine learning system that is trained to associate a set of tuning parameters for each DUT based on the waveform. The process can be iterated until the DUT has been determined to have passed or failed. The machine learning system accelerates the cycle of tuning the DUTs and then testing them to determine if the operation has passed or failed. The testing process involves a measurement process that also relies on the machine learning system. The system has demonstrated that it reduces the tuning interval for each emitter at each temperature to 12 seconds. With the ramp period for each oven reduced essentially to zero, the system can achieve a tuning interval of 36 seconds per emitter across three temperatures.

[0030] Figure 2 One of the pipelines is shown, and Figure 3 Four parallel pipelines are shown, each sourced by an oven. Figure 2 , oven 1 with a tuned set of emitters receives a new set of emitters at 60. At 62, the oven is raised to a first temperature, which takes 180 seconds. The emitters are then put through a tuning and test using a machine learning system at 64, which takes 12 seconds for each emitter, or 96 seconds for a set of 8 emitters to be tested at that temperature. Upon completion of the tuning and testing, the oven is then raised to a second temperature. There is an unused time interval of 96 seconds at 66. The process then continues with tuning and testing, raising to a third temperature, another unused time interval of 96 seconds, and then a final tuning and test at 68. Figure 2 Only one oven and one line view is shown. Figure 3 As shown in , the efficiency and advantages of the pipelined approach come from combining multiple pipelines.

[0031] exist Figure 3 In the figure, you can see four ovens feeding parallel pipelines. The top pipeline is Figure 2 The pipeline shown in . More pipelines can be added as more ovens are added. The process loads the second oven when the loading of the first oven is complete. As mentioned above, the tuning time for each transmitter takes 12 seconds. This breaks down into 2 seconds for each acquisition of the four waveforms, 1 second for the machine learning DSP to predict the tuning parameters, 2 seconds to calculate the TDECQ (Transmitter Dispersion Eye Closure Quadruple) measurement used to determine pass / fail, and 1 second margin for a total of 12 seconds per transmitter. Each test and tuning block performs calculations while the other three ovens are in different stages of heating up.

[0032] After each test and tune block, there is another block for that pipeline to ramp up to the next temperature. After each pipeline has calculated the tune for each temperature, the 8 emitters in that oven are ready to be unloaded and a new set of 8 emitters are loaded to repeat the cycle. The unused time interval allows the system to perform other operations that may not be covered by the optical tuning algorithm. The pipeline serialization time can then be adjusted if needed.

[0033] like Figure 3 The parallel pipeline shown in provides a system that outputs fully tuned transmitters at an average ideal rate of one transmitter every 36 seconds. Assuming the machine learning predictions are 100% accurate, the given tuning times represent ideal values. The 96 second overlap margin can cover some or all of any inaccuracies. Even with these built-in intervals, actual results are expected to be many times faster than current methods used on production lines.

[0034] As discussed above, embodiments may have been discussed with respect to tuned optical transmitters, but the pipelined architecture and temperature testing may be applied to many different types of DUTs (optical or electronic).

[0035] In this way, embodiments pair the mentioned optical tuning machine learning system with a novel parallel pipeline and instrument channel switching architecture to drive the oven cycle time to almost zero. This results in a speedup of machine learning-assisted predictive optimization of tuning parameters. Embodiments achieve this tuning speed improvement of almost two hundred times without the need to use parallel instrument acquisition channels.

[0036] [New content here]

[0037] In other embodiments, multiple channels of the test and measurement instrument can acquire waveforms from the DUT in parallel. For ease of discussion and understanding, the following discussion focuses on the specific dimensions of the DUT, oven, and switch. More general embodiments follow the initial discussion.

[0038] Embodiments route the outputs from four emitters in each oven into four channels of the instrument at a time for parallel instrument acquisition. This requires an additional layer of optical switches. These embodiments rely on multiple acquired channels with 2 to 3 seconds of acquisition time, which becomes the largest portion of the time used for tuning and TDECQ and other measurement value calculations with the help of machine learning. A test system using serial oscilloscope channels in accordance with embodiments of the '342 application achieves a nominal speedup factor of 75 times for machine learning tuning parameter predictions compared to a customer's typical tuning time. The embodiments here represent a 320 times speedup over a customer's original 2 hour tuning time for a worst case scenario.

[0039] According to some embodiments of the present disclosure, Figure 4 illustrates an example serialization of test operations, and Figure 5 A block diagram of an example test system is illustrated. The system is novel in the sense of how it pipelines the processing and includes a novel machine learning element for predicting the DUT transmitter tuning parameters. It is also novel in how it adds a second layer of switches, which enables acquisition of four channels at a time on an oscilloscope to significantly speed up the tuning process.

[0040] refer to Figure 4 and 5 , various embodiments of the present disclosure may include a combination of one or more of the following aspects. Figure 5 The parallel pipeline architecture shown in has four ovens; each oven includes one pipeline, so four ovens equal four pipelines. In this system, the DUT (which may include an optical transmitter) undergoes tuning at three or more temperatures. Fig. 9 As discussed in more detail, the temperature increase affects the timing of each pipeline. Figure 4 and 5 In the embodiment of the present disclosure, each oven takes 180 seconds to ramp up to a given temperature. Therefore, the embodiment of the present disclosure overlaps the ramp time of the temperature changes in each oven. This results in the oven ramp time having almost 0 seconds. Similarly, the time to put the transmitter in and out of the oven is also almost 0 seconds. Each pipeline also contains a DSP task that operates serially in the ovens within the temperature range.

[0041] exist Figure 4 and 5 In the embodiment of FIG. 1 , each oven 20, 22, 24, and 26 contains eight emitters to be heated, which represents another parallel aspect of the architecture. As mentioned above, each oven represents the source of one of the serial pipelines, so Figure 4 and 5 The embodiment of has four serial pipelines in the architecture. Each oven cycles through three temperature settings. It should be noted that embodiments of the present disclosure can be configured for more or fewer ovens, and more or fewer emitters per oven, and configured for one or more temperatures. The three temperatures are the nominal requirements of the most likely customer on the production line. In addition, the term "oven" should be understood to generally mean a controlled temperature chamber. Therefore, embodiments of the present disclosure can also control the oven temperature (one or more) to be below the ambient temperature to cool the DUT (if necessary).

[0042] exist Figure 5 In the particular embodiment shown in FIG. 1 , switches 60, 62, 64, and 66 each include four SPDT optical switches or two DPDT optical switches (per oven). Discussions herein refer to these as "oven switches." They include a first layer of optical switches that select four of the eight emitters in the oven that have reached temperature. These four waveforms are then input to a second layer of optical switches.

[0043] The second layer of optical switches 70, 72, 74, and 76 comprises a second layer of single-pole, four-throw optical switches. There are four of these 4x1 switches in the second layer. Each switch selects one of the four waveforms from the oven that is currently at temperature and routes them to the four channels of the oscilloscope. One oscilloscope acquisition will acquire four waveforms simultaneously. This discussion refers to these switches as "channel" switches because each channel switch is connected to one channel of the test and measurement instrument 52.

[0044] Some embodiments may use optical splitters, such as Figure 1 As in the embodiment of . The splitter 40 picks up some of the signal from one of the four optical switch outputs to be applied to the CLK recovery input of the sampling oscilloscope. This allows clock recovery for the oscilloscope to acquire the signal on each channel. Figure 5 In an embodiment of the present invention, the oscilloscope may be an equivalent time (ET) oscilloscope (oscilloscope), which is an example of a sampling oscilloscope, as a test and measurement instrument. An embodiment using an ET oscilloscope would use splitter 40. An embodiment using an RT oscilloscope would not have splitter 40 and would not have a connection CLK on instrument 52.

[0045] Embodiments that require an optical splitter to branch off a portion of the signal for input to the clock recovery input of the oscilloscope may require further processing. Sampling oscilloscopes may include mathematical scaling factors that need to be applied to the input waveform to make it compatible with the waveforms on the other three channels of the oscilloscope. For embodiments using an RT oscilloscope, the scaling block and splitter are not required.

[0046] exist Figure 5 In the embodiment of FIG. 1 , the instrument acquires waveforms from the transmitter. One of the key aspects of this system is that four channels are acquired at a time in parallel. The time used to tune the transmitter is limited only by the acquisition time of the oscilloscope, and the DSP processing time of the optical tuning application.

[0047] like Figure 1 As in the case of parallel channel acquisition Figure 4 The embodiment includes a customer test automation application 12 as the main controller for the entire system. It is responsible for the timing and sequencing of all tasks for all four parallel pipelines. This includes setting the temperature of the oven. This includes pausing while an operator or robot loads and unloads transmitters into and from the oven, loading tuning parameters into the transmitter, and controlling the instrumentation to acquire waveforms from the transmitter.

[0048] This also includes collecting waveforms and parameters and sending them to the optical tuning application to train the deep learning network and receive back predictions of the optimal tuning parameters and receive measurements on the waveforms to verify the tuning. The control also includes controlling a first layer of 2x1 optical switches to select (one or more) emitters and collect four waveforms from the ovens currently reaching temperature. The client application 12 controls a second layer of switches that selects the oven pipeline that is heating up and controls the oscilloscope to acquire waveforms from the oscilloscope while keeping track of which oven is currently heating up to temperature for the tuning operation. The timing includes an unused time interval of 96 seconds that can be allocated to additional unspecified operations.

[0049] like Figure 1 As in FIG. 1 , block 14 is a machine learning system 14. This block uses a deep learning network to predict optimal tuning parameters and accelerate the measurement. The client application and the machine learning system may reside on a host PC 16, which may have a graphics processing unit (GPU). According to some embodiments, in order to obtain the desired acceleration time, the host PC may have a clock speed of 5 GHz and use a GPU.

[0050] exist Figure 4In the pipeline task structure, how to execute tasks is specified, with some tasks executed in parallel and some tasks executed in series. This shows four parallel pipelines sourced by four ovens. The timing block shows a 96-second time interval for loading and unloading the ovens with emitters. The nature of the parallel pipeline organization makes this time block almost 0 seconds. The larger blocks represent the heating time of the ovens when temperature changes are required. These are up to 180 seconds in time. These are also almost zero seconds, where the tuning time depends only on the time length of the CHX 60s box. The blocks marked with CHX (where X is 1, 2, 3 or 4 in this embodiment) represent the time used to tune 8 emitters at a given oven temperature. For each emitter, the time to collect four waveforms in parallel each time is 2 to 3 seconds, the time for DSP to predict tuning parameters is 1 second, the time to calculate the classic TDECQ is 2 seconds, and the margin is 1 second. Refer to the Mathcad model equation below, where greenBlock is the amount of time used for calculation and acquisition. Each CHX 60s block is calculated when the other three ovens are in different stages of heating up. After each CHX60s block, there is another TEMP block for that pipeline to ramp up to the next temperature for testing. After each pipeline has calculated the tune for each temperature, the 8 emitters in that oven are ready to be unloaded, and a new set of 8 emitters are loaded to repeat the cycle. The other blocks are spare, unused time intervals that allow the system to perform other operations that may not be covered by the optical tuning algorithm. If desired, the pipeline serialization time can be adjusted to split the gray blocks to include other operations at each temperature after each CHX 60s block. These blocks take almost 0 seconds to execute.

[0051] The Mathcad model equation is as follows:

[0052] refPar:=3

[0053] scopeAcq:=2 seconds

[0054] TDECQ: = 2 seconds

[0055] Tune: = 1 second

[0056] numTemp:=3

[0057] numTXperOven:=8

[0058] nominalTune:=15·60 seconds

[0059] numChannels:=4

[0060] The timing estimate for a parallel oven pipeline with parallel oscilloscope channels is:

[0061] For 1 DUT / transmitter:

[0062]

[0063] greenBox=60 seconds

[0064]

[0065] timeParallel = 7.5 seconds

[0066]

[0067] speedUp=120 seconds.

[0068] like Figure 4 The end result of the parallel pipeline shown in is to output fully tuned transmitters at an average ideal rate of one transmitter every 22.5 seconds. This is for tuning at three temperatures. Assuming the machine learning predictions are 100% accurate, the given tuning times are ideal. But there is still a 96 second overlap margin to cover some or all of the inaccuracy. However, the expected actual rate of tuning transmitters may be slower because the transmitters that fail must be tested with the user's slower tuning algorithm. The actual results are still expected to be many times faster than the current methods used on the production line.

[0069] As an example, and for ease of discussion and understanding, refer to Figure 4 and Figure 5 .for Figure 5 For the oven 20 pipeline in , the example switching sequence is as follows:

[0070] 1. Load oven 1, 20 with 8 emitters.

[0071] 2. Start heating oven 20 to the first temperature and wait 180 seconds. While oven 20 is heating, load ovens 2, 22 with 8 emitters and start heating oven 22 to the first temperature for 180 seconds. While oven 22 is heating, load ovens 3, 24 with 8 emitters and start heating oven 24 to the first temperature for 180 seconds. While oven 24 is heating, load ovens 4, 26 with 8 emitters and start heating oven 26 to the first temperature for 180 seconds.

[0072] 3. Set the first level switch to select the four TXs in oven 20 using switch 60.

[0073] 4. Set the second level switches to apply the four Oven 1 transmitters to the four oscilloscope channels. Figure 5 , switches 70, 72, 74 and 76 will receive waveforms from transmitters 1, 3, 5 and 7 from oven 20, respectively.

[0074] 5. Put tuning reference parameter set 1 into TX.

[0075] 6. Acquire four TX waveforms from each of the first four selected transmitters.

[0076] 7. Set the first level switch, switch 60, to select the second four TXs in oven 20.

[0077] 8. Collect four TX waveforms from each of the second set of four selected transmitters in oven 1. Each of switches 70, 72, 74 and 76 will receive waveforms from transmitters 2, 4, 6 and 8 from oven 20, respectively.

[0078] 9. Place reference parameter set 2 into TX and acquire the second four waveforms.

[0079] 10. Change the first layer switches to select the other four transmitters. In this case, the other subset will include transmitters 1, 3, 5, and 7.

[0080] 11. Acquire four additional waveforms in parallel on four oscilloscope channels with one acquisition cycle.

[0081] 12. Place the third set of reference tuning parameters into the transmitter.

[0082] 13. Collect four TX waveforms with one oscilloscope acquisition.

[0083] 14. Change the first layer switches to select the other four TXs in oven 1. Now the transmitters switch to the second subset 2, 4, 6, and 8.

[0084] 15. Acquire four waveforms with one oscilloscope acquisition.

[0085] 16. At Temperature 1, the three waveforms from each transmitter are sent sequentially into the OptaML Tuning Prediction block and a set of predicted optimal tuning parameters are received back for each of the 8 TXs in Oven 1.

[0086] 17. Place the 8 best tuning parameter sets into the associated transmitters.

[0087] 18. Collect four waveforms from the first four transmitters.

[0088] 19. Change the first layer switch settings to select the second four emitters.

[0089] 20. Perform an oscilloscope acquisition to acquire four waveforms.

[0090] 21. Calculate TDECQ and other measurements on the 8 acquired waveforms to determine if each transmitter is optimally tuned.

[0091] 22. Change oven 1 to start increasing to temperature two.

[0092] 23. At this point, oven 2 has reached temperature 1 for its 180s soak time.

[0093] 24. Toggle the second level switch to select the emitter from oven 2

[0094] 25. Repeat steps 3 to 23, but use oven 2 at temperature 1.

[0095] 26. Toggle the second level switch to select the emitter from oven 3

[0096] 27. Repeat steps 3 to 23, but use oven 3 at temperature 1.

[0097] 28. Toggle the second level switch to select the emitter from oven 4

[0098] 29. Repeat steps 3 to 23, but use oven 4 at temperature 1.

[0099] 30. Toggle the second level switch to select the emitter from oven 1

[0100] 31. Repeat steps 3 to 23, but using oven 1 at temperature 2.

[0101] 32. Toggle the second level switch to select the emitter from oven 2

[0102] 33. Repeat steps 3 to 23, but use Oven 2 at Temperature 2.

[0103] 34. Toggle the second level switch to select the emitter from oven 3.

[0104] 35. Repeat steps 3 to 23, but use oven 3 at temperature 2.

[0105] 36. Toggle the second level switch to select the emitter from oven 4

[0106] 37. Repeat steps 3 to 23, but use oven 4 at temperature 2.

[0107] 38. Toggle the second level switch to select the emitter from oven 1

[0108] 39. Repeat steps 3 to 23, but using oven 1 at temperature 3.

[0109] 40. Unload the 8 tuned transmitters from oven 1 and load 8 new transmitters for tuning.

[0110] 41. Repeat steps 3 to 23, but use oven 2 at temperature 3.

[0111] 42. Output 8 tuned transmitters from oven 2 at all three temperatures.

[0112] 43. Unload 8 transmitters from oven 3 and load 8 new transmitters for tuning.

[0113] 44. Output 8 tuned transmitters from oven 3 at all three temperatures.

[0114] 45. Repeat steps 3 to 23, but use oven 4 at temperature 3.

[0115] 46. ​​Export the 8 tuned transmitters from oven 4 and load 8 new transmitters for tuning.

[0116] 47. Keep repeating the above cycle to continuously output a tuned transmitter with almost 0 time for loading and unloading and warming up. The time depends only on the acquisition time of the oscilloscope and the DSP time used to predict the tuning parameters and verify with TDECQ.

[0117] Figure 6 and 7 A closer view of the first and second layers of optical switches is shown. Figure 6 An embodiment of half of switch 60 is shown, but may also apply to switches 62, 64, and 66. These types of switches are commercially available, such as through Luxlink, part number OS-3141. In the first state, this half of the switch selects DUTs 1 and 3 from the first oven, so 1-1 and 1-3. In the second state, the switch will select DUTs 1-2 and 1-4. The second half of the switch will have the same configuration, but will select 1-5 and 1-7, or 1-6 and 1-8.

[0118] Figure 7 Shows Figure 5 70, 72, 74, and 76. The current configuration is for switch 70 connected to instrument channel 1. The first setup is for the first oven. Referring back to the sequence set forth above, after all three sets of waveforms have been acquired from all eight DUTs in the two 4-DUT subsets, the switch will then acquire waveforms from oven 2, and so on.

[0119] While the above discussion has focused on the number of DUTs per oven, the number of ovens, and two-tier switches, it is not intended to be limited to those specific dimensions, nor should any dimensions be assumed. Fig. 9As shown in , generally, embodiments of the present disclosure may include a universal optical switch matrix that can connect any number of ovens (with any number of transmitters in each oven) to four channels of an oscilloscope for parallel acquisition. It is this basic test configuration combined with a novel machine learning-assisted tuning prediction block for optical transmitters that constitutes the novel invention.

[0120] As examples of other dimensions of this test system, Figure 8 A parallel oven pipeline using only two ovens is shown. It requires more emitters per oven, 32, making the tuning process time shown in the CHX 60s block longer than the time it takes for the ovens to heat up. These are the longer blocks. The "swap" blocks represent the time spent unloading the ovens and loading the ovens with emitters for the next round of testing. As long as the prediction time in the CHX 60s block is longer than the temperature ramp time in the TEMP block plus the load / unload blocks in the swap block, then the overall tuning time depends only on the oscilloscope acquisition time plus the DSP prediction time plus the overhead associated with that. This means that the temperature ramp time and the oven loading and unloading times become almost zero and have no effect on the speed of the overall tuning.

[0121] Fig. 9 A more general block diagram is shown, which shows that the number of ovens 80, 82 and 84 can be any number and accommodate different numbers of transmitters. As long as there are enough transmitters in each oven so that the temperature rise time is shorter than the time it takes to process the transmitters in the oven, then it can operate at the maximum speed, which depends only on the CH processing block used for prediction to determine the speed of tuning. The switch layer can be changed as needed and form an optical switch matrix 90. As discussed above, this embodiment shows an ET or other sampling instrument that requires the use of a splitter 40 for clock recovery. Different instruments (such as RT oscilloscopes) will not need to do this.

[0122] These embodiments use an additional layer of optical switches to allow the use of parallel oscilloscope acquisition in pipeline processing. The optical tuning machine learning assistance block combined with the novel parallel oven pipeline and parallel oscilloscope channel acquisition and switching architecture results in the oven cycle time becoming almost zero and results in machine learning assisted acceleration to predict optimized tuning parameters. This combination of features results in an ideal 320x tuning speed improvement for the worst case customer taking 2 hours to tune.

[0123] Aspects of the present disclosure may be operated on specially created hardware, on firmware, digital signal processors, or on a general-purpose computer including a processor that operates according to programming instructions. The term controller or processor as used herein is intended to include a microprocessor, a microcomputer, an application-specific integrated circuit (ASIC), and a dedicated hardware controller. One or more aspects of the present disclosure may be implemented with computer-usable data and computer-executable instructions (such as one or more program modules) executed by one or more computers (including monitoring modules) or other devices. Typically, a program module includes routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types when executed by a processor in a computer or other device. Computer-executable instructions may be stored on a non-temporary computer-readable medium, such as a hard disk, an optical disk, a removable storage medium, a solid-state memory, a random access memory (RAM), etc. As will be understood by those skilled in the art, the functions of the program modules may be combined or distributed as needed in various aspects. In addition, the functions may be implemented in whole or in part with firmware or hardware equivalents (such as integrated circuits, FPGAs, and the like). Certain data structures may be used to more efficiently implement one or more aspects of the present disclosure, and such data structures are considered to be within the scope of the computer-executable instructions and computer-usable data described herein.

[0124] In some cases, the disclosed aspects may be implemented in hardware, firmware, software, or any combination thereof. The disclosed aspects may also be implemented as instructions carried or stored on one or more or non-transitory computer-readable media, which may be read and executed by one or more processors. Such instructions may be referred to as computer program products. Computer-readable media as discussed herein means any medium that can be accessed by a computing device. As an example and not limitation, computer-readable media may include computer storage media and communication media.

[0125] Computer storage media means any medium that can be used to store computer-readable information. By way of example and not limitation, computer storage media may include RAM, ROM, electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital video disk (DVD) or other optical disk storage, cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, and any other volatile or non-volatile, removable or non-removable media implemented in any technology. Computer storage media excludes the signal itself and the transient form of signal transmission.

[0126] Communication media refers to any medium that can be used for communication of computer-readable information. By way of example and not limitation, communication media may include coaxial cables, fiber optic cables, air, or any other medium suitable for communication of electrical, optical, radio frequency (RF), infrared, acoustic, or other types of signals.

[0127] Example

[0128] Illustrative examples of the disclosed technology are provided below. Embodiments of the technology may include one or more of the examples described below and any combination thereof.

[0129] Example 1 is a test system comprising: a test and measurement instrument having a plurality of channels; an oven configured to accommodate a plurality of devices under test (DUTs); oven switches, each oven switch selectively connected to a DUT in a dedicated oven; channel switches selectively connected to the oven switches, each channel switch connected to a channel of the test and measurement instrument; one or more processors configured to execute code, the code causing the one or more processors to: select an oven as a selected oven and select a corresponding oven switch as a selected oven switch; control the selected oven switch to connect to a subset of the DUTs in the selected oven; control the channel switches to connect to the selected oven switches to receive signals from a subset of DUTs from a selected oven and send the signals to channels of a test and measurement instrument to acquire waveforms from the subset of DUTs in parallel; repeatedly controlling the channel switches and the selected oven switches until a waveform from each DUT in the selected oven has been acquired; using machine learning to tune each DUT in the selected oven to a set of parameters; acquiring waveforms from the DUTs in the selected oven and calculating measurements on the waveforms to determine whether each DUT in the selected oven is optimally tuned; and repeatedly selecting a next oven, a corresponding next oven switch, and controlling the channel switches until each DUT in each oven has been tuned and tested.

[0130] Example 2 is the test system of Example 1, wherein the number of channel switches corresponds to the number of channels on the test and measurement instrument.

[0131] Example 3 is the test system of any of Examples 1 and 2, wherein the oven switches include a plurality of switches based on the number of DUTs in each oven.

[0132] Example 4 is the test system of any of Examples 1 to 3, wherein the one or more processors are further configured to control the oven to cycle the oven through a plurality of temperatures, and the one or more processors repeatedly control the selected oven switches and channel switches for each temperature.

[0133] Example 5 is the test system of any of Examples 1 to 3, wherein the DUT includes one of an electronic device, an optical transceiver, or an optical transmitter.

[0134] Example 6 is the test system of any one of Examples 1 to 3, wherein the code causing the one or more processors to tune each DUT using machine learning includes code causing the one or more processors to: sequentially transmit acquired waveforms from each DUT in a selected oven to a machine learning network; and use the machine learning network to analyze the waveforms and provide tuning parameters for each DUT.

[0135] Example 7 is the test system of Example 6, wherein the one or more processors are further configured to control the oven to cycle through multiple temperatures, and the one or more processors repeat the serialized transmission of the acquired waveforms and the use of the machine learning network for each temperature for each DUT in each oven.

[0136] Example 8 is the test system of any one of Examples 1 to 3, wherein: there are four ovens, and each oven accommodates eight light emitters; there are four oven switches, each oven switch is configured to select a subset of the four light emitters; and there are four channel switches, each channel switch is connected between the four oven switches and a channel of the test and measurement instrument.

[0137] Example 9 is the test system of any of Examples 1 to 3, wherein the test and measurement instrument includes an oscilloscope.

[0138] Example 10 is a method for testing a device under test (DUT) in multiple ovens, the method comprising: selecting an oven from multiple ovens as a selected oven and selecting a corresponding oven switch as a selected oven switch; controlling the selected oven switch to be connected to a subset of DUTs in the selected oven; controlling a channel switch connected to the selected oven switch to receive signals from the subset of DUTs from the selected oven and send the signals to channels of a test and measurement instrument to acquire waveforms from the subset of DUTs in parallel; repeatedly controlling the channel switches and the selected oven switches until a waveform from each DUT in the selected oven has been acquired; using machine learning to tune each DUT in the selected oven to a set of parameters; acquiring waveforms from the DUTs in the selected oven and calculating measurements on the waveforms to determine whether each DUT in the selected oven is optimally tuned; and repeatedly selecting the next oven, the corresponding next oven switch, and controlling the channel switches until each DUT in each of the multiple ovens has been tuned and tested.

[0139] Example 11 is the method of Example 10, wherein the number of channel switches corresponds to the number of channels on the test and measurement instrument.

[0140] Example 12 is the method of any of Examples 10 or 11, wherein the oven switches include a plurality of switches based on the number of DUTs in each oven.

[0141] Example 13 is the method of any of Examples 10 to 12, further comprising controlling the oven to cycle the oven through a plurality of temperatures, and repeatedly controlling the selected oven switch and channel switch for each temperature.

[0142] Example 14 is the method of any of Examples 10 to 13, wherein the DUT comprises one of an electronic device, an optical transceiver, or an optical transmitter.

[0143] Example 15 is the method of any of Examples 10 to 14, wherein using machine learning to tune each DUT comprises: serializing acquired waveforms from each DUT in a selected oven to a machine learning network; and using the machine learning network to analyze the waveforms and provide tuning parameters for each DUT.

[0144] Example 16 is the method of Example 15, further comprising controlling the oven to cycle through a plurality of temperatures, and repeating the serialized transmission of the acquired waveforms and using the machine learning network for each temperature for each DUT in each oven.

[0145] Example 17 is the method of any of Examples 10 to 16, wherein using the machine learning system includes: receiving a waveform from each DUT; and applying machine learning to the waveform to generate tuning parameters for the DUT.

[0146] Example 18 is the method of any of Examples 10 to 17, wherein the test and measurement instrument comprises an oscilloscope.

[0147] Example 19 is a test system comprising: a test and measurement instrument having multiple channels; an oven configured to accommodate multiple devices under test (DUTs), the number of DUTs in each oven being high enough to allow a temperature rise time that is less than the time to process all DUTs in the oven; a switch matrix configured to connect the DUTs in each oven to the channels of the test and measurement instrument; one or more processors configured to execute code that causes the one or more processors to: control the switch matrix to connect to at least some of the DUTs in an initial group of ovens set to a first temperature to acquire a set of waveforms using the test and measurement instrument; set a next group of ovens to the initial temperature; transmit the set of waveforms to a machine learning network; receive optimal tuning parameters for the DUTs at the initial temperature; test the DUTs using the optimal tuning parameters to determine whether the DUTs are optimally tuned; set the initial group of ovens to the next temperature; and repeat the control, transmission, reception, and testing of the next group of ovens at the next temperature until all DUTs in all ovens have been tested at all temperatures in a predefined set of test temperatures.

[0148] Example 20 is the test system of Example 19, wherein the test and measurement instrument includes an oscilloscope.

[0149] In addition, this written description refers to specific features. It is to be understood that the disclosure in this specification includes all possible combinations of those specific features. For example, where a specific feature is disclosed in the context of a particular aspect, the feature may also be used in the context of other aspects to the extent possible.

[0150] Furthermore, when reference is made in the present application to a method having two or more defined steps or operations, the defined steps or operations may be performed in any order or simultaneously, unless the context excludes those possibilities.

[0151] All features disclosed in the specification including claims, abstract and drawings, and all steps in any method or process disclosed, may be combined in any combination except combinations in which at least some of such features and / or steps are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in the specification including claims, abstract and drawings may be replaced by an alternative feature serving the same, equivalent or similar purpose.

[0152] Although particular aspects of the present disclosure have been illustrated and described for purposes of illustration, it will be appreciated that various modifications may be made without departing from the spirit and scope of the invention. Accordingly, the invention should not be limited, except as by the appended claims.

Claims

1. A test system comprising: Test and measurement instruments with multiple channels; An oven configured to accommodate multiple devices under test (DUTs); Oven switches, each of which is selectively connected to a DUT in a dedicated oven; a channel switch selectively connected to the oven switch, each channel switch connected to a channel of the test and measurement instrument; One or more processors configured to execute code that causes the one or more processors to: Select an oven as a selected oven and select a corresponding oven switch as a selected oven switch; controlling the selected oven switches to connect to a subset of the DUTs in the selected oven; controlling a channel switch connected to a selected oven switch to receive signals from the subset of DUTs from the selected oven and to send the signals to channels of a test and measurement instrument to acquire waveforms from the subset of DUTs in parallel; Repeating the control of the channel switch and the selected oven switch until the waveform from each DUT in the selected oven has been acquired; Use machine learning to tune each DUT in a selected oven to a set of parameters; acquiring waveforms from the DUTs in the selected ovens and calculating measurements on the waveforms to determine whether each DUT in the selected oven is optimally tuned; as well as The selection of the next oven, the corresponding next oven switch, and the control channel switch are repeated until each DUT in each oven has been tuned and tested.

2. The test system of claim 1, wherein the number of channel switches corresponds to the number of channels on the test and measurement instrument.

3. The test system of claim 1, wherein the oven switch comprises a plurality of switches based on the number of DUTs in each oven.

4. The test system of claim 1, wherein the one or more processors are further configured to control the oven to cycle the oven through a plurality of temperatures, and the one or more processors repeatedly control selected oven switches and channel switches for each temperature.

5. The test system as recited in claim 1, wherein the DUT comprises one of an electronic device, an optical transceiver, or an optical transmitter.

6. The test system of claim 1 , wherein the code causing the one or more processors to tune each DUT using machine learning comprises code causing the one or more processors to: sequentially transmitting the acquired waveforms from each DUT in the selected oven to the machine learning network; and A machine learning network is used to analyze the waveform and provide tuning parameters for each DUT.

7. The test system of claim 6, wherein the one or more processors are further configured to control the ovens to cycle through a plurality of temperatures, and the one or more processors repeat the serialized transmission of the acquired waveforms and the use of the machine learning network for each temperature for each DUT in each oven.

8. The test system as claimed in claim 1, wherein: There are four ovens, and each oven houses eight light emitters; There are four oven switches, each oven switch being configured to select a subset of the four light emitters; as well as There are four channel switches, each of which is connected between the four oven switches and one channel of the test and measurement instrument.

9. The test system as recited in claim 1, wherein the test and measurement instrument comprises an oscilloscope.

10. A method for testing a device under test (DUT) in a plurality of ovens, the method comprising: selecting an oven from among the plurality of ovens as a selected oven, and selecting a corresponding oven switch as a selected oven switch; controlling the selected oven switches to connect to a subset of the DUTs in the selected oven; controlling a channel switch connected to a selected oven switch to receive signals from the subset of DUTs from the selected oven and to send the signals to channels of a test and measurement instrument to acquire waveforms from the subset of DUTs in parallel; Repeating the control of the channel switch and the selected oven switch until the waveform from each DUT in the selected oven has been acquired; Use machine learning to tune each DUT in a selected oven to a set of parameters; acquiring waveforms from the DUTs in the selected ovens and calculating measurements on the waveforms to determine whether each DUT in the selected oven is optimally tuned; as well as Selecting the next oven, the corresponding next oven switch, and controlling the channel switch are repeated until each DUT in each of the plurality of ovens has been tuned and tested.

11. The method of claim 10, wherein the number of channel switches corresponds to the number of channels on the test and measurement instrument.

12. The method of claim 10, wherein the oven switches include a plurality of switches based on the number of DUTs in each oven.

13. The method of claim 10, further comprising controlling the oven to cycle the oven through a plurality of temperatures and repeatedly controlling the selected oven switches and channel switches for each temperature.

14. The method of claim 10, wherein the DUT comprises one of an electronic device, an optical transceiver, or an optical transmitter.

15. The method of claim 10, wherein using machine learning to tune each DUT comprises: Serializing the acquired waveforms from each DUT in the selected oven to the machine learning network; as well as A machine learning network is used to analyze the waveform and provide tuning parameters for each DUT.

16. The test control system of claim 15, further comprising controlling the ovens to cycle through a plurality of temperatures, and repeating the serialized transmission of the acquired waveforms and the use of the machine learning network for each temperature for each DUT in each oven.

17. The method of claim 10, wherein using a machine learning system comprises: Receive waveforms from each DUT; as well as Apply machine learning to the waveform to generate tuning parameters specific to the DUT.

18. The method of claim 10, wherein the test and measurement instrument comprises an oscilloscope.

19. A testing system comprising: Test and measurement instruments with multiple channels; An oven configured to accommodate a plurality of devices under test (DUTs), the number of DUTs in each oven being high enough to allow a temperature ramp time that is less than the time to process all of the DUTs in the oven; a switch matrix configured to connect between the DUTs in each oven and channels of the test and measurement instrument; One or more processors configured to execute code that causes the one or more processors to: controlling the switch matrix to be connected to at least some of the DUTs within an initial set of ovens set to a first temperature to acquire a set of waveforms using a test and measurement instrument; Set the next set of ovens to initial temperature; transmitting the set of waveforms to a machine learning network; receiving optimal tuning parameters for the DUT at the initial temperature; Testing the DUT using the optimal tuning parameters to determine whether the DUT is optimally tuned; Set the initial group oven to the next temperature; as well as The control, transmission, reception and testing are repeated for the next set of ovens at the next temperature until all DUTs in all ovens have been tested at all temperatures in the predefined set of test temperatures.

20. The test system of claim 19, wherein the test and measurement instrument comprises an oscilloscope.

Images