Oh this is mostly for horns for drivers?
But what then if sides are straight and at one certain angle like a tapped horn, that's NOT parabolic, I thought THAT was conical (and how most THs are)
+This shows the opposite to what you're saying about corner reflectors:
Martinsson's Blog - Showdown in Stenungsund (THAM 15 & MKII proposal)
When parabolic is NOT straight sides, google it!
A quadratic equation makes a parabolic curve
Parabolic is different from parallel!
Conical is straight sides, but on a round pipe, was thinking conical in two dimensions.
So how do you simulate horn segments with two straight sides that keep expanding at a fixes degree?
I mean exactly like something conical seen from the side...
Tham12 is like this, but sims with "con"
Martinsson's Blog - THAM12
A quadratic equation makes a parabolic curve
Parabolic is different from parallel!
Conical is straight sides, but on a round pipe, was thinking conical in two dimensions.
So how do you simulate horn segments with two straight sides that keep expanding at a fixes degree?
I mean exactly like something conical seen from the side...
Tham12 is like this, but sims with "con"
Martinsson's Blog - THAM12
Last edited:
You're both wrong.
Everyone appears to build a conical TH
Read this "Simple Tapped Horn Tutorial using Hornresp"
Simple Tapped Horn Tutorial using Hornresp - AVS Forum | Home Theater Discussions And Reviews
The confusion seems more widespread here than on speakerplans
But the first question that got me here again was what hornresp-inputs Djim used for the comparison?
Everyone appears to build a conical TH
Read this "Simple Tapped Horn Tutorial using Hornresp"
Simple Tapped Horn Tutorial using Hornresp - AVS Forum | Home Theater Discussions And Reviews
The confusion seems more widespread here than on speakerplans
But the first question that got me here again was what hornresp-inputs Djim used for the comparison?
No, you are wrong, and this has been discussed to death already. Maybe you should google it instead of telling the people that have been following this CON vs PAR discussion for years to google it.
I'm not mathematically capable enough to explain it in math terms but here's a few quotes from David McBean.
From here - http://www.diyaudio.com/forums/subwoofers/143714-lab12-tapped-horn-15.html#post3146632
From here - http://www.diyaudio.com/forums/subwoofers/143714-lab12-tapped-horn-15.html#post3149933
Click the link to see the attachments, the attachment pics are very important and if you do the math you will find the same results.
Pay particular attention to the sentence I bolded, it's the key to everything.
From here - http://www.diyaudio.com/forums/subwoofers/119854-hornresp-470.html#post3996300
I'm not going to read your THAM link or your Hornresp tutorial link to find out what they say about PAR vs CON but rest assured, if the THAM is simulated using CON segments it is wrong and if the tutorial is saying you should sim horns with two parallel walls and two expanding walls as CON it is wrong.
Last edited:
Yes to the concept but no to the technicalities.
Danley's famous example which was (IIRC) a 6 meter horn. He said that you could build it as an expanding single segment or as 6 meter long stepped sections and you would get the same thing.
You will get CLOSE to the same thing, but if you sim it or build and measure it you will find that it is not the same thing at all. Close but still quite different.
As to the sentiment itself, absolutely yes you can build a stepped horn (or tl) without any problems and it will be similar but it won't be the same thing as a smoothly expanding (non stepped) design. Close but not the same.
Just a few more notes on CON vs PAR.
Not many people are even aware of this issue unless they are very good at math and understand the CON and PAR expansion rates OR unless they have followed this forum and in particular the Hornresp thread. So it's no surprise that there's a ton of bad info out there from other forums and resources. And anything related to Hornresp that was published more than a few years ago won't talk about PAR at all since this is a fairly new feature in Hornresp, so if you are looking at old tutorials you'll find a CON recommendation for this type of enclosure made of flat sheet goods since PAR was not available.
Second note, if you are dealing with a segment with a constant cross sectional area (straight duct, beginning CSA = end CSA) it doesn't really matter if you use CON or PAR, you will get the same result. In fact you could use EXP if you really wanted to since the result will be exactly the same if cross sectional area doesn't change. But when CSA of a segment does change it's important to use choose the right option. It's not the end of the world if you use the wrong option but results are more accurate with the proper selection.
And finally, if you wanted to sim or build ANY type of horn flare other than CON, PAR or EXP, any flare can be approximated with CON, PAR or EXP if you break it up into enough segments. For example if you simulated a HYP (hyperbolic exponential) horn in Hornresp but wanted to approximate it with PAR segments (there's a lot of reasons to do this), you can. Check this post from LTD02 for details. From here - SubMaximus -- A Large Front-Loaded Horn for UXL18 and Stereo Integrity HT 18" - Page 2 - AVS Forum | Home Theater Discussions And Reviews
Relevant info copied and pasted -
here is the idea put another way. we want to approximate the exponential curve with straight sections. can it be done? yes. it works "good enough".
this is the horn we want to model (just an example, not to scale or anything):
this is approximating the exponential curve with straight sections (illustrative example, not to any scale). straight boards, of course, create straight segments that are technically modeled with "par" in hornresp. but by going the other way, straight sections can be used to create an exponential flared horn. if that is what one wants.
it is the same concept as this, but with fewer straight lines, so the curve isn't quite as accurate, but still "good enough":
some folks may not believe that curve comes from straight lines. if you haven't seen this before, give it a shot. it is fun to draw it out.
Last edited:
Seeing that I have previously demonstrated it with a built AND measured example (my POC #2), where the major differences started to occur above 150 Hz (actually more like 300 Hz), I stand by my previous statement - it does not make a difference at bass frequencies.
I'm not sure how you performed the comparison. Hornresp can only sim a stepped horn with two steps, since S23 is dedicated to the step transition and then S4 can't be used (it would be another step transition but then there's no more segments to create another segment). To really compare you would need to either use Akabak to sim a bunch of steps or build BOTH a smoothly expanding horn AND a stepped horn and compare the measurements.
To be clear, if there are enough steps in the horn you can approximate any flare. If there's an infinite amount of steps you can copy any flare perfectly.
But since Hornresp can only do two steps and I'm not going to break into Akabak to sim one step per meter here's a quick comparison of an exponential horn vs a two step horn. Both horns are the same length (within .01 cm), both are the same volume (within less than 1/2 liter), both stepped segments have the same volume as the EXP counterparts. EVERYTHING is the same except stepped vs EXP flare profile.
As you can see the differences start as low as an octave below the low knee and there are some pretty huge differences above that.
I've done this exercise before with Akabak and more steps and it does get more accurate pretty quickly when using more steps but I will stand by my position that unless you use a lot of steps it's not going to be the same thing.
As it is, this is a 3 meter horn with 2 steps, Danley's example (which he showed no sims or measurements for) was (IIRC) a 6 meter horn with 6 steps so would be more accurate due to having more steps, but still not the same thing.
Pics on left - exp horn
Pics on right - stepped horn
Final pic - frequency response overlay
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
I believe that Tom Danley had also said or suggested that the LENGTH of the steps also had an impact on the FR (compared to a fully-flared horn). If I remember correctly, they start to have an impact at either 1/4W or 1/2W (it's been a while - I can't remember). For my designs, I assume 1/4W to be safe. You used 144 cm and 160 cm stepped segments in your model (I'm not sure how you chose that as the best approximation, but that's another subject). Given the 1/4W "rule", you'd start to see an impact on the FR at around 60 Hz and 54 Hz. So obviously trying to approximate your bass-horn example with two steps is not going to work.
In any case, I haven't seen anyone try to build a stepped horn using just two steps. Have you? Even my little PO2 build uses three steps AND two flares, and approximating it with a HornResp fully flared sim worked out well as the impedance response shows.
Just because I'm bored and TL.app is super easy to sim with I did another comparison even though I already knew what was going to happen because I've done this before. This example is a 1235 liter front loaded horn that's 6 meters long. In both cases there are six segments, each one meter in length. In both cases I've shown all the work, the schematic and the info for each segment.
Along the top row I've shown the somewhat exponential version made up of six PAR segments.
Along the bottom row I've shown the stepped version made up of six PAR segments.
In red I've outlined the important info, start and end CSAs for each segment, length of each segment and volume of each segment. Segment lengths and volumes are exactly the same in both examples so obviously start and end CSAs are different to accomdate the exponential vs stepped models.
Now the results. Exponential is on the left and stepped on the right in the image below.
As I said a few times, close but not the same. There's about 1 db difference at the low knee and about 2 db difference in some places up around 100 hz.
Also the peaks are all at different frequencies, the peaks in the stepped (right) graph are all shifted slightly to the right. It's hard to tell but I think the tuning is slightly higher in the stepped version, I can check that later. The curve shape itself is also slightly different.
Close but not the same. it would take more than 6 steps to make it the same, and at the 6 step stage we are already well into diminishing returns as is clearly evident compared to the two step comparison example in the last post, so it would take quite a few more steps to improve the correlation significantly.
Along the top row I've shown the somewhat exponential version made up of six PAR segments.
Along the bottom row I've shown the stepped version made up of six PAR segments.
In red I've outlined the important info, start and end CSAs for each segment, length of each segment and volume of each segment. Segment lengths and volumes are exactly the same in both examples so obviously start and end CSAs are different to accomdate the exponential vs stepped models.
An externally hosted image should be here but it was not working when we last tested it.
Now the results. Exponential is on the left and stepped on the right in the image below.
As I said a few times, close but not the same. There's about 1 db difference at the low knee and about 2 db difference in some places up around 100 hz.
Also the peaks are all at different frequencies, the peaks in the stepped (right) graph are all shifted slightly to the right. It's hard to tell but I think the tuning is slightly higher in the stepped version, I can check that later. The curve shape itself is also slightly different.
Close but not the same. it would take more than 6 steps to make it the same, and at the 6 step stage we are already well into diminishing returns as is clearly evident compared to the two step comparison example in the last post, so it would take quite a few more steps to improve the correlation significantly.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Given the 1/4 wave rule, the FR of your example will started to be impacted at 345/(1x4) = 86 Hz, given its meter-long steps. Now, review what your sims are showing you with that info in mind.
Also, a 6M long basshorn is a ridiculously long one - double or more the length of some of the well-known DIY examples featured on this forum. Try repeating your example using one of the well-known DIY builds (maybe the THAM15) as a guide, and stepping it like you did in your example above.
I showed a quick example with 2 steps because that's all that Hornresp can do. The dimensions chosen were just random. i wasn't expecting to do a more complex 6 step model but I ended up doing it anyway.
The info I am showing shows clear differences even at the low knee around 16 hz there's a 1 db difference and it's hard to tell but I think the tuning is shifted a bit as well, I'll check that later.
I did 6 meters with 6 steps because that what I remembered from Danley's example (in which he showed no proof or info of any kind). There's no chance at all that I'll be doing another example (at least not tonight), I've clearly shown here that the results are close but not the same, as I said repeatedly. You can do all the different examples you want and the results will be close but not the same. if you have a specific example you want to see break it up into 6 segments and give me all the segment info (and the rest of the horn and driver info) and I might do it later.
I looked up the Danley quote after I posted the info I showed. If you want to see it, it's here. The 6 meters with 6 steps was correct, the other dimensions and tuning were off though in my example and his. Only the bolded part is important here but the rest of the post is interesting so I left it.
Ah, that is the passage. I think I got my 1/4W guide from the first paragraph, as it appeared to pretty obvious and I've seen it used before (e.g. in estimating where cabin gain starts in cars).
I'm not sure I follow TD's jump to using 1/2W as a guide in his example.
In the last paragraph he's talking about the horn's parallel walls (the sides). There's a width mode that 1/4 wave simulators don't even bother with, they only calculate the length modes. So just like in a room with parallel walls, you will theoretically get a notch at the 1/2 wave frequency between the walls.
In the 3rd last paragraph he's talking about the same behavior for waves going around bends. The air is dispersed (not evenly but still dispersed) as it goes around the bend so some of the air takes a shorter path around the inside of the bend and some takes the longer path around the outside of the bend. When they recombine they have traveled a dissimilar distance so he is saying there will theoretically be a notch due to that.
As far as the 1/2 wave notches between parallel walls and due to path length differences around bends, I'm not sure how much these will show up in measurements. First of all, unless the distances in question are pretty large, the notches will be way above the passband of a subwoofer. Second, there's internal damping due to enclosure walls not being infinitely rigid.
I haven't seen any really deep notches in any measurements in the subwoofer passband frequencies that were not predicted in simulations so I wouldn't worry too much about the last few paragraphs. Technically they might be correct but in practice it isn't something to really worry about, especially with smaller subs (less than Labsub size).
I was referring to this part of the quote: "Imagine a horn, that was 6 meters long, with a 28 Hz flare and a 7 sq. meter mouth area. Instead of making a traditional horn, one made it out of pipes of stepped diameters, a "digital" horn. How many steps do you have to have before it works like a horn, any guesses?. Well if one made each pipe 1 meter long and did the expansion in 6 steps what would you get? You get the exact same thing as a true exponential horn up to 150 Hz."
Given the use of 1M segments, the 1/4W guide gives 344/(1*4) = 86 Hz, not 150 Hz.
Given the use of 1M segments, the 1/4W guide gives 344/(1*4) = 86 Hz, not 150 Hz.
If you want the compiled Labhorn notes (where this came from) I can send them to you.
Just remember that when reading Danley he plays REALLY loose with the details. For example, when he says the expo and the 6 step digital horn are "the exact same thing", you will find if you do the exercise that they are not exactly the same. Close but not exactly the same.
Another example is the Labhorn itself. Before folding the plan was for the horn to be as such:
If you analyze the folded Labhorn it's not even close to that spec. The rear chambers are massivly undersized (due to a mistake that he openly admitted) and the flare is undersized as well compared to the spec above. Some of the losses he attributed to "tuning for maximum smoke", but it should be obvious that the spec'ed design never would have fit into the Labhorn final truck pack dimensions he was shooting for.
He also said that stuffing the rear chambers in the Labhorn would fix his folding mistake and gain back the volume he lost during folding, but this is not true either, and will only make the thermal issues that the Labhorn is famous for much worse.
The guy is a genius but he's very loose with the details so you can't take everything he says to be fact. It's hard to even guess where he came up with the 150 hz figure in that quote. I have a couple of guesses (not related to 1/4 or 1/2 wave theory) but that would just be speculation.
Last edited:
Oops, that's not correct and it's too late to edit. I forgot to include the volume of the sealed chamber which is 500 liters, so it's actually 1735 liters total. Large but still much smaller than Danley's proposed example.
Just be aware that a hyperbolic-exponential horn approximated using multiple segments will always be simulated using a plane wavefront model, even when Cir > 1. The HYP option on the other hand will simulate a hyperbolic-exponential horn using an isophase wavefront model when Cir > 1, which gives more accurate results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.