Something stirs in the Undergrowth

[chris661]

How steep is your ribbon XO?

They do like to distort if you push them too low...

[IslandPink]

Hi Chris. Yes, I've heard about that problem. The crossover I have now is 1st-order HP at 5kHz, but increasing to 2nd-order below 2kHz . I figured that if the G3 can be run to 60W using a 3rd or 4th-order at 2.3kHz, I should be OK with the above.
What I can do is check later today, listen to some music with the Fostex disconnected , just see how the ribbon sounds .
I have noticed that the residual problems are mostly present when dead on-axis to the Fostex . I'm thinking it's quite peaky ( and you can see this in the curves ) in the region from 4kHz to 10kHz but very directional . A few degrees off axis and this improves noticeably .

[g3dahl]

Your crossover is probably fine as far as driver protection is concerned, but I don't think it is steep enough for optimum sonics with the G3. If you don't get the sonics you're hoping for, I would suggest a steeper filter for the ribbon.

Gary Dahl

[IslandPink]

OK, well, I started from a crossover with a 4th-order HP on the ribbon at 3.5kHz, and it sounded poor . There was no integration of the ribbon to the mid-driver.
I just listened to '21st Century Schizoid Man' by King Crimson on the ribbon ( with its HP crossover ) and it sounded clean and trouble-free . Just lots of fine detail and tone on snare drums, cymbals and triangles in the mix .
Listening to the Fostex , alone, with its crossover, there's a bit more that could be tweaked . Some of it will be improved a lot by going to the perspex baffle. The rest is just a case of doing some measurements and adjusting the depth of the notch filter and the exact point of the 1st-order LP ; perhaps a zobel too if the >5k is still up too much in the mix .

Can't reply soon on anything - I'm away for a week now - cheerio !

[IslandPink]

Just been checking the two drivers with a small inductance/capacitance meter . Fostex shows 1mH ( datasheet 0.05mH ) , the G3 ribbon shows 4.5mH
I wonder if i should believe ( roughly ) these numbers ?
I think the measurement frequency is about 250Hz - yes, just checked .

[Paul Barker]

the question is what is the capability of the tester at 1mH level?

Take a standard ohm meter for instance. Short the leeds and it may show 1 or 2 ohms.

top quality kit you can zero with leeds shorted, then do the reading knowing you have a zero point to start from.

Might be similar issues with very low inductance you are trying to find.

the manufacturer can be absolutely sure about the inductance he specifies because it is a given. He knows the diamater of the voice coil and he knows the number of turns. He should know the perm of his pole piece, and it shouldn't be so far out as you are finding. The inductance follows without question. no need to measure it.

So you have two possible scenarios. The manufacturer has misprinted the spec, or your meter is not accurate at this level.

[IslandPink]

It's as accurate as you could want on known inductors from 0.22mH upwards - seems spot-on . Just not quite sure if speaker measurements are frequency-dependant or if some other issues are there. I think the ribbon will need a zobel as 4.5mH is quite a lot - and I think the RAAL ones ( only ones I could find data on ) are eg. 2mH , so I think it may be a real measurement .
I'll just check the Eminence Delatlite II 2515 .... hmm... measures 0.57mH ... datasheet says 1.15mH
Not sure now . I'll check what Dick Olsher used on his crossovers for the G2Si . Think I had 12R + 0.1uF on his crossover .

[IslandPink]

OK, some progress yesterday and today .
First, a pic of the 'Pinkzilla_ver1.0' crossover. I now have this attached to the pair of drivers on the perspex baffle. It's sounding pretty nice, but some of the same niggles I heard on the other baffle are still there .

So, today, I managed to fit a proper 16-bit / 48kHz soundcard and get that working so I can see beyond 8kHz .
With AudioTester , I then did a number of white noise measurements , from a distance of 1.7m ( arbitrary - best I could rig-up quickly ! ) . Overall, I'm pretty pleased - not bad for a few guesses and some tweaks by ear ! ( levels vary on these, as I was adjusting the pre-amp level during testing to find the max level possible without clipping ) .

First plot is for axis, and second plot is for about 10 degrees off-axis .

More to follow .

[brig001]

Just been checking the two drivers with a small inductance/capacitance meter . Fostex shows 1mH ( datasheet 0.05mH ) , the G3 ribbon shows 4.5mH

I don't think you can measure voice coil inductance with the voice coil moving as the back EMF will (I suspect, because I'm making this up) make the reading higher as it will reduce the current. I assume that when you are testing that you can hear the test frequency through the speaker.

Brian

[chris661]

FWIW, the driver inductance is usually measured at 1kHz.


I suspect the driver's electrical and mechanical inductance sum to give a higher total. Some drivers have inductance values at 1kHz and 10kHz, which would (I expect) be identical if the inductance of only the coil was being considered.

Chris

[IslandPink]

Thanks for the comments . I had misgivings about the cone/voicecoil moving myself, Brian . Probably is an issue.
Chris - I'm not aware of 'mechanical inductance' . Are you perhaps thinking about the damping coefficients Qts, Qes and Qms ?

More plots shown below . These are from the Fostex in-situ on the baffle , with its crossover, but the tweeter is disconnected. I took the trouble to put a 10R resistor in place of the tweeter in case it had any effect on the crossover and hence the Fostex .
Note the fairly limited effect of the Low-pass part ( a 0.22mH coil ) which doesn't do much to knock down the 7kHz peak . I really need a 0.33mH part in there to put the roll-off at 4K instead of 6K as it is at present .

Second picture : Fostex as before but 10 deg off-axis. Note that the 7kHz peak drops a bit, and the 14kHz peak drops a fair bit more . This ties in with the listening impressions . I suspect these peaks are pretty directional and from the centre of the cone and dustcap .

The other point I've gleaned from these plots is that the notch filter at ~4kHz is possibly doing too much . This is good news , as a shallower notch has less phase shift, which ought to further improve the tone. I just need to reduce the resistor value on the LCR-group to 3.3R or 3.9R instead of 4.7R . This was a surprise, as I calc-ed it in Spice using the Fostex FR curve as a guide, thinking it might not be enough. It suggests the uncorrected peak at 3-5k shown on their graphs is not as bad as they show .

[IslandPink]

Here's the tweeter in a similar way : in-situ on the baffle ; using its part of the crossover ; and the Fostex disconnected but replaced by a 10R resistor .
This could do with re-doing, along with the Fostex-alone plots, to get the levels higher . Will do this next time i get all the kit out . On this plot you can see it's going down into the noise floor around 1kHz .
Anyway, looks OK - about what I expected . I will probably bring the HP1st at 5 or 6kHz down a bit lower once I've altered the Fostex as discussed above.

[IslandPink]

Based on some of that data, I've been adjusting the crossovers slightly .
From what I can tell from the overall response, there's a slow rise of 3-4dB going from 1kHz to eg. 5kHz making it too bright . So , I reckon the Fostex needs to start its low-pass a bit earlier , and the overall tweeter level needs to come down .
I was also not sure if the notch at 4kHz on the Fostex was a bit too aggressive, based on the Fostex-alone trace, so I tried making that less .
Results were that the tweeter level decrease and the lower onset of LP for the Fostex worked . Decreasing the 'notch' at 4kHz on the Fostez was not helpful ( to the ears ) so I went back to 4.7R on that .
I did a fair bit of listening this morning, using Joni Mitchell and Joan Armatrading for reference. I tried increasing the tweeter cap to 2.42uF to bring the HP lower on the G3, but it started to sound too forward , so I went back to 2.2uF on that one .
If I had more strength ( hampered badly by a chest-cold at the moment ) I would get the drivers off the baffle and rebate the tweeter cut-out to get it flush to the surface ( and also aid the time-alignment, I think ) but that will have to wait .

So, pic below shows current crossover for the Fostex - 'rev2' .

[IslandPink]

First test of three-way using bass pre-amp to feed Pass F4 for Eminence 15" bass speaker . 300B SE feeding crossover and mid/HF units.
I can't help thinking there must be a more elegant way of doing this, see below :

[IslandPink]

Although everything sort-of works ; pre-amp has enough gain , not much hum, low bass comes up due to the filter elements ; it doesn't sound much good
Overall the bass speaker doesn't add much if anything to the music. There's not much bass tone or 'snap' .; not much integration to the midrange .
I have tried both polarities of connection to the bass, and it sounds better ( as expected ) with the bass driver in reverse compared to the Fostex .

So, I feel a bit 'stalled' at the moment . Motivation is down .
I'm wonder if it's the driver or the equalisation ? ( or both )

I have looked at the phase shift from the baffle roll-off ( in Hornresp ) , and that from the low-pass element in the pre-amp ( in LT Spice ) . In the sign convention of the two pieces of software , both show increasing phase rotation ( more positive ) as you go down in freq : about +60deg and +90deg from 200Hz to 40Hz from the two contributions. If these are going the same way then that would be nearly 180 degrees mismatch from 200 to 40Hz, which would explain the problems . I was kinda-hoping that the two would cancel out ...