Analysis

Version 6.1

Introduction

The Analysis tool enables expert users to analyze the performance of NDI network connections by providing key statistics that would otherwise be difficult to recover separately. It offers detailed information about the sources available on a network and their connection types, making it a valuable tool for debugging systems and identifying the root causes of potential issues.

Analysis is currently not available on the Tools Launcher. You can download it here from our main website

"There are three kinds of lies: lies, damned lies, and statistics." (Mark Twain, Benjamin Disraeli, et al)

In general, if you don’t actually see a problem, you don’t need this tool.

Analysis is capable of identifying deviations that are, for all practical purposes, meaningless. Much ado about (very, very slightly more than) nothing is effectively time and effort wasted. The Analysis tool is provided to help you better diagnose and understand network issues. While it produces a deluge of detail about network connections and NDI streams, the human eye is by far the best arbiter of video performance and quality.

Ultimately your eye provides a superior real-time analysis, and one which is inherently more tolerant. In certain cases when subjective issues are visible, lower-level analysis of the root problem may be required to understand the underlying factors. While we would love to help every user with every problem, the sheer numbers of those populating ‘the NDI nation’ make this quite impossible. We do, however, offer professional services to those who need assistance in diagnosing issues or help with large scale, complex NDI networking issues.

You can always reach out to support here or visit our community forums here.

Finding All Sources on the Network

Analysis can be used in order to locate NDI sources on the local network. To achieve this you would run:

NDIAnalysis.exe /find

Because it might take a few seconds to locate all sources on a network, you can specify a timeout – i.e., how long Analysis should scan the network. The default timeout is 5 seconds, but you can specify any number you wish. The following example invokes a 10 second time out for locating sources:

NDIAnalysis.exe /find /time:10

If you need to quit, press CTRL+C to cause Analysis to exit

By default Analysis will only only locate sources that are in the Public group. To find sources in other groups, one or more groups can be specified to locate sources in:

NDIAnalysis.exe /find:"Cameras"

NDIAnalysis.exe /find:"Public,Studio 1"

Example output from the finder is outlined below.

NDI Analysis v6.0.0.0 ()
Copyright (C) 2023 Vizrt NDI AB. All rights reserved.

NDI version : 6.0.0.0 ()

Searching for NDI sources ... (4 found) ()
    1: name="IPC (Channel 1 - Virtual PTZ)" ()
       host="192.168.1.191" (
       )
       type="NDI|HX (Camera)" ()
    2: name="IPC (Channel 1)" host="192.168.1.191" type="NDI|HX (Camera)"
    3: name="QUASIRANDOM (FaceTime HD Camera (Built-in))" ()
       host="192.168.1.193" ()
       port="5962" ()
       type="NDI" ()
    4: name="QUASIRANDOM (NVIDIA GeForce GT 650M 1)" host="192.168.1.193" port="5961" type="NDI"

Getting Source Statistics

Analysis implements a special version of NDI that will receive all data from a source, and provide diagnosis on that data. It is important to know that Analysis does not decode any data. This means that the CPU performance of the host machine on does not significantly impact the results shown. Analysis can thus be considered to correctly measure stream details as they might be received in an ideal receive instance. That is, only up-stream CPU performance, network, network infra-structure and machine network performance have an impact on the results.

To perform an analysis of a source you could enter the following:

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)"

Don't forget to press CTRL+C to exit quickly. Otherwise, to limit the duration of the analysis to just one minute, you could enter the following:

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)" /time:60

By default, Analysis will request the video stream to be in high bandwidth mode. High bandwidth mode is the full resolution and best quality mode for the video stream. If you wish to run analysis in low bandwidth mode, you could enter the following:

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)" /lowQ

The Analysis tool also has a frame checker mode, by default this is disabled. This is mostly useful when there is a suspected corruption in the video bitstream. To enable the frame checker mode, you could enter the following:

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)" /framecheck

Normally Analysis will output statistics about both the video and audio data from a source. It can be restricted to video only or audio only, for example:

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)" /videoonly

NDIAnalysis.exe /source:"QUASIRANDOM (NVIDIA GeForce GT 650M 1)" /audiooonly

All of the above modes can be used together. The following is an example run of NDI analysis application with all of the fields described in the notes, along with comments on the meaning and interpretation of the information provided.

NDI Analysis v6.0.0.0 
Copyright (C) 2023 Vizrt NDI AB. All rights reserved.

NDI version : 6.0.0.0

21:15:19.257: Video receiver creation succeeded. ()
21:15:19.257: Audio receiver creation succeeded. ()
21:15:23.287: Audio connection opened. ()
21:15:23.287: Video connection opened. ()
21:15:23.295: Capabilities, Recordable=false ()
21:15:23.295: Capabilities, PTZ control=false 
21:15:23.295: Capabilities, KVM control=false 
21:15:23.295: Product, Long name=Scan Converter v2 ()
21:15:23.295: Product, Short name=Scan Converter v2 
21:15:23.295: Product, Model=Scan Converter v2 
21:15:23.295: Product, Version=2.0000.0000 
21:15:23.296: Product, Serial number=0000 
21:15:23.472: Audio format changed. 44100Hz, 2 channels. ()
21:15:24.756: Video format changed. 1920x1200, progressive, aspect ratio=1.60, frame rate=60.00, no alpha channel. ()
21:15:24.258: Audio data rate (MBps). Avg=1.43 ()
21:15:24.258: Audio size (kB). Min=5.89 Avg=5.89 +/- 0.00 Max=5.89 
21:15:24.259: Audio recv (mS). Min=29.05 Avg=33.16 +/- 2.96 Max=39.23 
21:15:24.259: Audio send (mS). Min=29.01 Avg=33.13 +/- 4.77 Max=42.53 
21:15:29.259: Video data rate (MBps). Avg=69.39 ()
21:15:29.259: Video size (kB). Min=289.90 Avg=292.33 +/- 8.09 Max=320.91 ()
21:47:48.216: Video recv (mS). Min=11.44 Avg=16.66 +/- 2.30 Max=24.73 ()
21:47:48.217: Video send (mS). Min=8.72 Avg=16.67 +/- 1.47 Max=24.70 ()
21:15:29.260: Audio data rate (MBps). Avg=1.42 
21:15:29.260: Audio size (kB). Min=5.89 Avg=5.89 +/- 0.00 Max=5.89 ()
21:15:29.260: Audio recv (mS). Min=30.58 Avg=33.33 +/- 2.80 Max=40.16 ()
21:15:29.261: Audio send (mS). Min=27.98 Avg=33.32 +/- 4.81 Max=42.53 ()
21:15:34.262: Video data rate (MBps). Avg=82.48
21:15:34.262: Video size (kB). Min=284.67 Avg=286.25 +/- 7.71 Max=334.25
21:15:34.263: Video recv (mS). Min=14.29 Avg=38.13 +/- 34.44 Max=387.31
21:15:34.264: Video send (mS). Min=6.58 Avg=38.13 +/- 34.81 Max=388.28
21:15:34.264: Audio data rate (MBps). Avg=1.42 
21:15:34.265: Audio size (kB). Min=5.89 Avg=5.89 +/- 0.00 Max=5.89 
21:15:34.265: Audio recv (mS). Min=30.55 Avg=33.37 +/- 2.74 Max=39.66 
21:15:34.266: Audio send (mS). Min=28.02 Avg=33.39 +/- 4.69 Max=42.04 
21:15:39.266: Video data rate (MBps). Avg=124.18 
21:15:39.266: Video size (kB). Min=281.54 Avg=297.39 +/- 8.12 Max=311.32 
21:15:39.266: Video recv (mS). Min=14.06 Avg=31.00 +/- 46.95 Max=315.53 
21:15:39.266: Video send (mS). Min=4.70 Avg=31.00 +/- 47.64 Max=317.90 
21:15:39.267: Audio data rate (MBps). Avg=1.42 
21:15:39.267: Audio size (kB). Min=5.89 Avg=5.89 +/- 0.00 Max=5.89 
21:15:39.267: Audio recv (mS). Min=30.53 Avg=33.32 +/- 2.60 Max=39.65 
21:15:39.268: Audio send (mS). Min=28.49 Avg=33.32 +/- 4.61 Max=42.05 
21:24:44.547: Video receiver destroyed. () 
21:24:44.652: Audio receiver destroyed. () 

Note 12

This is one of the most important measures. and represents the frame rate of data actually received from the network. In this example we are watching 60 Hz video; the frame rate indicates an average frame time of 16.66 ms, which is expected. Because this is on a real network with computers that are doing other things, the variance is 2.3 ms, a value which is much smaller than a frame. If the variance number becomes too large there might be dropped frames.

All NDI applications should be designed to support frames within the Nyquist sampling limit, which would mean that any frame variance in the range of half the average should be the lowest reliable limit that would not require additional buffering or latency to display smoothly. At 60 Hz, this would require that the maximum reliable zero latency transfer rates would be 16.68 ms +/- 8.34 ms.

The minimum and maximum times are important; thus the largest delivery time (which is 24 ms in the example) provides a very important diagnostic tool. If you see values that are over 150 ms or so, it is very likely that network frames were dropped, and you should take steps to ensure that there is no packet loss on the network. It is important to bear in mind that some NDI sources might not send video frames if there is no video change occurring. Indeed, the example listed above is from Scan Converter 2 which does not provide new frames when there is no activity on-screen (thus reducing CPU time, GPU time and bandwidth or network time).

For advanced diagnosis of problems, the video receipt time can be combined with the video sending times described in the next note. It is important to observe these two values when looking at large maximum frame receipt times.

Some key reference times related to video timing are listed below (in milliseconds):

• 60 Hz = 16.66 ms

• 59.97 Hz = 16.67 ms

• 50 Hz = 20 ms

• 30 Hz = 33.33 ms

• 29.97 Hz = 33.36 ms

• 25 Hz = 40 ms

Source Color Information and HDR

In frame checker mode (see above), NDIAnalysis will also output any NDI color information if present. For example:

12:01:14.674: Color info changed. primaries=bt_709, transfer function=bt_709, matrix coefficients=bt_709
... 
12:01:18.573: Color info changed. primaries=bt_2100, transfer function=bt_2100_hlg, matrix coefficients=bt_2100 
...
12:01:23.165: Color info not found.

To reduce clutter in the output, color information is only shown when its status changes rather than on every frame.

If the source is HEVC compressed, then frame checker mode can be used to determine if the video uses 8-bit (SDR) or 10-bit (HDR) color values. NDIAnalysis will usually output this when a keyframe in an HEVC stream is received:

12:16:39.748: HEVC frame with timecode 0 is a keyframe and indicates a bitdepth of 10 bits

but may also output the bitdepth on other frames where an SPS element is found with bitdepth information in it.

Understanding Performance

Analysis now includes the ability to write out a CSV file that includes information about the timing of every video frame captured. This is enabled at the command line with the following example:

NDIAnalysis.exe /source:"My Source (Channel 1)" /csvvideo:"my source.csv" /time:120

This will then write out a file called “my source.csv” that will include a huge list of every video frame received, the time at which it was received and the details of every frame. An example of this output is illustrated below.

Audio frames can also be captured, for example:

NDIAnalysis.exe /source:"My Source (Channel 1)" /csvaudio:"my source.csv" /time:120

Video and audio frames can be captured together, as long as separate files are used:

NDIAnalysis.exe /source:"My Source (Channel 1)" /csvvideo:"my video source.csv" /csvaudio:"my audio source.csv" /time:120

The created files are text files in comma separated values format. They can be opened in any spreadsheet application, or inspected in a text editor.

Beware that timestamp and timecode values may be very large whole numbers, so when opened in a spreadsheet a suitable number format should be chosen for those columns which does not lose precision or truncates the values.

System time (utc)	Elapsed (ms)	dTime (ms)	Timestamp (100ns)	dTimestamp (ms)	Timecode (100ns)	dTimecode (ms)	X resolution	Y resolution	Aspect ratio	Frame type	Codec	Frame rate	Frame duration (ms)	Stream duration (ms)	Data size (Bytes)	Notes
00:17:54.73 8	0	0	17212618746625500	0	17212618746966600	0	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	41.71	237893	8 Mode 2
00:17:54.76 8	30.01	30.01	17212618747035200	40.97	17212618747383700	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	83.42	248295	[dTime < Prev frame duration] 8 Mode 2 Frame arrived early compared to intended frame rate.
00:17:54.80 9	70.98	40.97	17212618747455500	42.03	17212618747800800	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	125.12	248146	9 Mode 2
00:17:54.85 2	113.98	43	17212618747875000	41.95	17212618748217800	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	166.83	241047	2 Mode 2
00:17:54.89 4	155.99	42.01	17212618748285100	41	17212618748634900	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	208.54	249379	4 Mode 2
00:17:54.93 5	196.92	40.94	17212618748704800	41.97	17212618749052000	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	250.25	240985	5 Mode 2
00:17:54.97 7	238.92	41.99	17212618749124400	41.97	17212618749469100	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	291.96	250443	7 Mode 2
00:17:55.01 9	280.91	41.99	17212618749534700	41.03	17212618749886200	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	333.67	241389	9 Mode 2
00:17:55.06 0	321.88	40.97	17212618749954300	41.96	17212618750303300	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	375.37	250595	0 Mode 2
00:17:55.10 2	363.91	42.03	17212618750374500	42.02	17212618750720300	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	417.08	242025	2 Mode 2
00:17:55.14 4	405.88	41.97	17212618750794300	41.98	17212618751137400	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	458.79	252542	4 Mode 2
00:17:55.18 6	447.87	41.99	17212618751204300	41	17212618751554500	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	500.5	253334	6 Mode 2
00:17:55.22 7	488.84	40.97	17212618751623900	41.96	17212618751971600	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	542.21	242861	7 Mode 2
00:17:55.26 9	530.81	41.97	17212618752043800	41.99	17212618752388700	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	583.92	255295	9 Mode 2
00:17:55.31 1	572.8	42	17212618752453700	40.99	17212618752805800	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	625.62	243193	1 Mode 2
00:17:55.35 4	615.81	43.01	17212618752873700	42	17212618753222800	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	667.33	254911	4 Mode 2
00:17:55.39 3	654.96	39.15	17212618753293300	41.96	17212618753639900	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	709.04	244730	3 Mode 2
00:17:55.43 6	697.69	42.72	17212618753714100	42.08	17212618754057000	41.71	1920	1040	1.85	progressive	SpeedHQ	23.98	41.71	750.75	253564	6 Mode 2

Video CSV File

The reason that this is very valuable is that if you wish to determine why you are seeing problems sending video over the network then you can dump the CSV data at two different locations on the network, for instance you can capture it on the sending machine and also on the receiving machine. You can now look at the “timestamp” data on each frame and locate frames that exactly match between different locations and verify whether any frames are dropped (the timestamp is present on the sender side but not on the receiver side) or if the frames are delivered late (measuring the dTime). Note that there is always some variability on most typical networks and so some level of jitter is to be expected but if there are large timing hits then there clearly is something on the network that is stopping the smooth flow of packets (e.g. congestion control on a router).

System time (utc): The local time in UTC at which the frame was received by NDI Analysis.

Elapsed time (ms): The number of milliseconds of system time elapsed in the current session. Notionally this is the time point at the start of this frame.

dTime (ms): The time difference in milliseconds between the receiving time of this frame and the previous frame. Ideally this should be equal to the frame rate, but usually varies depending on network conditions and fluctuations in frame generation.

Timestamp (100ns): The timestamp of the frame, as specified in the frame's header. This is typically set when the frame is submitted to NDI by the system creating the stream. It is in units of 100 nanoseconds. It is usually related to the originating system's clock, but the exact relationship (particularly when the zero time was) is undefined.

dTimestamp (ms): The difference in milliseconds between the timestamp of this frame and the previous frame. Typically this represents the rate at which the source is submitting the frames to NDI.

Timecode (100ns): The timecode of the frame, as specified in the frame's header, in units of 100 nanoseconds. Usually this is an ideal time mark for the frame, set by the source generating the frames.

dTimecode (ms): The difference in milliseconds between the timecodes of this frame and the previous frame. Usually this is equal to the duration of the frames as indicated by the frame rate.

X resolution, Y resolution: The image resolution of the frame, in pixels.

Aspect ratio: The aspect ratio of the frame, as specified in the frame's header.

Frame type: Indicates if the frame is progressive, fielded, or interleaved.

Codec: The codec used for the frame, plus if the frame includes an alpha channel or not.

Frame rate: The frame rate of the frame, as specified in the frame's header.

Frame duration (ms): The frame duration in milliseconds. This is derived from the frame rate in the frame's header.

Stream duration (ms): The sum of the frame durations of all the frames received so far (including this one). As this uses the frame rate specified in the frame header, it is an ideal value. It should remain close to the elapsed time with some fixed offset.

Data size (Bytes): The number of bytes of image data in the frame. If the frame is compressed then this is the size of the compressed data.

Notes: NDI Analysis will report any potential issues here, such as a frame arriving significantly later than expected or the stream not providing sufficient frames to keep up with the intended frame rate. These can be starting points for further investigation.

Audio CSV File

The audio CSV file is similar to the video CSV file and shares many of the same columns. Instead of the video frame specific columns it has these audio frame columns:

Sample rate: The sample rate of the frame, as specified in the frame's header.

Num channels: The number of channels in the audio frame, as specified in the frame's header.

Num Samples: The number of samples in the audio frame.