JPW AP3 Crossover Upgrade

[istari_knight]

Both AP3's & Type K's are of very similar overall design but differ in two areas, the cabinet dimensions & the crossover... Both have the same internal volume but JPW abandoned the "golden ratio" used by Snell for a more aesthetically pleasing look [IMO.] I cant do anything about that but I can change the crossovers... I was intending to do a long write up about the why's & how's but I cant really see the point now... Suffice to say, the Snell crossover design is superior to that employed by JPW so replacement should yield some good improvements.

Now, I've spent an age hunting down the values used used in the original Type K crossover & its difficult to nail down as they tweaked the values to suit each individual drive unit but we can average out which is what I decided to do :eyebrows:

This is what I've come up with:

http://farm4.staticflickr.com/3809/11175091795_51db6ae494.jpg (http://www.flickr.com/photos/87920837@N06/11175091795/)

Snell used air-core inductors for both HF & LF drivers, bipolar electrolytic caps bypassed with small value films & a 5R variable resistor which was usually set between 1R6-1R8.

This is what I'm putting together:

http://farm6.staticflickr.com/5543/11175769994_fb46344c67.jpg (http://www.flickr.com/photos/87920837@N06/11175769994/)

It does differ from the original slightly... I am exclusively using polypropylene caps, I have kept the HF inductor an air core but decided to opt for a cored variant on the LF mainly as I'm diving into the unknown here so not really willing to throw £30 away on inductors that might not be an improvement over the £5 one's... My thinking is if I like the results I can always upgrade these in time. I've also opted for a fixed value resistor in place of the variable jobbies which IME are invariably crap after a few years use.

Thats it for now until the bits arrive... Here's a picture of the original crossover's currently fitted to the JPW's for comparison :)

http://farm4.staticflickr.com/3686/11175141536_04fd6ec47f.jpg (http://www.flickr.com/photos/87920837@N06/11175141536/)

[Reffc]

Any reason for using a series resistor in place of an L-pad James? You get the advantage of consistent impedance using the L-pad but not with the series resistor. I'd be tempted to put a zobel in the bass after the filter to flatten out the impedance rise too.

Snell used various crossovers for their various incarnations of the K (2nd order for the Mk1 and 3rd order for the Mk2 for flatter off axis response after the drivers used changed) but that is not to say they can't be improved upon as they were fairly basic in their Mk1 guise. By impedance flattening and matching tweeter to woofer impedance, you may get a better response.

[istari_knight]

I hadn't considered it to be honest... I've not read of any complaints regarding the Snells performance [not the mk1's anyway] so was just aiming to replicate that but if the L-pad & zobel will improve them further its a no brainer !

I'll keep it as planned for the minute but will definitely have a play with your suggestions at some point, this little project is actually a test... If I like the results I'll be building some proper "K clones" from scratch for myself, I suspect your suggestions will come into play then & they can be "improved K's" :eyebrows:

[Radford Revival]

You'll definitely need an adjustable L-pad on the tweeter instead of that variable resistor - otherwise the crossover will only work correctly in one position of the control!

Edit: Never mind, I see you're using a fixed resistor, I haven't had my coffee yet!

[istari_knight]

Hehe I did originally consider using the same part as Snell, the exact ones can still be purchased here: http://www.digikey.co.uk/product-detail/en/D25K5R0E/D25K5R0E-ND/823538?WT.mc_id=PLA_UK_823538

[brian2957]

Wish I could do things like this James . I recently had Paul at Reffc repair and refurbish the crossovers on my Rega EL8s and it has totally transformed them :D. Hope you get the same result :cool:

[istari_knight]

Cheers Brian, here's hoping !

The bits arrived this morning from audio-components.co.uk who when it was realised one of the inductor value's I ordered were out of stock upgraded to the next size up [AWG] FOC... Now that's what I call service !

A quick "mock up" ... Dont they look empty ?! At least everything will be nice & separate :eyebrows:

...Now just need to find my drill bits :scratch:

http://farm6.staticflickr.com/5486/11251200274_90d0ae5212_b.jpg (http://www.flickr.com/photos/87920837@N06/11251200274/)

[system7]

Interesting thread, considering what a legend that old Vifa M21WG-09-08 is.

http://www.wilmslow-audio.co.uk/vifa-m21wg-09-08-701-p.asp

But there is a glaring error in your JPW AP3 schematic:

http://farm4.staticflickr.com/3686/11175141536_04fd6ec47f.jpg

The coil must come after the capacitor in the treble filter, or it's a dead short at low frequency. FWIW, your Snell Type K circuit is a lowish crossover around 2kHz.

[istari_knight]

Interesting thread, considering what a legend that old Vifa M21WG-09-08 is.

http://www.wilmslow-audio.co.uk/vifa-m21wg-09-08-701-p.asp

But there is a glaring error in your JPW AP3 schematic:

http://farm4.staticflickr.com/3686/11175141536_04fd6ec47f.jpg

The coil must come after the capacitor in the treble filter, or it's a dead short at low frequency. FWIW, your Snell Type K circuit is a lowish crossover around 2kHz.

You are quite right Steve, not sure how I missed that :scratch:

The Snell crossover was compiled using averaged component values taken from 6 different Type K [mk1] crossovers... The examples I saw didn't vary that much anyway [within 10% +/-]

[istari_knight]

:D

http://farm4.staticflickr.com/3787/11254159816_c3d00b4b8a.jpg (http://www.flickr.com/photos/87920837@N06/11254159816/)

[system7]

I have done a lot of work on 8" bass plus 1" tweeter. In fact I've got the best results where I've used cone tweeters, and given the bass some second order rolloff with the crossover. 8" bass has a LOT of harsh cone breakup above 3kHz, and taming it is the issue. It's quite hard to get good phase alignment with 8" bass. Here I used the old KEF trick of flipping tweeter polarity on a third order XO of about 3.5kHz. Decent sounding speakers these. They (Sony E44) originally just had a simplistic 3.3uF on the tweeter. :)

http://imageshack.us/a/img51/2661/k4su.jpg

http://imageshack.us/a/img46/8922/sggh.jpg

http://imageshack.us/a/img40/4937/x9cz.jpg

http://imageshack.us/a/img845/3347/e3g1.png

The 10 ohm resistor on the bass shunt is a moveable feast. It can be smaller with lower inductance woofers like yours. It stops the midrange being peaky.

[Reffc]

If you can put a multimeter across your woofer James and PM me the DC resistance, I can supply you with a simple Zobel circuit for the woofer. This keeps impedance flat to the crossover point. You can't really use the stated nominal impedance for crossover design although it's surprising just how many manufacturers have! If the drivers are original, the equivalent coil inductance should be about 0.9mH.

Something doesn't look quite right with those crossover values either James. The D19 was a nominally 8 Ohm tweteer but impedance between resonance and rising curve after resonance was closer to 6 or 7 Ohms. With no zobel to flatten the rising impedance curve, the value at crossover point corresponding to your values is roughly 8 Ohms if the crossover is at 2,100 Hz

If you look at the value for the woofer, it's miles too high to be anywhere close to 2100Hz. In fact, 1mH will start rolling the woofer off at 6dB/Octave from 900Hz upwards (ie, by 2100 Hz, response will be roughly 7 or 8dB down on peak).

Looking at the curves for each drive unit, this puts the bass unit quite a few dB down on the tweeter by 2100 Hz, hence the resistor in the tweeter network which brings down tweeter response at crossover by roughly 3.5 to 4dB. However, the tweeter starts rolling off very steeply after 10KHz so you may want to reduce that resistance value slightly (suck it and see). The reason for the woofer being rolled off so early is probably due to avoiding the cone breakup resonance point so why they didn't use a 2nd order to achieve this is beyond me.

If it were me, I'd be using a 2nd order network with impedance flattening for the woofer to avoid any significant woofer output at resonance and either eliminating that resistor or reducing the value. remember that the tweeter is an 89dB sensitive model whilst the woofer is 90dB minus a little bit for the impedance of the inductor, so both units should be well enough matched.

[Reffc]

I have done a lot of work on 8" bass plus 1" tweeter. In fact I've got the best results where I've used cone tweeters, and given the bass some second order rolloff with the crossover. 8" bass has a LOT of harsh cone breakup above 3kHz, and taming it is the issue. It's quite hard to get good phase alignment with 8" bass. Here I used the old KEF trick of flipping tweeter polarity on a third order XO of about 3.5kHz. Decent sounding speakers these. They (Sony E44) originally just had a simplistic 3.3uF on the tweeter. :)

http://imageshack.us/a/img51/2661/k4su.jpg

http://imageshack.us/a/img46/8922/sggh.jpg

http://imageshack.us/a/img40/4937/x9cz.jpg

http://imageshack.us/a/img845/3347/e3g1.png

The 10 ohm resistor on the bass shunt is a moveable feast. It can be smaller with lower inductance woofers like yours. It stops the midrange being peaky.

Some valid points on cone breakup (which is why 1st order is a bad idea imho) but a 2nd order (bass plus treble) is 180 degrees out of phase, so should be perfectly phase aligned by simply reversing polarity as you've pointed out. The bass circuit would be better with the zobel after the filter (ie calculate the proper values for C and R based upon DCR of the driver and it's inductance). You take the guesswork out of the picture then and there's no moveable feast, it just works. By going 3rd order on the tweeter, and 2nd order on the woofer, you've also introduced 90 degrees of phase difference. Leaving the tweeter 2nd order and keeping polarity reversed should do the trick though.

[istari_knight]

If you can put a multimeter across your woofer James and PM me the DC resistance, I can supply you with a simple Zobel circuit for the woofer. This keeps impedance flat to the crossover point. You can't really use the stated nominal impedance for crossover design although it's surprising just how many manufacturers have! If the drivers are original, the equivalent coil inductance should be about 0.9mH.

Something doesn't look quite right with those crossover values either James. The D19 was a nominally 8 Ohm tweteer but impedance between resonance and rising curve after resonance was closer to 6 or 7 Ohms. With no zobel to flatten the rising impedance curve, the value at crossover point corresponding to your values is roughly 8 Ohms if the crossover is at 2,100 Hz

If you look at the value for the woofer, it's miles too high to be anywhere close to 2100Hz. In fact, 1mH will start rolling the woofer off at 6dB/Octave from 900Hz upwards (ie, by 2100 Hz, response will be roughly 7 or 8dB down on peak).

Looking at the curves for each drive unit, this puts the bass unit quite a few dB down on the tweeter by 2100 Hz, hence the resistor in the tweeter network which brings down tweeter response at crossover by roughly 3.5 to 4dB. However, the tweeter starts rolling off very steeply after 10KHz so you may want to reduce that resistance value slightly (suck it and see). The reason for the woofer being rolled off so early is probably due to avoiding the cone breakup resonance point so why they didn't use a 2nd order to achieve this is beyond me.

If it were me, I'd be using a 2nd order network with impedance flattening for the woofer to avoid any significant woofer output at resonance and either eliminating that resistor or reducing the value. remember that the tweeter is an 89dB sensitive model whilst the woofer is 90dB minus a little bit for the impedance of the inductor, so both units should be well enough matched.

Thats alright thanks Paul, as said before I'm just trying to replicate the Type K with this project... I measured the original JPW crossover's myself & can confirm the values are as stated however strange they look.

FWIW the woofers measure 5R6 DCR & coil inductance is 1.3mH :)

[istari_knight]

Two new crossovers all wired up, tested & ready to go...

http://farm3.staticflickr.com/2836/11258651144_5ed53f1e70_b.jpg (http://www.flickr.com/photos/87920837@N06/11258651144/)

[synsei]

Two new crossovers all wired up, tested & ready to go...

http://farm3.staticflickr.com/2836/11258651144_5ed53f1e70_b.jpg (http://www.flickr.com/photos/87920837@N06/11258651144/)

Wow James they are beautiful, and I don't get to say that too often in regards to crossovers... ;)

[istari_knight]

Thanks Dave, that's the beauty of using acrylic... It makes everything look good ! Well everything's back together & they're making music. They certainly sound different :eek:

Thats all for today, too late to do any proper listening :(

[Reffc]

Very neatly done James.

[brian2957]

How are the speakers sounding now James ?

[istari_knight]

How are the speakers sounding now James ?

Meh :lol:

My "close enough is good enough" method isn't good enough. With the new crossover's in place the bass gained weight & authority that was previously missing so an obvious improvement there, everything else is hard to judge as it sounds disjointed & altogether not right... Like the tweeter is at least 6dB down. I've tried bypassing the resistor & reversing the tweeters polarity without much success... I've run out of time to play as I've got people waiting to have their amps repaired so I reinstated the original crossovers & normality is restored... For now :eyebrows:

[brian2957]

OOPS . Hope it's not too difficult to put right mate . Nothing worse than all that effort going to waste James :rolleyes:

[istari_knight]

No I'm not worried, Its all a learning curve & I just enjoy playing :thumbsup:

[brian2957]

Glad you're enjoying yersel mate. You have PM BTW.

[system7]

Hmm, bit of a disappointment. But we move on. :scratch:

I did think that capacitor was a bit on the big side for most applications with a 2.5mH coil. Very low crossover.

I rather liked this KEF Celeste 3 circuit that we dug up somewhere. It's for a fairly typical 8" paper woofer and tweeter.

In fact it's the basis of what I do with this combination these days. I doubt it would work with bextrene woofers. But you've got near enough the values to try it.

[Reffc]

I do think Steve that it needs to be pointed out, before someone mis-understands and wastes time and money, that you simply cannot take an approximate circuit value and apply it to any old drivers. Fine for experimentation but it's like shooting in the dark trying to hit a target if crossovers are to be properly designed.

The circuit above is still a very basic circuit and there's no attempt by the designer to address impedance flattening which any decent crossover addresses unless the drivers have been specifically designed to control resonance and have very low impedance peaks.

You NEED to know the specific driver impedance (taken from the minimum impedance on the curve after peak resonance), the cone break-up point for the woofer, the resonance point for the tweeter and the power handling of the tweeter before you can get close to ascribing the right values.

Whilst I agree that there may be values "typical" of many used in 2-way designs, each has to be tailored accurately for the specific drivers in order to work properly. You cannot just take values "off the web" and apply them to any two way using an 8 inch woofer.

You also need to understand the differences between crossover profiles (Bessel, Linkwitz, Butterworth) and how each of those may be appropriate or inappropriate for a specific order of crossover (eg, you would not normally use a Butterworth filter with a 2nd order design and expect a flat response at crossover).

It is important to make these distinctions otherwise people could be easily misled. Crossover design can be a very involved and complex area and the proper measurement tools are essential to getting it right.

[istari_knight]

As I've discovered ! Changing existing component value's even by a small amount makes a huge difference in sound let alone adding further components for impedance flattening etc... Trying to design/build crossovers without the measurement kit like wot Paul's got is evidently just silly unless you don't mind wasting time/money trial & error in which case that's fine.

I for one assumed passive crossover's were piddly simple little things [well they are in isolation] but trying to get them working right with your chosen drivers is another matter entirely & makes building an amp seem like child's play... Trust me, I speak from experience :D :doh:

[system7]

Let's start again. 8" plus 1" units only really have 2 solutions that work. The first is the rough and ready JPW approach, also use in my Monitor Audio R300/MD:

http://img35.imageshack.us/img35/6638/monaudiocrossover.png

http://img580.imageshack.us/img580/713/monitoraudior300md008.jpg

This is not hard to get right. You just adjust that 2.2R input resistor for tweeter level. I added a little extra 5kHz 22R + 0.68uF tank/notch across the bass coil for phase alignment. That was based on modelling it.

The second is the KEF circuit I showed. This is fiddlier to adjust, but not really hard. Both circuits give good phase alignment. The second one deals with cone breakup better. In fact a 4th order tweeter filter is ideal. Probably as good as this one gets. With a typical 1mH woofer, you don't need much shunt resistance in the bass filter.

But I don't accept that this is hard to tune by ear. But it happens that I model filters and know what does what. 8" bass has some severe time alignment problems that limit the solutions, because it is DEEP.

[Reffc]

Let's start again. 8" plus 1" units only really have 2 solutions that work. The first is the rough and ready JPW approach, also use in my Monitor Audio R300/MD:

http://img35.imageshack.us/img35/6638/monaudiocrossover.png

http://img580.imageshack.us/img580/713/monitoraudior300md008.jpg

This is not hard to get right. You just adjust that 2.2R input resistor for tweeter level. I added a little extra 5kHz 22R + 0.68uF tank/notch across the bass coil for phase alignment. That was based on modelling it.

The second is the KEF circuit I showed. This is fiddlier to adjust, but not really hard. Both circuits give good phase alignment. The second one deals with cone breakup better. In fact a 4th order tweeter filter is ideal. Probably as good as this one gets. With a typical 1mH woofer, you don't need much shunt resistance in the bass filter.

But I don't accept that this is hard to tune by ear. But it happens that I model filters and know what does what. 8" bass has some severe time alignment problems that limit the solutions, because it is DEEP.

Sorry, but there's more to crossover design than that. If you model filters then I'd have assumed you knew that. You most certainly DO need to understand the driver impedance curves and resonance points to get it right and also where to choose the right crossover point. Not all 8 inch drivers are the same (in fact they can be very significantly different in response both on and off axis), unless of course you are referring specifically to one particular driver and have assumed that the impedance, break-up points and choice of cross-over point are already known. Tuning by ear is one way of fine tuning (blending tweeter integration etc and I agree it's also an essential part of development) but you most certainly cannot design a crossover from scratch not knowing the driver parameters and just doing it by ear unless you have a lot of time, experience and money to go through a lot of permutations. Lets agree to disagree.

[system7]

Paul, trust me. I know what I am talking about with this 8" driver. Don't try and make it more difficult than it is. I know exactly what that drivers impedance (Le 0.9mH and dc 5.6R) and frequency response is:

http://img10.imageshack.us/img10/7186/vifam21wg0908.jpg

That's easy. Give it a bafflestep coil of 1 or 1.5mH and it becomes a tidy bass crossing over at 3kHz with a more level slope. What's hard about that?

Where it's awkward is the Qts of 0.7 on a Vas of 81L. That means it needs a big box.

You will understand how coil inductance affects rolloff with the 3mH SEAS CA26RE4X, which runs unfiltered in SEAS's new A26 kit:
http://www.seas.no/index.php?option=com_content&task=view&id=118&Itemid=140

The SEAS CA26RFX has less voicecoil inductance at 1.95mH, so you need to add it to get it flat:
http://www.seas.no/index.php?option=com_content&task=view&id=117&Itemid=139

The second driver is more versatile IMO. Bit like this Vifa. There's just one thing to remember when you go second order here, you need a bigger coil still, because the shunt cap lifts the midrange slope. Hence 2.5mH is not unreasonable. Sorry if I sound like a bighead, but sometimes filters are quite easy.

If James wants to improve this speaker, and it's up to him, my advice is exactly right. His mistake was crossing too low. 1.7kHz is absurd for most tweeters. That's what he was doing. That's why it didn't work. A 3kHz crossover with steeper second order on the bass will be much better. :)

[istari_knight]

I'm sure there's a middle ground here ! I don't doubt you get good results using the methods you've mentioned Steve, IIRC you had some input on a diyaudio thread replacing HF2000's with Seas 19TFF1 which greatly helped me. You just have to understand that Paul's a perfectionist, if something isn't 100% it wont sit easy with him ! :)

I find it odd the replacement crossovers sounded so crap due to the too low crossover point as from what I've seen those are the values used in the Type K [+/-10%] or is there something blatant I'm missing ? :scratch:

Here's a photo of a Type K mk1 crossover:

http://i26.photobucket.com/albums/c129/istari_knight9/SnellKfilter_zps57e8a1ec.jpg

LF filter above & HF filter below. The LF consists of 2x6uF 'bipolar lytics with a 1uF film bypass and a 2.5mH air core. The HF consists of a single 6uF bipolar 'lytic with a 0.68uF film bypass, a 0.6mH air core & 5R variable resistor set to 1R7. The only thing I found unusual was that both drivers are wired in phase which I hadn't seen on a 2nd order filter before but some searching revealed that although uncommon its not unheard of.

I wont be doing any further work on the AP3's as I've just sold them to fund a CD player purchase and now it turns out I didn't need to sell them [bugger, bugger & bugger :doh:] Alas, I would still like to understand what went wrong... Could it be the inductors have the wrong DCR ? Or am I simply misinformed of the Type K values in which case I was doomed from the start :D

[brian2957]

Yup , Paul is a perfectionist which was evident when he refurbished my Rega crossovers .

[system7]

I thought you might find it interesting, James, to know how your JPW speaker works.

http://imageshack.us/a/img268/5864/tdg.PNG

This is the JPW simple filter I talked about earlier. You can see that the tweeter gets an easy ride, but that a lot of the rough stuff above 3kHz comes through from the 8" paper bass. This will sound rough on loud male voices particularly, and at high volume with complex music.

This is my current similar version of your speaker:

http://imageshack.us/a/img856/1741/w3ab.jpg

IMO, it's about as good as it is possible to get a soft dome working under these circumstances. The trick is in the crossover:

http://imageshack.us/a/img69/556/djei.jpg

It is the KEF acoustic butterworth idea, that notches the 900Hz Fs of the Morel CAT 298 tweeter.

http://imageshack.us/a/img849/4408/96si.png

http://imageshack.us/a/img843/8213/odae.png

You can see that the drivers get some considerable rolloff. This gets rid of a lot of the nasties.

http://imageshack.us/a/img40/3750/yy9.PNG

And phase is nicely aligned around crossover. I could do better, but probably as good as it gets for this combination. This speaker sounds very detailed. Any criticism of it which I may have, is probably a criticism of soft domes in general. :)