The "Elsinore Project" Thread

The cap has a reactance curve. Microphonics created by the AC current passing through the cap causes the reactance to change, depending on the construction and materials used (obviously mass and damping as well as no air gaps etc, these all come into play). As it is in series with the voice coil, then we have some kind AM, amplitude modulations, of the current. This ends up in the voice coil, and hence we hear it. Different caps will not all behave them same. Cause and effect would seem to exclude other explanations. What do you think?
Capacitive reactance is the variable resistance of a capacitor being controlled by the applied frequency.

Microphonics describes the phenomenon where a component transforms mechanical vibrations into an electrical signal, the reverse would make the capacitor make noise.

I think if either of those phenomenon have an audible effect in the output of a passive crossover, something is wrong with the circuit or components chosen for it 😉


https://www.diyaudio.com/community/threads/capacitor-microphonics-question.384835/#post-6986865

Art
 
Concerning capacitor noise, it is interesting that the most highly regarded capacitors (Duelund, Miflex, and even for that matter the Audyn True Copper Max) are potted withing a rigid fiberboard tube and use dense foil plates (copper, even silver for the Duelund). The construction is optimized against vibration. It makes me wonder if as a DIY effort potting less expensive but well regarded tape covered capacitors would have a significant benefit. That may be an experiment for me to try, I have some good sounding CDE 942 series hybrid film/foil caps I use for coupling in tube amps, and it would be nice to see if they can be improved cheaply.

Anyway, any verdict yet on the Miflex caps in the ULD tweeter crossover?

Thanks.

David
 
Anyway, any verdict yet on the Miflex caps in the ULD tweeter crossover?

Works great!!!

The air I am hearing on this track link below, if you have Tidal, at the very beginning, it is so obvious and it in on the percussion and the harmonics, not some obvious treble sound. The improvement is wider in frequency than you would be expected handled by a treble driver. But I stress this is what I hear in the context of the system I have here.

https://tidal.com/browse/track/25029966?u

Hope the link works for you guys, one half of Yello is behind this, Boris Blank.

1719819921948.png
 
Folks:

Joe's experiments with various copper caps in the Elsinore's tweeter crossover section has been fascinating and I purchased a pair of Miflex KPCU-03 1.8uF caps based on his assessment. I also purchased a pair of KPCU-01 0.027uF caps for bypass purposes. The Miflex caps were installed in my Elsinores last week, replacing Jantzen Superior Z and Cornell Dubilier 940C 0.01uF 3000V bypass caps, and were burned in with a FryBaby2 at moderate volume for 5 days. I've been listening to the upgrade now for a couple of days and am more than a little surprised by the improvement. There is increased clarity, air and precision in the upper frequencies. Spatial cues, which were already excellent, have also improved. The treble is both lighter and more realistic. What was magical is now even more so.

If you're contemplating this upgrade, check out pricing at Miflex's website ( audiomiflex.pl ). My usual sources here in North America (partsconnexion and soniccraft) sell Miflex caps but not the values needed, and while HiFiCollective had them, they were more expensive.

Following in the footsteps of experts is sometimes especially rewarding. Thank you, thank you, Joe!

Regards,
Scott
 
I've been listening to the upgrade now for a couple of days and am more than a little surprised by the improvement.

Yep. Hear those same improvements pretty much straight away.

If you're contemplating this upgrade, check out pricing at Miflex's website ( audiomiflex.pl )

Alas, they sell to Europe and the US, but not Australia. I don't need any now, but chances are that I will at some stage. And yes, 1.8uF is not as often covered compared to 1.5uF and 2.2uF.

I just checked out audiomiflex.pl pricing. Has it gone up?
 
Joe:

Maybe, but the pricing on the audiomiflex.pl website is what I paid. Their price for the KPCU-03 1.8uF capacitor is 90.58 euros; hificollective's price is 111.23 British pounds (or 132.23 euros, per xe.com). In any event, a bargain at that price.

Regards,
Scott
 
Progress report on my ULD build. Just finished with cutting and routing out all the panels. The cabinet is 1" bamboo plywood, except for the front baffle, which is 3/4" bamboo plywood. The sub front panel is 24mm baltic birch plywood (2-12mm panels glued together in a press) as well as the tweeter stiffening panel. The main brace is a laminate of 6mm baltic birch ply on the outer surfaces and a 1/2" MDF core. I was actually surprised that an unscientific rap on this composite (identical sized) was a higher pitch than 24mm baltic birch. I didn't weigh it, so don't know if it is stiffer or just less massive.

I have a little work to do prior to starting assembly. On the East Coast of the US, the most practical and cost effective way of getting 1" bamboo plywood are pre-finished desktop/countertops, which ships free to a home improvement stores (Lowe's) as easily to handle 60"x30" panels. They have a poly varnish, so need to strip off the finish for glue-up. The shipping cost for full sheet (8'x4') panels is several hundred dollars on top of the $200 to $300 cost, and since they have to be transported via truck, I am responsible to get them off the truck at delivery, which is more than I can handle Yes, I probably could have used BB, but even that is expensive and difficult to find now days given the Russian embargo. I wanted to use a superior panel material after my experience building MDF cabinets on the MFC build a few years ago.

Anyway, picture of the parts.

David
 

Attachments

  • 20240715_120009.jpg
    20240715_120009.jpg
    726.4 KB · Views: 96
Scott,

I have at least one taker for the speakers...I built him a pair of my 211 A2 SET amps (25W) and he really loved the sound of my Elsinores with the amps (as has all who have been at my place and have heard them together). He also lives 4 miles away, so it will be easy to take them over, but figure if I get parts cost out of them, it will be fair for both of us.

I notice that you very much favor the Pass "First Watt" designs on your Elsinores. I have also experimented in that direction, and am planning on an dual mono F4 build after the ULDs are finished (I have a 6SN7 Aikido preamp in my stash that should work well in driving it). I built a dual mono version for a friend of mine at his request and he loves it. I built a stereo F4 (not with the Elsinores), that morphed into a BA3, as well a tube front end to the BA3 (Moskido), but still prefer the sound of my tube amps that have enough gain that I can run passive or DAC directly into them. Maybe using the F4 with an active tube pre will convince me otherwise.

David
 
David:

A friend of mine uses balanced F4 mono blocks with his Elsinores to great effect. I don't recall what he uses as a preamp (perhaps one of Jim Hagerman's designs?), but with good quality material his system is very compelling.

Regards,
Scott
 
Scott,

Good to know. 100W Class A should be plenty dynamic. Something I have been tempted to explore, my sources have balanced output capability, just need to build a balanced tube preamp in addition to the amps. Building solid state amps is much cheaper than big SET amps, that is for sure. A pair of Tamura 10K/40w output transformers costs me $2000 from Japan, and the rest of the parts come in for about the same.

Well, back to work on the new Elsinores.

David
 
In this case a sixth order rolloff for reduced excursion, and a third (or forth) order fitting phase response.
Where the roll-off is steepest, it's 26dB/octave, and then for about 100Hz or so. That's 10dB/octave short of 6th order. There also seems to be a bit of a downside in that at below about 600Hz the filter network is providing a lot less attenuation than the desired target function response. Maybe it would have been better to include a twin-slope approach to the target function, with an initial 3rd-order/4-order fitting response, followed by a 5th-order/6th-order low-frequency fitting response below some chosen frequency. The total component count would likely have been similar.
 
I went back and checked it by drawing 36dB/oct slopes on the plots, the one on the left is the model and the one on the right is measured.

I agree that it's not a precise copy. Some areas where there is room for deviation.. the notch filter doesn't have orders, the resistance of the inductor and vituixcad smoothing. I took the slope asymptote further down in frequency than you appear to have, as it's where the slope stabilises.

In any case this is not what was being demonstrated. As I'm going to point out on the other thread, the tweeter has an standing 12dB/oct slope before the crossover is added and that resumes where it deviates below 700Hz.

Screenshot from 2024-07-17 13-55-32.png
 
I misread the graph, and you are quite right – the slope is 6th-order over a large part of its range. Once the magnitude response starts diving into the notch, the slope becomes even greater, about 8.5th/order. In any case, the target response is a 6th-order high-pass Bessel response, and the simulated response lines up with that over most of the range.
 
Last edited:
Here's a simulation of the response obtained with the circuit including the notch filter leg, after some optimization to compute suitable component values for an 8-ohm resistive load. The target response was a 6th-order Bessel high-pass filter with a cutoff frequency of 2450Hz.

1721195627063.png


1721195609193.png


Adding an extra capacitor produces a markedly better fit to the target 6th-order Bessel high-pass response. As the magnitude response is matched better, it might be expected that the phase response inherent in the chosen target function is also more closely matched.

1721195933181.png


1721196030915.png
 
Last edited:
I've added a 2nd-order high-pass response function for the tweeter, with a −3dB cut-off at 1000Hz to the model in the previous post. The circuit and the optimized response that was obtained are shown below.

1721200386385.png

1721200406328.png


Adding in an extra capacitor produces the following result, which is a better match to the target function. Note that the notch is now much less sharp, and is also much lower in acoustic output level.

1721200986873.png

1721201012386.png