James
Lord of the Erg\o/s
I had been disturbed by the marginally imperfect performance of the E-IIIs for a while now. It's not wrong sounding by any stretch of the imagination; just not as sublime as the E-VIIs and E-IVs in the midband. It sounds good enough to be enjoyable, but it does not make you sit up and pay attention like a live band in your living room might. OK, so I have high aspirations.
When Mr Tibbs built the E-IIIs, I acquired a pair of Seas M15CH001 pure mids that are supposed to be one of the best drivers on the planet in its class. Not uncoincidentally, it shares the same chassis as the CA15RLY midbass that I designed the E-IIIs with. So it's a doddle to swap the drivers physically. Electrically, it's a bit more complicated as they have quite different T/S parameters and a somewhat different frequency response. The only real way of incorporating them into an existing crossover design is to start from scratch.
I never got around to doing that because the E-IVs, E-V, PFM-Specials, E-VII, E-VIII and now the E-IX got in the way. I had always meant to take another set of measurements to take another stab, but you know about good intentions. Anyway, Mr Tibbs' reinstatement of and wonderment in his old Epos ES14s got me doubly motivated. So, whilst taking measurements for the E-IX, I took the time to take a fresh set of measurements - both of the original set-up, and of the new with M15CH001 in place.
To understand how drivers crossover, it is important to consider the phase response of each overlapping driver. For even-order crossovers, the amplitude sums correctly only if the phase is exactly the same. In practice, a number of things conspire against this. Not the least of which is how accurately the driver is following a symmetrical transfer function acoustically, which is a combined result of the native response of the driver and the electrical effect of the filter. This is where actual measured data and a good simulation software comes in rather handily.
Let's have a look at the original E-III phase response.
As you can see, they don't line up, but they do track in the same direction and have the same relative difference across the crossover ranges' midpoints of 350Hz and 3.5kHz. This isn't half-bad. At least they will sum consistently if not 100% accurately. But what they won't do is sound like a seamless trio of drivers.
With the M15CH001 midrange in place, I remodelled the crossover and got this instead.
Here, the midrange phase and woofer phase are almost perfectly in line. They should sound as coherent as a single driver, and much more so compared to the original iteration. The real difference is I've learnt a lot more about crossovers since I built the E-IIIs. The revised XO includes resonance traps for both the midrange and tweeter so that they behave properly in their stop-bands. Even though the M15CH001 crosses to the woofer at 350Hz, and almost 2 octaves above its resonance frequency, the latter was sufficient to produce a kink in the transfer function and more than enough to throw the phase askew.
The resulting frequency response is also telling. Here is the original E-IIIs simulated summed response.
The revised E-IIIs compares more than favourably, don't you think?
There is only one downside to this story so far. The M15CH001 is no longer available at the usual outlets, and so unless Mr Tibbs happens upon a pair on eBay, this is really more an exercise to try prove that 3-way designs could work as well as a two way. But fear not, Mr Tibbs. I can redesign the E-III using the CA15RLY driver using the same techniques.
Interested?
James
When Mr Tibbs built the E-IIIs, I acquired a pair of Seas M15CH001 pure mids that are supposed to be one of the best drivers on the planet in its class. Not uncoincidentally, it shares the same chassis as the CA15RLY midbass that I designed the E-IIIs with. So it's a doddle to swap the drivers physically. Electrically, it's a bit more complicated as they have quite different T/S parameters and a somewhat different frequency response. The only real way of incorporating them into an existing crossover design is to start from scratch.
I never got around to doing that because the E-IVs, E-V, PFM-Specials, E-VII, E-VIII and now the E-IX got in the way. I had always meant to take another set of measurements to take another stab, but you know about good intentions. Anyway, Mr Tibbs' reinstatement of and wonderment in his old Epos ES14s got me doubly motivated. So, whilst taking measurements for the E-IX, I took the time to take a fresh set of measurements - both of the original set-up, and of the new with M15CH001 in place.
To understand how drivers crossover, it is important to consider the phase response of each overlapping driver. For even-order crossovers, the amplitude sums correctly only if the phase is exactly the same. In practice, a number of things conspire against this. Not the least of which is how accurately the driver is following a symmetrical transfer function acoustically, which is a combined result of the native response of the driver and the electrical effect of the filter. This is where actual measured data and a good simulation software comes in rather handily.
Let's have a look at the original E-III phase response.
As you can see, they don't line up, but they do track in the same direction and have the same relative difference across the crossover ranges' midpoints of 350Hz and 3.5kHz. This isn't half-bad. At least they will sum consistently if not 100% accurately. But what they won't do is sound like a seamless trio of drivers.
With the M15CH001 midrange in place, I remodelled the crossover and got this instead.
Here, the midrange phase and woofer phase are almost perfectly in line. They should sound as coherent as a single driver, and much more so compared to the original iteration. The real difference is I've learnt a lot more about crossovers since I built the E-IIIs. The revised XO includes resonance traps for both the midrange and tweeter so that they behave properly in their stop-bands. Even though the M15CH001 crosses to the woofer at 350Hz, and almost 2 octaves above its resonance frequency, the latter was sufficient to produce a kink in the transfer function and more than enough to throw the phase askew.
The resulting frequency response is also telling. Here is the original E-IIIs simulated summed response.
The revised E-IIIs compares more than favourably, don't you think?
There is only one downside to this story so far. The M15CH001 is no longer available at the usual outlets, and so unless Mr Tibbs happens upon a pair on eBay, this is really more an exercise to try prove that 3-way designs could work as well as a two way. But fear not, Mr Tibbs. I can redesign the E-III using the CA15RLY driver using the same techniques.
Interested?
James
