audio-talk

Apologies if I've missed something obvious here -almost certain as I'm down with covid (again...) & my head's feeling like a kango hammer. But from my current less-than-good status, a couple of thoughts

-Re distortion, system HD plots will always show variation relative to individual driver HD measurements, unless you happen to be taking them under identical conditions. Its main value is confirmatory, to show relative trends, where individual driver HD plots is one of the tools in helping you evaluate the performance of the driver & how best to use it. Individual driver HD testing is usually performed nearfield, typically between about 100mm - 350mm or so depending on details, ideally on either a fixed / consistent voltage, SPL or at a pinch, amplifier setting used consistently for all tests. So you'll always get variation there just from the shift in distance if you do a system HD test of a 'finished' speaker which will typically be further out at 1m or > for the usual reasons.

-I see the SPICE plots above, but since you have the on-baffle measured FR, impedance & phase responses, I'm wondering why you don't seem to be using them in a proper crossover design package that lets you see & model the actual FR, impedance & phase with that data & adjust your filter that way. It's a lot more effective than using a purely electrical circuit modeller.

I'm going to mull over the distortion comments for a bit before I comment. I'm trying to visualise how this could happen with a swept frequency test where there's only one frequency being driven at any time.

Scottmoose wrote: ↑Sun Oct 16, 2022 10:02 am -I see the SPICE plots above, but since you have the on-baffle measured FR, impedance & phase responses, I'm wondering why you don't seem to be using them in a proper crossover design package that lets you see & model the actual FR, impedance & phase with that data & adjust your filter that way. It's a lot more effective than using a purely electrical circuit modeller.

Well.. I haven't got one or used one before, and I was hoping the combination of the driver data, the audio tests and the Spice design would get me there. It has before, with the main system. But you've got a good point. Is there a package you'd recommend that isn't too expensive, and can handle the type of circuit I have on the Woofer, which has the parallel notch ?
It might be good to see what it predicts for the 2.5kHz area, becuase I think this has up/down polar behaviour, the dip seems to fill in if you put the microphone above tweeter level.

Meanwhile here is the latest FR plot from last night, getting closer. i will dig out the Spice LP filter model later.

Distortion measurements of individual drive units are usually taken nearfield to avoid room, baffle diffraction &c. effects corrupting the response. This is standard practice.

Almost all good free crossover design software allows you to model a majority of, (or in many cases every possible) circuit topology; XSim, PCD & VituixCAD are all free and certainly don't have any difficulty modelling a simple notch filter; the first & last are complete freeform and will model any crossover circuit topology you can dream up; the middle is not freeform but within the available ladders can model all but the most complex designs (what you're doing is, as crossovers go, extremely simple, so no issues at all there). VituixCAD is the most complex & the most comprehensive, at the price of a steeper learning curve; if you have the measurements it will model & display on & off axis data in realtime, rather than only displaying one axis at a given moment. Swings & roundabouts; they're all better than just going off electrical slopes.

Aside from their own varying behaviour off-axis with frequency, all filters have off-axis lobes, & for conventional speakers these tend to be most obvious as you move off the design listening axis on the vertical, so I'm not surprised you're finding some shifts in that.

Thanks, I'll have a look at XSim first, it's the one I've most heard mentioned !

The Mk.2 crossover breadboard - now it's on one piece of wood !
Lovely bit of varnished 25mm birch ...

Put better bits of birch ply on my fire.....

Ha ! I have a rather large stock of 'might come in useful' bits.
But I do keep up a steady supply of kindling for my friend Mark who has a wood-burner. I used to regularly bring a bag of bits into work for him, and call him up saying "I have a bag of horn shavings for you, upstairs"

I had chance on Saturday to talk to Chris and get some more measurements done, with his knowledge and microphone. He also showed me a bit about how XSim works.
More to follow, but briefly we can say two things from the measurements :
1. The drop-off above 10kHz that I'm seeing on my plots is an artifact of the microphone, not real.
2. There is still a hump around 3.5kHz in the SPL, so the notch I'm using on the Purifi is not aggressive enough. This was likely causing some harshness on vocals and other (selected) sounds during my demo.

Mark,

PM me and I'll upload your REW file to my cloud storage for you to download.

There's potential with this monitor for sure, and I think an updated crossover will help it to shine.

Chris

Trying to kick-start myself into this again. The Covid is not helping !
I hadn't made the effort to sit down and compare the SPL measurements that Chris took in the open in the Owston room, against my own at home. This should be interesting.
First mine from a week before Owston :
Then from Chris :

Some things are clear -
1. Most of the structure from 50Hz to 1500Hz is just room-dependant, there's hardly any correlation between the two traces.
2. The tweeter level beyond 10kHz is fine - Chris has a more expensive microphone, and i suspected that the generic calibration for my ( oops Nick's ) ECM8000 is not correct for this specific example.
3. one of the few common factors is a small SPL peak at 3.5kHz as suspected.

First, i should mention that the low-pass crossover I showed above ( previous page, Spice ) was not exactly what i brought to Owston. For some reason I had a couple of components wrong or hadn't got round to changing them. The upshot was probably that the 3.5Khz notch wasn't as deep as I intended.

Here is what I tried the day after Owston, and this is now correctly implemented on the hardware.
This was another slight improvement on the sound.

Meanwhile I have installed XSim and started looking at vids on Youtube about how to use it. I'll have a go at creating something on there tomorrow.

I missed your demo at Owston Mark - Is this the direction you're now heading in for your own listening?

No, not just yet, but having this nice Mk.2 test cabinet makes me think that getting Dave to do a second one after he's made Jon's final ones would be a good thing to have in the 'stash' for the future if I have to downsize.
The quality of the sound is close to my main system, in fact the bass extension might be a smidge lower.
The main system still has the edge on 'low level detail' from the all-low-Rms drivers. What you might hear as 'room-filling sound' at low listening levels.