2123H vs 2122H in a 4343 monitor

[4313B]

Originally posted by Earl K
Hi Guido

What crossover modifications have you done with your 43xx clones to flatten the 2012s' rising response curve ?

( I've looked through most of your postings but can't find this info if it was ever posted )

thanks <> Earl K

Yeah Guido, have you tried the notch filter yet? If so, what's the scoop?

[Guido]

Originally posted by Giskard
Are you hinting that I should try a 2012 with my 2242's and see what happens?
Frankly, I don't have the physical room for such unless I build it all in the backyard

Yeah! Build it in the backyard

[Guido]

Originally posted by Earl K
Hi Guido

What crossover modifications have you done with your 43xx clones to flatten the 2012s' rising response curve ?

( I've looked through most of your postings but can't find this info if it was ever posted )

thanks <> Earl K

Until last week I did nothing. The result was good but not perfect.

One thing was this peak in the 1kHz region that I know well from former projects. Peaks in that frequency range become disturbing exactly when listening at higher levels.

So I popped into this thread and asked Giskard for that notch posted above.

Also @ Giskard
With the components I had at home I did a fast and dirty notch.
Cap 50 uF Electrolytic (Ugh!)
Resistor 5,6 Ohm
Inductor 0,25mH

The peak is gone but the sound quality suffered (no wonder with that cap). All in all the mids are more ?punching? without that peak.
As soon as I have the better components I will tell you.

IMO the 2012 and the 4343 is a good combination. But be aware that I never compared (A/B'd) them with a 2121 or 2122.

[4313B]

Originally posted by Guido
But be aware that I never compared (A/B'd) them with a 2121 or 2122.

Ignorance is bliss and that's a fact. Nothing can take the place of a direct A/B.

Originally posted by Guido
IMO the 2012 and the 4343 is a good combination.

If it sounds good go with it!

[Earl K]

Hi Guido

Until last week I did nothing. The result was good but not perfect.

Does that mean you used/followed the original 4343 crossover component values ?

If it was my project using a 2012 ( and I may be doing this sooner than later ) I would implement/experiment with some of the LC ratio techniques that JBL utilise to "flatten" the midband response curves of their hightech K2 cone transducers. These K2 woofers , most likely ( I've never seen a published curve ) exhibit some portion of a "rising midband response" such as found in your 2012 .

One of those "techniques" is to effectively lower the turnover frequency by increasing the value of the second pole/element in a second order network. The capacitor values JBL plays around with are often 1.5 to 3 times the classical stated "Butterworth" values. The final value of this capacitor ( for a low pass circuit ) really needs to be determined by empirical tuning methods ( or a very good crossover modelling package ). To understand what I'm talking about , study/analyse the LC values for the S3100, 4348, 4344mkII, S9800. It would seem that all these "deep-gapped" woofers need special midband attenuation to give optimal ( flat passband ) results . The 2012 "fits" into this deep-gapped transducer family.

Another technique used to accomodate single area "peaks" ( if they exist at the top end of the pass-band like your 2012 ) is to employ an 18 db /per octave low pass. The trick here is to set the first pole/element (filtering action) just below the offending peak while playing with the LC values of the second & third pole/elements. This will tune the steepness ( or Q ) of the whole roll-over section of the filter. This tuning/tinkering allows you to decide how much to include/exclude the specific area of the offending peak. A quick analysis of the 4344mkII circuit ( for the 2123 mid-bass ) would indicate that something similar has been employed to avoid some of the peaks as found in the 2123h . Here, the first poll/element seems to be set at around 900 hz.


The 2012 would seem to need both of the above design tricks to be engaged to offer a net "flat" power response - for a net 95 db sensitivity from 300 to 1200 hz .

Do you have crossover modelling sowtware ?

regards <> Earl K

[4313B]

One question Earl? Do you have my phone line tapped?

The staggered poles idea is another solution, which I wasn't going to get into if the notch filter worked. It's definitely nice to see someone really tearing these filters apart and giving them a thorough examination

[Earl K]

Hi Giskard

No phone tap here - though I have relatives in the Cdn security services. I'm sure my border-crossings are well scrutinized.

regards <> Earl K

[Guido]

Originally posted by Earl K
Hi Guido


Does that mean you used/followed the original 4343 crossover component values ?

Do you have crossover modelling sowtware ?

regards <> Earl K

I had the original X-Overs and just plugged the 2012 in.

Result was good as I described except the peak around 1k.

I will be travelling next 2 weeks so I can earlist tell the notch results in 2 to 3 weeks.

No, I don't have x-over modelling sw.

If you do some x-over design to fix compensation for the 2012 please let me participate.

Thanks!

[Earl K]

Hi Guido

Here's something you can look at. This is a response curve from one of my 2012h mids as used in one of my companies small boxes. This is full range pink noise run into the woofer section. The box has an approx. effective volume of .65 cu' with a 110hz tuning.

As you can see, the response curve is actually a lot flatter than the JBL cut sheet would lead one to believe. Of course the original 1985 , 2123h cut sheet shows a straight line - and that's not correct.

You can easily see a 2/3 wide - 3 db high bump in the 800 to 1100 hz area. I haven't yet run sine wave test to see what frequencies this "bump" starts & stops at.

Be aware that this RTA response is taken with the enclosures' protective foam backed - metal screen still in place. The mic was on axis to the center of the 2012.

I see this driver as needing similar treatment as the 2123h as used in the 4344mkII. You might be able to add a third pole or element ( inductor ) to your existing crossover - but since the present values are chosen for a 2 pole design - only trial & error or someone with design software can give you the needed value . If it was me - I'd try a .5 to .7 mh inductor in line ( easy for me to say since I own some ) with the midbass circuit. I haven't done any math on what that "T" pad is doing to the circuit impedance - but I'd be inclined to wire in the new inductor just before the 16 ohm Lpad.

It seems I have the necessary parts on my bench that will allow me to wire up a close approximation ( including the Tpad) of this modified 3143 . I'll let you know when that happens and what the results are .

,> Earl K

[Guido]

Thanks Earl K!

This driver looks good. It isn't as bad as I thought.

But anyway it has that peak at 1 k.

I would appreciate the results from your test.

[Earl K]

Hi Guido

- You're correct - the published JBL response curve doesn't do this driver justice.

Here's a pic of my todays' efforts to see what happens when the second pole ( the capacitor ) in a 2-pole low pass is increased beyond the "text-book" butterworth value.

This picture is taken after I built up the low pass portion of the midbass section of the 3143 crossover. I didn't have the exact LC values but I came very close. Instead of a 1.7 mh inductor - I used my 1.6 mh aircore type.
The capacitor was 22uf instead of the 20uf in the original circuit. The "T" pad
was builtup from 2 x 2.5 plus 30 ohm shunt resistor instead of the twin 2 ohm with 30 ohm shunt. The Lpad was a stock 16 ohm JBL unit.

- I obtained this curve by adding 15uf to the existing 22uf for a total of 37uf.

- As you can see - this method works quite well .

- My next posting will show the 2012h curve with my version of the stock values .

- This curve gives 1K as minus 3.75 db referenced to 500 hz while 1.6k is minus 13.75 - also referenced to 500 hz. The low-pass crossover slope ( between 800 and 1600 hz ) is 13.75 db per octave .

- Running pink noise through this 2012h as I added the additional capacitor was very enlightening. It was very obvious a " lot of hi-mid energy was disappearing as the capacitor was clipped into place .

- This is a "Simple - Stupid" approach to the 900 hz peak. It's the one I'll use if I ever build a 43xx thing and if I don't triamp. If in your shoes - I'd first try adding a 15uf capacitor in parallel to the existing 20uf cap. Give it a listen.

- It may be just the ticket for integraton to a 4343 system .

regards <> Earl K

[Earl K]

Hi Guido

- Here's the curve that shows the resulting peak when I use the 2012h with the mock-up of the midbass , lowpass section of the 3143 xover. The 900 hz peak is quite evident.

It's worth pointing out that the "Q" of your circuit is now a little "off". Just dropping in the 2012h without some other changes seems to result in a bit of sagging in the middle of the passband. This becomes more evident when the Lpad is used a lot. Where do you usually set the Lpad ? ( as a percentage of its' rotation ). Altering the values of the resistors in the "Tpad" may mitigate this effect. I won't know until I get to my local electronics surplus store later this week.

<. Earl K

[Guido]

Hi Earl!

Great work!

This looks very good and very easy. I will try this within the next days.

2 Questions:
What was the peak in dB when you just connected the 2012 to the original x-over mockup?

Where should I connect the additional 15 uF cap? Just parallel to driver or do you think I should dismount the whole xover network?
Maybe you can try with your test installation?


I usually set the LPad middle, means 0dB.

[Earl K]

Hi Guido

Parallel the 15uf capacitor to the existing 20 uf capacitor - it's found in the midbass section of the 3143 crossover.

I can't remember offhand how large that peak was when referenced to 500 hz. It was at least 3 db - I can remeasure later today since this mockup is still on my bench.

regards <> Earl K

[Earl K]

Hi Guido

- The peak was 3 db.

- My mistake ; the additional cap I added to my mockup circuit was 22uf not 15uf. So start with an additonal 18uf cap and then add a little more capacitance on 2uf increments. Tune this by ear. FWIW ; I just threw the 15uf ontop of the 44uf ( for 59uf total ) and it started the rolloff around 600 hz ( this is just mentioned as an indicator of the effects of adding even more capacitance. )

- a word of caution, there's no free-lunch to this exercise. At some point changing the LC ratios like this will result in the creation of an underdamped resonant circuit . As a guideline for how far one can wander away from the path ( of years of proven circuit design ) - look to JBLs' own schematics for guidance . They don't seem to get much beyond doubling the text-book capacitance value.

regards <> Earl K