Klipsch woofer - tweeter crossover network.

All the parts have to be put together at some point. Have fun!
Post Reply
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Klipsch woofer - tweeter crossover network.

Post by dave slagle »

Hey all!

Iain dropped me a line to discuss the parts for replicating this circuit for the Klipsch Museum and the thing that caught my eye was the unique case of L2 which just happens to be seemingly really close to bifilar. The thing I cannot figure out is why the need for the low leakage found in bifilar for the use in a low frequency crossover. As near as I can tell, for the frequencies in question the core will more than adequately couple the two tubes and the capacitance added by the phase and close coupling of the inductor windings will prematurely roll off high frequencies.

I asked for permission to discuss here to coax JC into the mix since he was all into the unity coupled idea. Turns out PWK had lots of discussions with the guys at McIntosh and this use of closely coupled windings predates anything mac did with the concept. I'll build and share the circuit in spice here but for now here is the existing info courtesy of the Klipsch Museum.

First the article...
Attachments
PWK Paper  Crossover 2.jpg
PWK Paper Crossover 2.jpg (367.52 KiB) Viewed 61955 times
PWK Paper Crossover 1.jpg
PWK Paper Crossover 1.jpg (340.05 KiB) Viewed 61955 times
PWK Paper Crossover .pdf
(2.04 MiB) Downloaded 2191 times
Last edited by dave slagle on Fri Feb 05, 2021 12:11 am, edited 1 time in total.
Get Your Fix
www.hifiheroin.com
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

here is an image that is a bit tough to decipher.... two versions of the same image and a higher res pdf of the two.

I do want to be clear that the idea here is to replicate the amp as it was built to be paired with what is essentially and early K-horn that had a WE-555 doing the HF duties.
Attachments
kc1.jpg
kc1.jpg (603.29 KiB) Viewed 61952 times
kc2.jpg
kc2.jpg (626.08 KiB) Viewed 61952 times
kc1..pdf
(3.08 MiB) Downloaded 2206 times
Get Your Fix
www.hifiheroin.com
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

here is the patent for this idea.
Attachments
US2612558A.pdf
(262.8 KiB) Downloaded 2243 times
Get Your Fix
www.hifiheroin.com
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

finally the schematic.
Attachments
1946PWK_Amplifier_Circuit.jpg
1946PWK_Amplifier_Circuit.jpg (107.94 KiB) Viewed 61949 times
Get Your Fix
www.hifiheroin.com
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

a start to the complete schematic.
Attachments
image.png
image.png (680.85 KiB) Viewed 61948 times
Get Your Fix
www.hifiheroin.com
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

here is a quick first pass at a spice sim. I guessed on a few things which i will play with and I'll upload the .asc tomorrow.

For the moment I left the feedback out.
Attachments
Screen Shot 2021-02-04 at 6.57.56 PM.png
Screen Shot 2021-02-04 at 6.57.56 PM.png (71.44 KiB) Viewed 61947 times
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

it's clear that L1 - L2 is a common mode choke. as soon as you hear that it simply means a balanced network, and not common ground. that makes the crossover much simpler. the bass is a balanced 2nd order filter and i love the idea of incorporating it on the high Z side to avoid large value caps... un-necessary today, but in 1945 a big expensive problem. when caps are the expensive component compared to inductors, it's a parallel universe you talking bout!

common mode chokes with a gap are going to correct themselves a little so saturation effects might shift the xover point a bit, but not too much... i am curious about how much precision is necessary? especially with cheap iron! the small low leakage inductance tweeter tranny is critical to the success of this circuit. it is interesting the inductive part of the hi pass is not ground referenced, but completely floating. nothing left to chance for high frequency extension!
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i imagine also that L1 - L2 are hum bucking... these chokes would otherwise be very sensitive to hum pickup. by making them bifilar, and common mode, you get hum rejection. very useful in a low frequency crossover!
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i do see the cleverness of loading the triodes with both L1 - L2 AND the output transformer! you can use less inductance in the output primary and still get good loading for the low end... because the common mode choke is also part of the load. very slick!
my skin is thin, my skull is thick.
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

the key thing i am trying to understand is why PWK insisted to wind the wires on top of each other layer to layer which in 1946 with the less than stellar insulations would be as close to bifilar as possible. this gives a decided lack of leakage between the halves of the choke but that comes at the cost of capacitance due to the need to reverse polarity of one of the adjacent wires.

In all bifilar designs the large capacitance is a non issue since the AC potential of adjacent wires is identical. When you reverse one of the wires you effectively get all of the interwinding capacitance across the device which essentially places a cap across the anodes represented by Cbifilar in the attached plot. I used 1p 1n and 10nf as a range and a 1:1 bifilar IT I have here measures 100nf winding to winding.

My guess is given the small inductance of the choke this capacitance has a trivial effect on the 15kHz bandwidth he was trying to obtain. Even so, I still see the increased coupling of those windings at high frequency at the cost of capacitance a move in the wrong direction.
Attachments
PP Klipsch.asc
(8.07 KiB) Downloaded 2205 times
Screen Shot 2021-02-04 at 9.12.49 PM.png
Screen Shot 2021-02-04 at 9.12.49 PM.png (73.64 KiB) Viewed 61939 times
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

because its low frequencies! and the windings are coupled... bifilar on the same core is a common mode choke. a 1:1 transformer functions exactly the same. you will not get much high frequency left on the other side but that is the point! the clever part isn't that. i don't think...? it's the simple low primary inductance output transformer. it would be too small to be used normally... but with a tightly coupled common mode choke in series you get a large total inductance. and the crossover caps are slipped in between.
my skin is thin, my skull is thick.
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Post by iain42 »

Thanks for the help Dave.

The Klipsch Museum has the first Khorn prototype from 1946. It consisted of a smaller LF section w/ Jensen A12 and the prototype horn for HF had a WE555. How high would a WE555 go on this horn?

The museum eventually wants to recreate a party they had in Hope to celebrate the end of WWII which utilized this very khorn in an old WWII hanger which also still exists next to the Klipsch factory. Wouldn't that make it an antifa party? Sorry couldn't resist.... Anyway, I offered to build this for the museum and I greatly appreciate all the help.
Attachments
khorn1sm.jpg
khorn1sm.jpg (47.12 KiB) Viewed 61933 times
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i think it is important to pay attention to Klipsch's initial statements about low cost. large value caps cost more than transformers at that time, and perform poorly. a low turns ratio transformer together with a 1:1 common mode choke gives a higher performance per dollar... at that time. the common mode choke is somewhat active. it is regulating because of the coupling... you get little or no loss of loading at the plates of the output tubes, and a less expensive output transformer that works just as good as a more expensive higher turns ratio interleaved arrangement.

i am not sure what catches your eye about the bifilar winding. it is for the same reason as the "unity coupled" phase splitter or Mac output circuit... the coupling is a feedback mechanism... it increases loading and regulates it's impedance under excitation. but the job of the low pass section of the circuit is particularly tolerant of capacitive losses at higher frequencies! it's a crossover! more care is needed for the hi pass...
Last edited by nanana on Fri Feb 05, 2021 3:44 pm, edited 1 time in total.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

below i add an experiment i did to test the relationships between "critical inductance", capacity and load. this is for power supply purposes but the idea is universal. i wanted to know what the effects were to work strictly in the AC realm. i wanted to put the choke ahead of the rectifiers so i could use less inductance (AC only... less heat, vibration, no gap and smaller size). of course there is no penalty at all for making a choke input filter ahead of the rectifiers. but, i found the regulation and filtering to be maximum with a coupled choke: common mode choke. you remember this i believe. it was used in the big DA-100 amps i made for MJ.

the experiment goes like this: 1H 20 ohm bifilar windings (not practical for high voltage) A) before the rectifiers... and in this way like adding leakage inductance to the power transformer. B) standard choke input, after the rectifiers (a DC component added). C) pi filter. all with the same common mode choke.

it is VERY clear that peak current is reduced! which any choke input filter should do. lower ripple, and better regulation. 1H in the AC realm means a LOT more than it does in the DC world. and that is just physics. the critical inductance for a certain regulation will always be less, or accomplish more, AC wise. not having a GAP is also good. coupling the two windings is an error correction system! just like in the "unity coupled" transformer.
Attachments
common mode choke .png
common mode choke .png (36.63 KiB) Viewed 61931 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here are the sweeps. the top is obviously pi filter. the middle is conventional choke input. and the bottom is AC choke. each is 1H each winding, common mode (coupled). so there is almost double the drop of a conventional choke input, but more regulation.

this action takes cost away from the output transformer in Klipsch's crossover. the load remains high under all conditions... the signal is purely AC realm. a 1K or 2K pp transformer for pp 2A3s is enough after the a common mode choke... instead of a highly designed interleaved 5K pp output trans. and it's only for bass. no highs needed.
Attachments
common mode choke sweep.png
common mode choke sweep.png (35.2 KiB) Viewed 61930 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i keep getting the sizing stuff wrong for this format... too big or too small!
Attachments
common mode choke.png
common mode choke.png (89.97 KiB) Viewed 61929 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here it is. note that the inrush current is completely regulated in the AC common mode choke. the choke input, with DC component, "swings". and the pi filter has an extra cap and does not regulate.

the ripple is clearly lowest in the AC filter, and the pi filter has more C! but the choke does a LOT more for the inductance in the purely AC realm.
Attachments
common mode choke sweep.png
common mode choke sweep.png (69.81 KiB) Viewed 61928 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

the cost of insertion loss is HIGH with a common mode choke. but for loading purposes, you get a lot more bang for your buck! using this technique is very interesting.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

if the grids get balanced drive there is no difference between a common mode choke and two separate chokes. or a center tapped choke. but in a balanced system, if there is unequal loading or drive, some error correction is forced upon the whole.
Attachments
2A3 pp w common mode choke.png
2A3 pp w common mode choke.png (107.5 KiB) Viewed 61925 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

like this... i get it. i don't think it's got anything weird going on. note the small size of the caps! i am sure the output trans on top could be optimized for minimum L to save money.
Attachments
pp 2A3 klipsch xover.png
pp 2A3 klipsch xover.png (106.48 KiB) Viewed 61923 times
my skin is thin, my skull is thick.
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

While the entire concept hasn't entirely sunk in yet this is my take on a few of your comments.

common mode: The signals at the anodes of the 2A3's is differential mode and the series choke behaves exactly as the primary of a PP output which is also differential.

The fact that it is not coupled to the OT makes "L1" a filter element that serves two purposes. It makes the filter 2nd order and it provides increasing load to the outputs as frequency goes up to allow for the 555 output to see full signal. L1 is the critical part of any low pass filter and if a more gradual slope is desired for the bass the 0.32µ caps can be omitted.

From the perspective of the filter which L1 is intended, the coupling is not required and two discrete 1hy chokes could be substituted. PWK is correct on the need for an airgap in this device. This is not to allow for DC current / imbalance but to linearize the inductance with applied signal / frequency. This is a filter element and a linear device is needed. My guess on the choice of coupled chokes is two fold... first a single device is cheaper which is the goal of the design. Second, the coupling between the coils from the common core will inherently help balance the circuit.

I do not see L1 doing much to alleviate the need for the bass OT to have ample inductance and think L1 actually does just the opposite. It allows the bass transformer to have a huge inductance with little concern for high frequency behavior since L1 is in series with the reflected load isolates the tubes from from that behavior. The device i see the need for here is tightly coupled in the low frequencies which is something even a pedestrian core can handle below 400hz and as frequency goes up and the coils become decoupled you still get the leakage inductance in series to isolate / prevent the HF behavior of the bass OT from reflecting back and unduly loading the source. I see any capacitance being a liability which short circuits the ideal which is why this all caught my attention.

If the whole idea can be distilled down to the simplest terms I would say putting the crossover before the outputs lets you get away with murder with each output. PWK's murder was economic and a dear friend who will remain nameless once (ok 1000 times) screamed in my ear a good transformer can do 10 octaves.... choose them wisely. Well My take on this is if you take the PWK approach you now have two transformers that can each do 10 octaves that need to overlap by say 5 octaves for a total range of 15 octaves. This mean 4Hz - 125kHz instead of 20-20kHz.

This entire idea isn't too far from what Jeffrey and I did in the Salt Cellar system. We actually did this one for economic reasons since the amp and the 555 and the horn mouth all had to travel with us on a trans-Atlantic flight and an additional checked bag would have been punitive.

dave
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i am in agreement with this... we are saying mainly the same thing with different terms. differential = common mode. the primary turns of the output transformer can be less than optimum because the chokes boost the load for the 2A3s. an economical tranny for bass. and a small economical one for the treble. yes, klipsch was frugal. he lived in Arkansas! a transplanted Stamford boy?! wtf!
my skin is thin, my skull is thick.
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Post by iain42 »

You bet he was frugal unless it was a plane.... Just before he started Klipsch PWK made the rounds to talk to the audio greats to discuss his Khorn. PWK met the Western Electric guys he was corresponding, Lincoln/Walsh, Fairchild, Jensen, and actually dug horn drawings out of Jensen's waste bin. PWK the dumpster diver... We still have these in the archives....


Thanks for the insight..... Gresatly appreciated...
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

nanana wrote:i am in agreement with this... we are saying mainly the same thing with different terms. differential = common mode. the primary turns of the output transformer can be less than optimum because the chokes boost the load for the 2A3s.
That choke does boost the load for the 2A3 particularly at high frequencies but from an output voltage perspective and AC generated across it does not make it to the speaker so I see its benefit apart from being a filter element minimal. An further increase in the value will simply push the crossover frequency lower and if the value is made = to that of the OT the inductive load on the tube will indeed double pushing the f3 down an octave but the output to the speaker will be 1/2.

dave
Get Your Fix
www.hifiheroin.com
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Post by iain42 »

Do you think you can wind these for us? It looks like this project might actually work out.

Now to shop for the Jensen A12 fortunately we do have a WE555 that was in PWK's collection for the first Khorn...

thanks again for all the help!!!!!
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

consider me onboard... we can discuss timeframe privately.

dave
Get Your Fix
www.hifiheroin.com
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Post by iain42 »

dave slagle wrote:consider me onboard... we can discuss timeframe privately.

dave
AWESOME!!!!!!
I am not in a hurry slap me at the end of the list.
dave slagle
Posts: 2108
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Re: Klipsch woofer - tweeter crossover network.

Post by dave slagle »

well it took a year but I finally got the transformers done for this circuit and they work as expected. (apologies for the cell phone screen grabs... I am out of my element when it comes to working on a PC).


Here is the spice simulation of the circuit that I started with.

IMG_8203.jpeg
IMG_8203.jpeg (80.36 KiB) Viewed 61380 times

This is the measured response of the above circuit. I just got close and did not match the cap values perfectly. (0.15µF and 0.33µF in place of 0.16µF and 0.32µF)
67129018514__A3BBCD0A-EE77-4E6E-9BE0-E3D32D6979A3.jpeg
67129018514__A3BBCD0A-EE77-4E6E-9BE0-E3D32D6979A3.jpeg (112.76 KiB) Viewed 61380 times


The slight dip in the 400-1000Hz range of the HF and the somewhat steeper rolloff slope is from the 0.04µF cap in series with the HF output and eliminating that cap altogether gives the below. My only guess is that was a tweak by PWK done in system or simply to keep the LF info off of the 555.

67129025646__2C0B2965-D66A-41AF-9ADA-1381CD4B0235.jpeg
67129025646__2C0B2965-D66A-41AF-9ADA-1381CD4B0235.jpeg (115.22 KiB) Viewed 61380 times


The speakers actually being used are a 8Ω jensesn A-12 for the bass and a 16Ω 555 for the treble so I had to do two different ratios on the transformers to reflect the needed 5K load for the crossover. This resulted in a 3dB increase in the HF output seen above. Since PWK abhorred L-pads it was only logical to use an autoformer to allow for level matching while still providing the required 16Ω termination for the crossover to work. Here is the autoformer giving 6dB of attenuation and providing the proper 16Ω load

IMG_8206.jpeg
IMG_8206.jpeg (116.26 KiB) Viewed 61380 times
Get Your Fix
www.hifiheroin.com
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Re: Klipsch woofer - tweeter crossover network.

Post by iain42 »

I can't express how much we appreciate the help with this project. It will really help us be historically correct. I'll make sure you get an invite to the END OF WWII recreation party. Period clothes are required. The WWII hanger is still in use and available to us.

I will be posting regular progress. This will be posted in multiple parts.
https://www.itishifi.com/hifi/1946-pwk- ... t-audio-p1

The parts quality is amazing. I'd like to think PWK would be excited about the quality of your speaker-level autoformer.

Image

https://www.instagram.com/p/Cca7TxrOcpT ... _copy_link
iain42
Posts: 6
Joined: Thu Feb 04, 2021 11:06 pm

Re: Klipsch woofer - tweeter crossover network.

Post by iain42 »

Thanks for you all your help Dave the #1 Klipschorn rides again thanks to your help on the crossover section. We managed to procure a Jensen A12PM from ebay and the museum has a few Western Electric 555s. This is what PWK used to built the prototype during WWII. Fairchild actually sent the plywood for PWK to be able to put this together because it was wartime there was rationing of most materials. The original amplifier artifact we found in the museum basement had a volume knob that went to 11... Eat your heart out spinal tap 8-)

https://youtu.be/4dN-PKubCt0?si=KWHQWThLshACzZxL

Image

Image
Post Reply