Yes, I use stock TAD adapter. I also think that it is far from optimal. But for the first iteration to compare the horns it will be ok. I can then try modeling with other adapters if any ideas come up.
Ok, I will provide Acoustic Impedance, Onaxis SPL at 3 meters, Farfield Horizontal/Vertical polars responses and the Balloon graph, and Directivity index.
Last edited:
That is not the polar map for the ES290. The latest test data is shown below.
The ES290 is designed to provide a gradual narrowing of directivity that is well behaved, consistent, and wide. Typically the treble is handed over to a dedicated HF horn with matched coverage anywhere between 4kHz-8kHz depending on the solution. The other critical metric is a smooth on-axis frequency response. All other comparisons should look at this closely, since a linear response is an indication of a trouble-free time domain (resonances, diffraction etc.). The response below is with the RCF ND850 compression driver, just for reference and comparison against other possible solutions (Yuichi A290 for example).
If a design achieves so called 'perfect' constant directivity, but at the expense of on-axis linearity in the frequency response, then you can kiss any sonic benefit goodbye.
These are results that imply some further questions. Is this the raw response on-axis without any eq?
If so I would say that it implies that your horn does neither provide sufficient acoustical loading nor directivity control and it would stand diametral to the horn class presented in this thread.
Also the Axi2050 has only about 5dB decrease in your presented measurement in this horn up to 20k. This is physically impossible (with horn loading) as in a horn with sufficient acoustic loading the Axi2050 is about 25dB down at 20k compared to the midrange. The amount of beaming in your horn seem to compensate some of the natural decrease of any large format compression driver (on axis) at the expense of the off-axis response. The same for the RCF driver. Less than 5dB decrease. How is this possible?
It would be interesting to see a free-air impedance shot of the ND850 and then attached to the ES290.
I am the totally opposite opinion corresponding strong beaming horns. This is the most annoying thing - a substantial amount of directivity control is a must imo without the need of of eq beyond the mass fall-off.
If so I would say that it implies that your horn does neither provide sufficient acoustical loading nor directivity control and it would stand diametral to the horn class presented in this thread.
Also the Axi2050 has only about 5dB decrease in your presented measurement in this horn up to 20k. This is physically impossible (with horn loading) as in a horn with sufficient acoustic loading the Axi2050 is about 25dB down at 20k compared to the midrange. The amount of beaming in your horn seem to compensate some of the natural decrease of any large format compression driver (on axis) at the expense of the off-axis response. The same for the RCF driver. Less than 5dB decrease. How is this possible?
It would be interesting to see a free-air impedance shot of the ND850 and then attached to the ES290.
I am the totally opposite opinion corresponding strong beaming horns. This is the most annoying thing - a substantial amount of directivity control is a must imo without the need of of eq beyond the mass fall-off.
Just reiterating the design goals for the ES290 Biradial, which puts emphasis on low coloration, perfect time domain performance, as well as a well behaved, consistent off-axis character. Typical of all exponential horns, loading is improved in the upper treble compared to an equivalent constant directivity design. These are all very well known trade-offs. I only caution that in the pursuit of constant directivity, on-axis frequency response may suffer as a result. The simulations I saw earlier in the thread only showed the normalized polar map characteristics. Like everything in audio, a balanced design looks at all the parameters and settles on something after juggling all the variables and deciding what is important. There's no right or wrong, just specific design goals.
The 0 degree axis will tend to suffer more from all diffractions than any other curve when the DI gets close to flat. That does not have to be the design axis and when it is at 10 to 25 degrees off axis in many cases those diffractions are gone, of course they still exist at whatever axis the 0 degree curve becomes. There is a large range between a truly flat DI and anything else that might be labelled as constant directivity. Normalized polar plots and polar curves don't hide these problems, particularly not if the zero degree axis is the normalizing one. The hot spots of dark red colour around the zero degree line indicate where the on axis differs from the other axes.
For me designing horns through simulation or evaluating them as here hinges on the directivity response, as that is the part that is simulated most accurately and is the most useful thing to see for me in comparisons. I think the Yuichi still remains a reasonable choice for anyone looking for more consistent directivity without giving up loading to low frequencies or significant beaming and without the need for a lot of corrective EQ. The measured on axis linearity with different drivers has already been established in the original article and other people's own measurements. This was meant to be a more modern presentation of the directivity through simulation.
The majority of these simulations were performed with constant velocity instead of constant acceleration. This makes no difference in normalized polars but it does in non normalized ones. A constant velocity will give a rising slope whereas constant acceleration will give a falling one which is the more common presentation. Neither is "right" so it is a matter of picking one for the job at hand. Posting constant velocity curves would just confuse most people and create a lot of noise about it so I didn't. This is the on axis of the V2-3 simulation, a relatively smooth upslope, no cause for concern there.
Indeed and for this reason, I would prefer the thread remain a discussion of technical aspects and performance rather than move into opinion as we all value things differently for different reasons.
It is interesting that google still has the file in cache what @NicoB posted here:
One big difference to your newly provided data is that you missed almost one octave to the low! They new graph begins with 500Hz where your old graph show some midrange narrow. This is some kind of clever if you want to fool us. Who is familiar also know how to tune these graphs with special settings/constraints to look better or worse.
I was always under the impression the Yuichi A290 was a biradial design, which in some ways helps in keeping coloration to a minimum while loading the driver down past 300 hz. There is a jump (flip) in vertical directivity around 1000 Hz. This isn't an issue most of the time, but it can make the ceiling and floor reflectons add up to a shouty kind of FR balance.
You state so many misleading things here and on your site regarding the ES290/Yuichi that I really do not know where to start.
Ok, is a linear response really a proof that the time domain is fine or that you do not have any resonances/diffraction? First you have to precisely define what you mean by using the terms resonance and diffraction, either driver or horn related. The time domain issues are largely related to the driver in most cases. When you measure what you say linear response then you should bear in mind that your driver/horn is radiation into space and your mic capsule is only measuring a very very small fraction of this. When you have a linear and even level on-axis response then this means in almost all cases with a general compression driver that you radiate your lower frequencies into a larger space because the horn is loosing directivity control at a certain frequency when the corresponding dimension is in the range of one wave length and this might happen in your horn.
You say that in exponential horns loading is improved to higher frequencies which is not the case. This has less to do with loading and is mainly related how you radiate the sound into space as already described. Pure exponential horns tend to beam towards higher frequencies and therefore tend to concentrate the radiated power into a smaller space along the central forward axis and this is the reason why they have a more "linear" or more on level response but only measured in the central forward axis.
On your home page where you sell the ES290:
https://josephcrowe.com/products/es-290-biradial-wood-horn
you state the following: "The ES-290 takes all that's great about the much loved A-290 and improves many aspects to it's overall performance."
This is simply not the true! Your ES290 has almost nothing in common with the Yuichi and it also improves almost nothing. Furthermore, it makes things worse. You have posted two sketches as comparison overview between your horn and the Yuichi and you simply need to compare the area inside the horn to see that Yuichi is in another league.
Your inlay looks like this:
The small red inlay is your horn (and it already has a rollback!) and it is looking very similar to a small Iwata horn. Although I have to admit that your products look stunning and are very well manufactured my pure technical point of view as designer is that the rest of the material is doing less to the horn performance and is integration waste. DonVK and I we did extensive investigations with respect to round overs and still a 1-2cm round over radius eliminates almost all of the diffraction issues at the outer edges of the horn. And off-axis performance is largely related to the horn profile itself and not the small portion of edge diffractions when treated well with a small round over.
Then you state that the fins in the Yuichi design are "dispersion fins" which is simply not true! This is a completely misleading advertisement of your product and bashes a design that is still way ahead of yours.
What about the vertical profile of the ES290? The vertical profile is the most difficult to get right in such horns. The generous round overs do not much anymore for horn loading or directivity control and the vertical of the ES290 is looking more like a slot radiator to me with a really generous round over. If we could see the vertical profile and resulting polar then I suspect that this is the "miracle" for your on level response as the horn starts to loose vertical control very early and tames the lower frequencies at the measured mic position.
"linear response really a proof that the time domain is fine or that you do not have any resonances/diffraction? "
If the frequency response is linear (free of ripple) then it indicates a trouble free time domain. The inverse is also true.
"They new graph begins with 500Hz where your old graph show some midrange narrow. This is some kind of clever if you want to fool us. Who is familiar also know how to tune these graphs with special settings/constraints to look better or worse."
I've included the ARTA settings window below. What specific settings have embellished the results? Please be very specific. If you'd like I can share the ARTA .pir files so that you can investigate further.
If the frequency response is linear (free of ripple) then it indicates a trouble free time domain. The inverse is also true.
"They new graph begins with 500Hz where your old graph show some midrange narrow. This is some kind of clever if you want to fool us. Who is familiar also know how to tune these graphs with special settings/constraints to look better or worse."
I've included the ARTA settings window below. What specific settings have embellished the results? Please be very specific. If you'd like I can share the ARTA .pir files so that you can investigate further.
This assumes that loudspeakers are predominantly linear phase, unfortunately it's really not that simple..
It's a long talk but here's a seminar on the subject, time well spent:
https://www.acculution.com/single-post/043-loudspeaker-phase-video
I'm specially referring to ripple in the frequency response. It will always show up in loudspeaker time domain measurements (Burst Decay, CSD) as well as a driver's impedance and phase sweep. A flat response is an indication of a trouble free time domain when looking at driver/horn combinations and it's raw frequency response, within it's intended bandwidth.
Minimum phase? The point he is making or trying to is that the time and frequency domains are linked.
"Fourier analysis refers to the mathematical principle that every signal can be represented by the sum of simple trigonometric functions (sine, cosine, etc.). The Fourier analysis enables a transformation of a signal in the time-domain x(t) to a signal in the frequency-domain X(ω), where ω = 2πf. In other words, a Fourier analysis is a mathematical operation for calculating the frequency-domain representation (frequency spectrum) of a signal in the time-domain."
https://link.springer.com/chapter/10.1007/978-3-031-14186-7_5
Whether a time problem can be easily seen in the frequency domain or vice versa depends, as there is a tradeoff in resolution in all plots derived from impulses. As the frequency resolution goes up the time resolution goes down, it is the reason why waterfall and CSD plots can be so variable and difficult to interpret. Burst Decay is a good way of comparing because it is a display in periods, so it fixes the relationship in a sensible way.
https://wiki.cimec.unitn.it/tiki-in...tween time,same time, with arbitrary accuracy.
It is a shortcut to say that if there are no obvious problems in the frequency domain there will be no obvious problems in the time domain.
However that is not always the case. If it were, we could equalise any horn and they'd sound the same.
A horn is not usually a minimum phase device.
When looking at measurements that have been processed with the various different Fourier functions, they must follow the rules. You can use the same data to represent the time or frequency domains so in that sense, what is in one is in the other. The Gabor limit determines how much resolution of either you can see at any one time.
Waterfalls and CSD's were mentioned which use the STFT, an extension of this is the wavelet transform where different time windows are used to represent different frequency ranges in an attempt to give a better view than could be had through a single window.
https://en.wikipedia.org/wiki/Wavelet_transform
Of course this does not perfectly describe all time related aspects of a device or speaker. Excess group delay doesn't change the frequency response, the step response can be wildly different without the frequency response changing.
The word resonance can be the Toole description of it or the wider description of any form of stored energy.
A smooth frequency response does not show that there is no diffraction, but it does show that any diffraction is not causing a peak dip combination.
So I think it is a matter of degree or nuance, because trying to distill a vastly complex subject down to a single line is never going to be all encompassing.
Of course you can mathematically, but inferring generality from a specific case (like a single gated FR magnitude sweep taken at a specific angle) is like looking at the field through a straw.
There is a bit much of that going on and I feel one should perhaps be a bit more cautious and humble towards the complexities involved and don't make too many assumptions based on rudimentary steady state data. You need to look at transient behavior as well, music is after all temporal in nature and we tend to forget that.
There is a bit much of that going on and I feel one should perhaps be a bit more cautious and humble towards the complexities involved and don't make too many assumptions based on rudimentary steady state data. You need to look at transient behavior as well, music is after all temporal in nature and we tend to forget that.
Last edited:
I'm quite confused as an impulse response describes the transient response and is not a "rudimentary steady state measurement". Clearly a single on axis measurement is not enough to characterize a speaker given the subject of the thread it should seem obvious.
@fluid I would like to bring the discussion back to the original thread topic because I think it’s a great idea to revisit the A290. I’ve built and tested many horns and it’s my preference to have low coloration as a “first” priority. I wasn’t implying this was the “right” direction, it’s just that I don’t know how to get 120-170 degree listening window AND low coloration to the point where there is literally no indication of artifacts. If I was designing a revised A290, that would be my goal, based on my own experiences on what sounds good and what I’m not willing to sacrifice in terms of sonics (not that there’s anything wrong with other design goals, which I’m well aware of). From a technical design standpoint I think a revised A290 would need very complex 3D CAD modelling through the throat section, something much more organic than the parallel walls and basic geometric shapes of the existing design.
I would like to be involved and help support the technical design, but I’m torn, on a personal level, if I should be helping at all considering @docali personal attack suggesting I’m trying to trick people by embellishing my test results against other test data. It’s one thing to question test results, it’s another to suggest I’m trying to “fool” people in some nefarious way, especially as a business owner myself. I’m open to criticism and actually enjoy being challenged, but it has to be professional, both in a technical AND personal aspect.
I would like to be involved and help support the technical design, but I’m torn, on a personal level, if I should be helping at all considering @docali personal attack suggesting I’m trying to trick people by embellishing my test results against other test data. It’s one thing to question test results, it’s another to suggest I’m trying to “fool” people in some nefarious way, especially as a business owner myself. I’m open to criticism and actually enjoy being challenged, but it has to be professional, both in a technical AND personal aspect.