Klipsch RF7 (old version) crossover modification

[Yasnyi Sokol]

Hi, everybody.

I made some research on Klipsch RF7 and modified the crossover, hoping that might be interesting for others.

RF7 is considered by many people as a good speaker, it was the reference of Klipsch. For a long time I actually loved its sound and promised I would never, never look inside and try to change inner components:).

I loved it so much that decided one day to simply buy better components. Better bass coil, perhaps better caps. Try Mundorf supreme caps etc. You know that feeling. Of course it ended by changing almost all parts and full re-design.

So, what I want to share.

1. The bass coil.

Oriignal coil is very small, it uses iron core and has small cross section. You might think it's OK, or even very good coil. But it isn't.

To compare 4 handmade coils were used.

a) Air coil with magnet wire 1.8 mm.

b )Core Inductor coil with steel laminates (german steel equivalent to M6), 0.35 mm, Similar design to Klipsch, but 4 sq.cm instead of about 1 sq.cm of original coil

c) Coil on two bobbins with amorphous Hitachi metglass core (size 32 to their specs), two U-parts with small air gap between. Rdc less than 0.1 ohm.

d) Coil on 1 bobbin with same metglass core but on one side only.

The 5th was Jantzen Wax Coil 750 uH, 0.16 ohm, 12 awg. I would like also to mention that isolation material in Jantzen coil is paper.

See pictures.

a) and d) where much better than Klisch's original. More resolution, more microdynamics. b )was harch, c) was slightly compressing the sound, perhaps due to small air gap. d) was the winner among handmades.

The absolute winner was Jantzen wax coil. It has a lot of transparancy, pure and delight presentation of sound. what I liked is the fact it can play well at low levels. I can say that compared to jantzen coil original coil from Klipsch is like a plug, almost half of the sound is lost in it.

However, the funny thing was new coil (Jantzen or d) sounded with some sibilants at the top. Now I understand that the losses in Klipsch coil are balanced in such way that they damp partially sibilants. But they also delete the half of the sound, and I could accept it already. It was time to look deeper at the crossover.

Alexander.

Edited April 2, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

The crossover.

Original crossover is shown in the picture. In some speakers there is 2 ohm resistor with 18 uf in the tweeter.

Here are the room measurements at 1 m distance which are not so bad, if we don't talk about bass..The hole at 150 Hz is due to interaction with floor when mic is to close to speaker (1 m).

Next, the drivers independently.

bass/mid driver (about 25 cm at axis)

tweeter

Not a surprise: the ceramic cone has high peak (3.2 khz). I knew about that when i was buying RF7s. So what? My doubt was that second-order crossover is not enough to deal with it, but at that time I closed eyes and beleived to Klipsch designers it is OK. And I was wrong.

Edited March 29, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

The drivers with crossover

So how do they work together? Here are the calculation and measument of the drivers with crossover.

The waterfall of the bass/mid driver with X-over,

Clearly, the 3,2 kHz resonance remains and it is not so pleasant. The quest was what to do next.

Edited April 2, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

There were two options

(i) damp it mechanically

(ii) damp it electrically in the X-over. First I was interested where it comes from. The first idea was to paste some absorption material (paper, leather etc) on the back side near the driver cone edge, at the place where usually resonances of hard cones appear, that is about at 3/4 of radius).

The driver was taken out and the cone was measured near the edge, between the edge and the dust cap, and at the axis, that is in front of dust cap, at about 1 cm.

At the edge

At the middle between edge and dust cap

at the axis

This shows that the main contribution to unpleasant 3 khz peak is due to the dust cap. I cannot fix the problem with damping the cone on back side. Some good idea was to delete the dust cap. I was told by a friend that usually helps much and "opens the sound". But I was afraid the point of cone break up will move down and I would not be able to fix it with two-way crossover.. The dust cap improves the rigity of the cone, without it the cone will bend easer. But You can try to remove dust cap.

Here are also some measurements with damping (scotch paper of 2 and 4 layers, and silicon on the paper). Doesn't help really. Moving mass is increased on the dust cap, the peak is increased or becomes wider with damping, but doesn't go away

paper 2 layers:

paper 4 layers

paper 2 layes plus silicon

So, the right way was (if one doesn't delete the dust cap) to damp the cone resonance electrically.

BTW, the dust cap is made from thin alluminium.

Edited March 29, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

Before I continue I would like to show some measurements of ribbon coil. Some say ribbons have huge capacitance. Is it why sibilants appeared? No. One graph is original klipsch (they say it is 0.75 mH but measuring two I found one is close to 0.75 mH but the second is slightly less). Another graph is jantzen wax 12 awg. Looking closer at phase one can see that jantzen is exactly with 90 degree phase with no appearance of capacitance effect. Klipsch coil starts to slowly reduce phase at the top end. Not to much, but this is to show that jantzen ribbon is almost perfect coil.

Klipsch original coil

Jantzen ribbon coil

Edited March 29, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

New crossover (original drivers).

So, here is another version of X-over for RF7. It was designed together with my friend (he asked not to give his name).

First, the notch filter at 3.2 kHz, and some mods with 16 uf-10 ohm on the woofers to better match in phase with the tweeter.

The idea of tweeter crossover was taken close to original Klipsch design. 18uf plus 0.12 mH reduce the range 1-5 kHz to flatten the tweeter (this is not the last opportunity, more on that later). We also increased the caps to 8.2uf and 15 uf based on Leap calculation, measurements and listening tests.

1.8 uf and new coil 500 uH slightly reduce peak at 5 kHz, which is also present in original klipsch sound.

The input resistor on tweeter can be adjusted arround 2.0-2.4 ohms up to personal taste.

bass/midbass waterfall

Room SPL 1 m

Room SPL 2m

All coils are Jantzen Wax (12 awg 750 uH on bass and 14 awg others). Caps are Mundorf supreme on bass with 120 uH, 16 uf with 10 ohm is Jantzen cross cap. 8.2 uf on tweeter is also Mundorf supreme. 15 uf was made by simple 12 uf Janzen cross cap by-passed with Fostex 3.3 uf cooper tin foil. 18 uf was also jantzen cross cap.

Later I ordered all caps Mundorf supreme, but already modified the tweeter. That is another part of the story.

At that level I enjoyed much new crossover. It has more resolution and is more transparent. Direct comparison to original crossover showed it is cleaner and more natural to my taste. Slightly darker on the top, but it is to personal choice (one can change the input resistor).

Edited March 31, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

Simplicity

Let's look again at tweeter equalization. What I didn't like in 3rd order crossover is the number of elements plus LCR filter working in the wide range for tweeter equalization. A lot of elements for the signal to pass through. This can be also understood considering the decrease of impedance, as some amount of signal current has to pass through LCR filter instead of the tweeter.
Can we do tweeter equalization more simple? Here is another method: first order plus damping of the main tweeter resonance.

 

 

One cap on the signal pass, which can be selected from top level caps due to its small value.
One can say that first order crossover is not good for the compression driver. However, look at the voltage graph and tweeter spl and compare them to the graphs in upper post. The simple version has similar sharp roll-off for tweeter which is not bad compared to the 3rd order design.

The other points are:
Klipsch-analog design has
(i) two caps on the pass plus coil (3rd order).
(ii) LRC filter in the wide range 1-5 kHz to equalize tweeter.

Alternative X-over has
(i) only one cap on signal pass
(ii) narrow LCR filter tuned to 1 kHz (main tweeter resonance), which works almost outside the working range.

I didn't make that exact crossover but one can try. Notice that 5 kHz peak still remains, as it is present in the original driver.
 

 

Edited April 1, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

deleted

Edited April 1, 2015 by Yasnyi Sokol

[Yasnyi Sokol]

deleted

Edited April 1, 2015 by Yasnyi Sokol

[robotc]

Hi Alexander

 

I was really interested to read about your measurements and modification to the RF-7 crossover. You have done a lot of work  testing and preparing the information you have posted. Congratulations.

 

I have been making similar measurements and modifications on my RF-3 speakers which, like the RF7, have twin aluminium cone woofers and a tractrix horn tweeter (speaker drivers are different models of course)

 

Like the RF7, the RF3 metal cone woofers have a large resonance ( at around 3-5kHz), which can only be reduced significantly with a notch filter similar to the one you used. The difference with the resonance reduced in level is clearly audible.

 

I believe this extra (unwanted ) sound energy around 3-5kHz is a big part of what gives the speakers the ‘klipsch sound’ which can be very impressive and realistic with some instruments (eg  saxophone) but which can become unpleasant and harsh when listening to music, especially at high levels.

 

 

Robert

[ninjai18]

This is super cool. Love the info! Thank you!!

[Yasnyi Sokol]

Thank You for Your responses!

[Yasnyi Sokol]

Effect of mesh on compession driver

Some folks said that without mesh the sound becomes better. And its role is reduced to protect the driver from dust etc.

I tryed to check that.

Technically the mesh screen is an acoustic resistance which slightly damps the diaphragm. Here are some measurements from a distance close to RF7's horn with and without mesh on the driver.

Green is no mesh, brown is with mesh.

The measurements were done at different time with different input level, so absolute value is not important. More important is that without mesh the response is higher at top frequences which seems to be predicted.

The waterfalls show the effect on diaphragm resonances.

With mesh:

Without mesh:

One can see that in most important region 2-10 kHz wihout mesh the peaks are slightly reduced, which is quite interesting and promissing.

If one is interested to compare the driver in RF7 with other comression drivers I can bring here the measurements of B&C de200 in original horn of Zingali Home Monitor and radian 475B in some horn:

de200:

radian475B:

This may give an insight to someone's consideration whether the klipsch driver is better on not . I beleive from these measurements the klipsch driver is quite good one.

Looking at these measurements i decided to build a test crossover for driver without mesh. The first one was made similar to the first modification i posted above with the 3rd order filter. However something strange happened to the sound. I heard the improvements and good articulation of vocals, but overall it was a feeling the driver is not very clean, perhaps due to something happening at top frequences. Maybe the tweeter sounded faster than mids. I tryed to put additional notch filter to reduce 10 kHz region, in next version a small input coil but it was not full satisfaction.

The next crossover was made similar to simple 1st-order:

This version sounds interesting and now i need to accept whether it is enough of highs for me. I found that the effect of small coil on the top makes the sound of the tweeter cleaner and consistent with mid drivers. Otherwise, if i don't finally like that sound the mesh will come back, of course.

I counted the threads of the mesh (btw, not so easy to do ). There are about 76 threads per inch. Some meshes can be found on ebay even from copper or brass, from the numbers in the range #40-80, that covers about the same cell size as in original one. Perhaps it would be possible to use an intermediate solution with mesh, but with sligtly bigger cells. I understand that engineers from Klipsch know what they are doing when use the mesh. But it is interesting to look at the things with my own eyes.

A friend of mine also having the RF7s asked me whether it is possible to do most simplest version of X-over without any notch filters and for the case without mesh (he also deleted it). He put soft cloth material instead of original mesh and says it is very pleasant sounding. Here is the calculation of 1st order X-over for his case (to be exact with no mesh), but i didn't listen to it myself.

Edited April 3, 2015 by Yasnyi Sokol

[Deang]

Stick around Yasnyl - we've lost many of our most technical folks.

I never would have thought the screen would have had that profound of an effect.

As for the experiments and modifications - very interesting stuff, and I can certainly appreciate the use of well constructed higher performing passive components.

[wstrickland1]

Yasnyi certainly seems to have his stuff together. Thank you very much. As an owner of (4) RF 7 (originals) I appreciate this extensive information.

[CECAA850]

Pulling the mesh certainly helps above 10K.

[Yasnyi Sokol]

Thank You, guys!

Think the coil on the top still slows the sound, not so agressive and sharp as I would prefer! The coil has gone and the test Xover now is as follows (1uf +0.15 mH + 6.8 ohm do some job at 10-12 kHz).

 

If one puts a soft cloth inside it is OK, but without it there is still too much energy on the top.

Edited April 5, 2015 by Yasnyi Sokol

[robotc]

I am not sure about the wisdom of using a first order crossover here. The output at the tweeter resonance (around 1kHz) looks to be only about 10dB down on the output at 6kHz.

 

Are you modelling the crossovers you have designed?

Using Passive Crossover Designer I modeled the electrical response of the crossover you posted in post #17 . The response is into 8ohm.

 

This is not the whole story because you need the frequency response and the impedance of the speaker driver to model the sound output. However the shape of the crossover electrical response might give a clue as to why there is too much output at the high frequency end.

 

 

[Yasnyi Sokol]

Hi, Robert,

There are voltage curves in posts #6 and #7. Blue line is for tweeter. I added it also for the scheme in post 17. For modeling I use Leap CrossoverShop with impedance measurements of the drivers first, of course. Otherwise it wouldn't coincide with acoustical measurements.

Original design of Klipsch is as follows:

The difference between the shelf 2-5 kHz and 1 kHz is slightly better for 1st order with notch filter. The 3rd order outperforms bellow 800 Hz, which is already -15 dB. Of course, it is always a question of compromise.

Edited April 5, 2015 by Yasnyi Sokol

[ninjai18]

Stick around Yasnyl - we've lost many of our most technical folks.

I never would have thought the screen would have had that profound of an effect.

As for the experiments and modifications - very interesting stuff, and I can certainly appreciate the use of well constructed higher performing passive components.

Why'd most of them leave? I've seen a few people say that. That's unfortunate. Edited April 5, 2015 by ninjai18