[dougsmith]

I am interested in building a sub with multiple vents of slightly different lengths to decrease port lengths and smooth out resonances. When you increase the number of vents in WinISD, the vent length goes up proportionally. I assume this is giving the total length of all the vents. If that is the case, then the first port resonance will not be correct because it is being calculated from the total length of all the vents and not from the actual lengths of the individual vents. Agree?

There is another thing that has been bugging me for a while… When you change the vent end correction from two free ends to two flanged ends in WinISD, the vent length decreases. However, flaring the ends effectively increases the port diameter slightly so the port length should increase. This is what happens in Unibox, which I believe is correct. Comments?

- Doug

 

[Mike P.]

The vent length in WinISD is for each port.

Flanged end does not mean a flared end. For an explanation look up Port Terminology in the help files.

The program states port length assuming a straight port. For flared ends add 1 inch to the length given.

 

[dougsmith]

Ok - my mistake - well, the documentation does not actually define clearly what flanged end means but I get it now... it is just an end which is flush to part of the box. So adding more ports requires each one to be substantially longer, then. Using my model if I had one 3" port the length would be 5.2", but using four would make them each 26", not so different from one 6" port.

-Doug

 

[willy-be]

Math errors...

Jack Hidey, someone who knows a thing or two about loudspeaker design, posted on AVS he found numerous math errors in WinISD Pro within the first few minutes of playing with it.

 

[Mike P.]

No modeling program is perfect.

 

[willy-be]

The math used in the algorithms should be correct

 

[Mike P.]

Agreed. Bottom line is these modeling programs work well enough. Thousands of subs have been built based on these programs with no complaints about the end product.

 

[willy-be]

The math errors in question are specific to the algorithms used in WinISD, other programs don't have these errors.

 

[Mike P.]

Real world results are what counts. Ilkka's subwoofer testing has shown that WinISD modeling is close to measured performance.

 

[DrWho]

Re: Math errors...

Jack Hidey, someone who knows a thing or two about loudspeaker design, posted on AVS he found numerous math errors in WinISD Pro within the first few minutes of playing with it.

Can you post a link that describes the math errors he found?

I have had extremely good correlation between WinISD and the real world so I'm kinda perplexed. At the end of the day, the whole purpose of modeling is to know what will happen in the real world.

Along that note, I've also had very good correlation between hornresp, akabak, winisd, and bassbox...

 

[willy-be]

Can you post a link that describes the math errors he found?

His reply when asked that question.

"With regards to the errors in WinISD, I'm not going to discuss them. If the writers of the software want to pay me to fix them, I'm game."

 

[Mike P.]

So he won't say what the assumed "errors" are unless he gets paid. That says it all.

 

[Ricci]

There are some errors and some very real limitations but there aren't glaring errors in the basic response calculations. It would've been widespread knowledge by now if it was. The models appear to match reality and other modeling program predictions quite closely actually.

 

[brent_s]

Re: Math errors...

Jack Hidey, someone who knows a thing or two about loudspeaker design, posted on AVS he found numerous math errors in WinISD Pro within the first few minutes of playing with it.

Just to pile on since I was the party having that debate with Jack. I'm pretty sure he didn't say "math" errors, but only said "errors" with no qualification of the type.

Right off the top of my head, I can name several "errors" in WinISD that don't effect its accuracy if you understand the physics.
1) When you change the distance on the signal tab, the output model doesn't change.
2) If you add any EQ to the model, WinISD computes output and excursion assuming you have an infinite power supply and the "system input power" is just a starting point. Example: if you enter 100 watts and then add +3dB of EQ, the model will show output/excursion for 200 watts.
3) Some of the EQ screens allow the user to enter fields that don't actually mean anything to that particular filter. Example: a subsonic filter is modeled with highpass, "S.O.S. user specified fc and Q" yet you can still enter an order for the slope...except I'm pretty sure S.O.S. represents Second Order Sallen-Key since that's the most common circuit for implementing rumble filters in plate amps. Technically, S-K filters can be any slope, but WinISD hasn't implemented anything except 2nd order so changing the order for an S.O.S. filter doesn't change the model's behavior. You can see the same WinISD behavior if you try to change the Q of a Butterworth filter (by definition, Butterworth Q=.7071) or change the order of a L-R 4th filter to 2/5/3/x.

Ultimately, like any tool, you still need to understand a bit about how it works. None of those 3 items are show stoppers and are easily recognized and compensated for if you understand a bit of the physics involved. With a little bit of knowledge, WinISD is a pretty good solution for free.

-Brent

 

[dougsmith]

Re: Math errors...

2) If you add any EQ to the model, WinISD computes output and excursion assuming you have an infinite power supply and the "system input power" is just a starting point. Example: if you enter 100 watts and then add +3dB of EQ, the model will show output/excursion for 200 watts.
-Brent

I guess I am not the only one who has been decieved by this one (with a Linkwitz transform).

Also, how hard would it be to add correction for port flares?

-Doug

 

[DrWho]

Re: Math errors...

2) If you add any EQ to the model, WinISD computes output and excursion assuming you have an infinite power supply and the "system input power" is just a starting point. Example: if you enter 100 watts and then add +3dB of EQ, the model will show output/excursion for 200 watts.

I wouldn't call that one a bug because you're just entering a nominal drive level. The real power handling requirements actually shows up on the VA plot...

But to add another one to the list...you have to manually compensate for thermal compression when modeling an increase in voice coil temperature. But like you said, if you understand the physics then it's easy to compensate for.

 

[brent_s]

Re: Math errors...

I guess I am not the only one who has been decieved by this one (with a Linkwitz transform).

Also, how hard would it be to add correction for port flares?

-Doug

The problem with compensating port length for flared ends is knowing the radius of the flare. The radius determines the transition point for the effective port length. I think Collo's got some of the math on his site. Like any variable, it could be added to the port tab *if* the project were still active, but nothing's changed since 2004. I think the WinISD guys offered to OpenSource the code at some point, but I'm not aware of anyone taking them up on it.

-Brent

 

[brent_s]

Re: Math errors...

I wouldn't call that one a bug because you're just entering a nominal drive level. The real power handling requirements actually shows up on the VA plot...

But to add another one to the list...you have to manually compensate for thermal compression when modeling an increase in voice coil temperature. But like you said, if you understand the physics then it's easy to compensate for.

Looks like to me VA increases based on that infinite power supply just like excursion/spl when you dial in some EQ. I'm too lazy to look up the formulas right now, but it appears to simply compute VA based on the signal power and calculated impedance values. Doubling system input power or adding +3dB of EQ produces the same change on the VA graph.

It's an error in the sense that I've seen a number of folks use SOP of entering the max sustained power for their amp into the signal tab and then dialing in some EQ and start getting excited about outrageous SPL numbers. Amazingly, you can use an SA240 with an LT to drive a Tempest-X/SDX15 to Xmax at 20Hz if you work the model "just right".

In an ideal world, the output models would reverse integrate the EQ'd response by treating the input signal as the max available power and labeling it as such. Then, for good measure, also plot the reponse in a different hatch pattern as if the infinite power were available, but that's likely to induce as much or more confusion as the current method.

-Brent