DW sound propagation model measurements

[Dr.Sid]

This is results of my measurements of DW sound propagation model (SPM). It contains raw data, methods how they were obtained and my conclusions about how the DW SPM works.

First the raw data
Sea state 1 for all SSPs. The target is surface car carrier.

X axis shows distance in nautical miles.
Left Y axis shows track's SNR (signal to noise ration) as it is displayed on FFG's broadband station.

There are these SNR graphs here:
- gray line shows how SNR changes with distance with no layer, with bottom limited conditions. It is here just for reference. Note that it is not straight line. More about this later.
- blue line shows SNR in surface duct conditions how it was measured above the layer. Note that it is almost straight line, the most simple situation, spherical spreading.
- red line shows SNR below the layer. Up to distance about 5.5 nm layer just makes the signal little weaker (about 10%). It is the same for convergence zones conditions, so that line is not displayed.
- purple line shows convergence zones conditions above the layer. It is very similar to the surface duct, except for the peak at 30 nm. There is another peak at 60 nm but it is not measurable with SNR (it does not get over 0). However it can be seen nicely on the BB or NB display. Note that convergence zone is really narrow, it takes seconds to cross it.

Right Y axis shows depth (in ft) to show SHADOW ZONE shape.
- Shadow zone shape is shown as dashed line, for both surface duct and convergence zones conditions.
For surface duct, with distances greater then 5.5 nm when listening from below the layer, you will hear nothing. Surface target will be lost. And also surface listener will not be able to hear you, nor ping you. This is the shadow zone.
Between 5.5 nm and 11 nm this shadow zone is not bottom limited, but it has it's lower border. It starts at 5.5 nm at layer depth and goes down really fast. I guess it goes down even further but I could not measure it with FFG and these depths are not accessible for subs either. In distances greater then 11 nm you can consider shadow zone to be bottom limited.
For convergence zones conditions the shadow zone has different depth because the layer depth is different, but otherwise the shape is the same.




Now how did I measure this, just fot your info

- I used custom made mission with good weather and only target over deep water.
- I started at 1 nm going parallel with the target and moved array into different depths, grasping the situation at first. I launched BT buoy to get real SSP and layer depth. I also saved the game because you can't reproduce layer depth in any other way.
- Then I made 1 run for each graph line. I put array into depth I want, then I went away from the target, keeping the array at constant depth (more or less). Any time SNR changed I paused, jumped to navmap and used R tool to measure distance from ARRAY TIP (beginning of LOS) to target. It's not much exact method, but lines came out quite straight, so I guess it is OK.
- Then I made little more measurements around shadow zone beginning to know for sure where it starts and how does the bottom border of the shadow zone look like, same for convergence zones.
- Then I made some experiments when both target and listener where submerged to see how it behaves. I did no graph since I only found when the signal is weakened by the layer and when it is not weakened.
- then with some papers covered with numbers I put it all into Excel.

Conclusion

There is no measurable difference if you change array depth, until you cross the layer or shadow zone border. There are simply 4 cases:

Above - target is above the layer, if there is any. In bottom limited conditions consider all situations as 'Above layer'.
Under - target is under the layer, but closer than about 6nm, so it is not in the shadow zone. This apply same for both convergence zones and surface duct.
Shadow - target is under the layer and beyond 6nm, inside the shadow zone.
CZ - convergence zone - target is right at 30 nm or 60nm, above the layer, and this applies only to convergence zones situation.

This gives these combinations. Note that transmission is symmetric. If A can hear B, B can hear A with same quality.

Above-above - signal is affected only by distance, let's call this 'normal transmition'.
Above-under (or under-above) - signal is somewhat weaker. For subs it can complicate ID, but you will be detected.
Above-shadow (or shadow-above) - signal is completely lost. Both passive and active is useless.
Above-CZ - slight signal increase. Allows very short detection of target. Signal is too weak for ID.
Under-under - signal is weaker, by the same amount as in Above-under situation. So even if both target and listener are bellow the layer, the signal is weakened. Anyway there is no shadow zone when both listener and target are below the layer, so target is lost depending only on distance.

Now some other informations and ideas:

- in DW, SNR is just signal strength. It has nothing to so with noise. It does not change with my speed, it stays high even near array washout, when you simply can't see anything on BB display. However I did not test if it is affected by weather (I guess it must be, because I know even in DW bad weather makes targets harder to track). With good weather SNR looks simply like signal strength in decibels. The logarithmic scale is obvious because spherical spreading looks like straight lines.

- with bottom limited conditions (gray line) I tried to get far enough to lost the signal. I gave it up at 50 nm (I've got even farther away when measuring convergence zones, signal still there even over 60 nm). There was no line visible on the display but the TAG line was still there and reacting correctly. There is little trick here with these tags. When signal is lost they vanish about 2 minutes after that happens. But SNR changes seem to be right on time. Signal lost is best judged on single beam NB display.

- note that surface duct and convergence zones conditions give better signal transition above the layer than bottom limited conditions. This is correct, because the signal is trapped in shallow depths, even the amount of the effect seems right.

- The bulge on the bottom limited gray line could be surface shadow. With SSP used here rays would bend down, creating something similar. I guess it should be much more prominent, but I have no other explanation what it should be.

Bottom type influence:
Bottom type affects general detection ranges. Rock gives best sound transfer, mud about double transmission loss, sand about triple transmission loss. Bottom type affects sound propagation over all possible depths, shallow and deep.

Other links:
Sonoboy made measurements to better describe shape of the shadow zone. You can find it in this topic http://www.subsim.com/radioroom/showthread.php?t=124973
Or if just the image is enough for you, here it is: http://img228.imageshack.us/my.php?image=layergk9.jpg

[Fish]

A litle OT perhaps, but a nice pic nevertheless.

http://www.globalsecurity.org/milita...ages/FIG47.gif

[Dr.Sid]

Or here you can check picture from my ray-tracer (download here). You can see that although DW has it simplified, it catches all important features, like stronger signal above the layer, shape of the shadow zone and weaker signal below the layer.
In fact below the layer (outside shadow zone) the signal is not weaker (compare it with bottom limited situation), it is louder above the layer because sound is focused there.

I'm quite happy that DW has it more or less right. I expected worse results, as we know DW sometimes promises more than it meets (like torpedo wires, or sound model before 1.03).

[PeriscopeDepth]

Cool to hear that this simulation does a pretty good job of...simulating.

PD

[GrayOwl]

Originally Posted by Dr.Sid

Or here you can check picture from my ray-tracer (download here). You can see that although DW has it simplified, it catches all important features, like stronger signal above the layer, shape of the shadow zone and weaker signal below the layer.
In fact below the layer (outside shadow zone) the signal is not weaker (compare it with bottom limited situation), it is louder above the layer because sound is focused there.

I'm quite happy that DW has it more or less right. I expected worse results, as we know DW sometimes promises more than it meets (like torpedo wires, or sound model before 1.03).

This tool is serious, or it is an --entertainment--?
Only beautiful figure....
What levels of a signal you measure?:hmm:

[Dr.Sid]

Well .. I'm no naval expert, just rogue programmer. This only simulates bending of rays in medium with gradual changes of index of refraction. It would be same for light in glass or whatever.
So as far as you take sound as rays and you enter correct speed of propagation it is correct, as far as I know. But in reality it is much more complicated.

As for the signal levels .. what do you mean ?

[nzm30]

I'm intersted to hear if anyone has tried importing SSP data (mission with convergence zone selected) into SoundPropagation.exe and what results were obtained?

I created a mission with SSP=CZ and Russian Sub (saved converting feet to meters). Started mission and collected SSP data from sonar station.
Now SoundPropagation seems to only support whole numbers so I rounded to nearest meter/second, set options (surface bounce=0, distance scale=0.01) and I seem to be seeing first zone at around 12 nautical miles??

Probably me doing something stupid - but I would be interested in hearing the experience of others.

[Dr.Sid]

First: SSP profiles in DW are mostly cosmetic feature. They are idealized and I doubt they affect actual sound propagation model except for SSP type and layer depth.

As for convergence zones and my SoundPropagation simulator, these effects can be demonstrated, but they were not compared with real data, so don't take the results as something correct.

I will compare them, I have some real data from books. I'm doing some improvements of the SoundPropagation now so I include descriptions of model verifications with the next version.

[Pisces]

While I'm no naval/sonar or programming expert, whether or not you are doing something stupid because you got convergence zones at 12 nm is difficult to say without knowing all settings you used (if I am allowed at all). I don't understand why you'd want to set 'surface bounce' to 0 though (which means it's off). That's an important part in (multiple) convergence zones. It needs reflection at the surface if all or most of the ray are to converge. Also I think the most important variable in soundray paths is the steepness in the speedgradient. Also the relative steepness of the different sections in the profile seem to make big differences in the path shapes according to my experimentations with the program. All I'm saying here is you left the juicy part out from your post.

I too noticed that this program doesn't accept fractional numbers well in the SSP. Atleast not real-life reasonable numbers. It seems to crash when speeds entered go above 99.7 m/s. But I allready emailed Dr.Sid privately (if he reads this: my real name is Rico) about this and he said he was working on a new version with some other improvements, to be expected soon. Also, all specified points in the profile get a zero speed gradient (the blue line is vertical at those circles) while everything in between is a smooth fit. This is not really compatible with the datapoints from the DW SSP profile table. In the SSP plot in DW there is a sharp crisp turn in the gradient at the layer. Only in the deep soundchannel section (convergence zone type) do the points indicate a smooth turn in the gradient. The DW SSP is very simple in shape btw, it's only made up of straight and parabolic (if convergence zone) gradient sections. I think I calculated some time ago that the true 0 speedchange depth of the deep soundchannel lies somewhere in between two specified depth points there. The minimum speed wasn't one of the points, but I'm not sure. And it would matter much. Dr.Sids program needs smooth curves everywhere (as he explained to me), so the local speed at depth is allways a bit off compared to what DW has (even if it didn't have round-off errors).

But you have to remember we are comparing apples and oranges here anyway. DW probably has a much more simplified sonar engine than this ray-tracer, if not radically different. There's no way this ray tracing could be done realtime in DW I think. But it's a great tool to get an idea of how those sound paths are created.

I would like to see from Dr.Sid if he can post a screendump of his program that shows some examples of sound paths he got with the SSP from DW in his tests. As far as the
SSP can be simulated ofcourse. Or atleast give us the datapoints in the SSP so we can try ourself.

As for the signal levels .. what do you mean ?

I think what GrayOwl is asking is what does the pixel intensity mean. Is it the summation of (fading out) amplitudes of all the passed rays through it. Or is it summed local soundpressure of each passing ray. (I hope you see the distinction I mean, I hardly do ) And/or is it in a linear or logarithmic scale. Which would make faint pixels way more fainter in terms of levels relative to source. I did notice that when the calculation an plotting is done the mouse can be used to look up decibel levels, which I have a hard time understanding in-and-of-itself, along the path. But it's not covered well inbetween the rays if the delta-rayangle is not fine enough. As you (Dr. Sid) allready know yourself.

[Dr.Sid]

I see .. the summing is done in linear scale, so does the colors on the display (multiplied by brightness). Ray intensity at source is 1, independent on rays count. Which means more rays give brighter screen with same brightness.

For TL display at mouse pos I use logarithmic scale, by dividing actual and base energy. Base energy is computed at 1 yard or meter (based on convention) from sound source and since I can't easily compute base energy (it depends on rays count and some other factors) I 'measure' it too. I pick some samples at some distance (not 1m, that would be 1 pixel and less), average them, make the correction for the distance and I use this as base energy. I take samples from about 10 pixels from left, which can be distance 10-100 meters based on X scale. I looked for best approach so sphearical spreading (straight SSP) gives correct numbers, and now it does with error not more than 1db, which is acceptable if you consider amount of random scattering.

I'm working on new intagration system which will fill area between rays uniformly and exactly. There will be no need for scatering, it will have better coverage and it will be even faster (I hope).

[nzm30]

Pisces, the reason I selected ‘surface bounce’ equal to zero was just to simplify the display. I assumed that surface refection would only become relevant in the creation of the 2nd. CZ. I was only interested in following the rays travelling down from the source into the deepsound channel and returning to the surface (1st. CZ). If I had surface bounce on the display became confused due to rays that did not penetrate below the layer but instead bent back towards the surface, reflected etc. etc.
So it was a ‘more out of interest’ experiment to compare the DW SSP data against SoundPropagate utility and I was interested to hear the views of others.
As it is being said, the SSP values being displayed in the sonar station are most likely only cosmetic and I am sure that the actual simulation of CZ in DW may be so simple as to just intensify sound level every 30 nautical miles. But I suppose this is only reasonable, as with a full simulation we would not have had resources (PC) available for much else.

You know, it is always fun to dig about.
So I look forward to the improved SoundPropagation.exe from Dr. Sid and any enlightenment as to the sonar model used in DW.

[Pisces]

Originally Posted by nzm30

Pisces, the reason I selected ‘surface bounce’ equal to zero was just to simplify the display. I assumed that surface refection would only become relevant in the creation of the 2nd. CZ. I was only interested in following the rays travelling down from the source into the deepsound channel and returning to the surface (1st. CZ). If I had surface bounce on the display became confused due to rays that did not penetrate below the layer but instead bent back towards the surface, reflected etc. etc.
So it was a ‘more out of interest’ experiment to compare the DW SSP data against SoundPropagate utility and I was interested to hear the views of others.
As it is being said, the SSP values being displayed in the sonar station are most likely only cosmetic and I am sure that the actual simulation of CZ in DW may be so simple as to just intensify sound level every 30 nautical miles. But I suppose this is only reasonable, as with a full simulation we would not have had resources (PC) available for much else.

You know, it is always fun to dig about.
So I look forward to the improved SoundPropagation.exe from Dr. Sid and any enlightenment as to the sonar model used in DW.

Ok, just so long you know you're seeing only half the picture it's allright. But I hope for Sonalyst sake they didn't make the 'simulation' that simple. They said outright (not to me but others here I believe) that alot of CPU power went into the sound propagation code of DW, aswell as SC. They better not be talking about sloppy and slow code. But I admit, I may have been tricked by that SSP eye-candy.

p.s. You may want to do something about the color you used in your post. I got to see black text on a black background. Well, this quote pretty much solved it, I removed the color coding statements.

[Dr.Sid]

Good trick for such texts is to select them either with mouse or CTRL+A combo.

As for sound simulation, it goes beyond transmision loss. You must still generate all broadband and narrowband displays, even with high time compressions. But for example adding several sonobuoys does not slow things down much, while adding few ships slows thing down nicely. So I guess AI is to blame.

I'm pretty curious about measuring convergence zones in DW, I have some time now. But upgrading SoPro is number one priority now (shorter name for sound propagation .. how do you like it ? lol).

[Pisces]

Pisces, the reason I selected ‘surface bounce’ equal to zero was just to simplify the display. I assumed that surface refection would only become relevant in the creation of the 2nd. CZ.

Took some time to think about this longer. But I don't thinks so. The convergence zone area at the surface is a small area. Which means that rays that strike the surface on the near end of (even the 1st) CZ reflect back down and immediatly reinforce the rays that come up to strike at the far side of the CZ. Which results in a lesser signal if reflection is off.

As an example see the following images:
With reflection:

No reflection!!!:

Strangely enough the center of the starpattern doesn't show increased signal level, but some of the arms do have higher dB levels

But to understand the paths the rays take it is smart to simplify this way and not get distracted.

[Dr.Sid]

dB measurement is not much exact .. just wait for the next version.
You are right .. with no reflection half of the sound energy is wasted, but the pattern will be more or less same.