Experimenting with Feedback

[Fatrecco]

Hi,
This forum is very helpful for me, problem is my bad english. So I read to much over

Since a few weeks I startet to build a new head out of car speakers. This time I'm experimenting with Feedback. I make paperglued cones with an aluminium push rod at the end. These pushrod is not connected to a torque tube. Around the rod is a feedback coil. Between the feedback coil and the driver is steel and aluminium to keep away the dynamic magnetic fields from the driver coil. I have tested it before without the extra magnet for the feedback. Now, I got a good feedback amplitude (with very little 50Hz noise) but another problem where I think you all know it -> phase issues
In this forum I see often a phase plot, but I don't know the software to make it. The only way I check out the phase correlation is to use a 2 channel oscilloscope or record it to a stereo wave, where the one channel is the original driver signal and the other channel is the feedback signal. I can also see the X/Y diagram.

One messure point is direct at the driver contacts. The other point is after the feedback amplifier. I build a simple amplifier with 2 non-inverted NE5532, without a capacitor. I tested this produces no phase shift up to 15kHz. But now with the mechanic everything goes bad. In my opinion the signal at the feedback is older then the signal at the driver coil .

Several problems I think about:
-The mass velocity rises to its maximum after(!) the maximum current is flow (?)
-sound velocity of the materials -> paper glued cone (?)
-capacity and/or inductivity of the wires (??)

Summing the signals can destroy my new expensive construction. So I don't want to try it without making things clear as possible. Here are many threads about problems of feedback, but I don't know how to handle with my results.

First Problem I cannot display my phase shift in a diagram. Only testing frequence by frequence and hold a ruler at the oszi

In this file is the left channel the original and the right is the feedback signal. Sorry for the different levels. The amp is from Sony and only for desktop.
orig-feed.wav


I hope someone can help me a little bit.

MfG Fatrecco

[Fatrecco]

I don't found out how to edit my text.
.
Here is a picture of the driver with mounted feedback. The Coil is < 5mm diameter wound around a 3mm polythylen socket at the aluminium rod.

feedbackcoil.jpg

[Fatrecco]

Hey I confuse the polarity of the feedback. This night I builded a mixer and added some feedback. There is no EQ in the signalpath. There are only some DC killing capacitors at the beginning of the mixer circuit. Very simple. I feel good cause it seems to eliminate the first resonance at ~830Hz. Blow up the Resonance at ~12kHz. But I have to be quiet, its 0:49 here and I live in a little flat.

feedbacksingletest.JPG

Next thing is to buy an equalizer for the feedback signal and mounting the torque tube.

[Fatrecco]

I made a phase diagram out of the wave files. Limiting and mixing. The rest is to check manual out how the phase is going.
I'm confusing

phase.JPG

The phase shift comes from +90° goes over zero at 1. Resonance and then -180° and then over -360° (2. Resonance). I can't understand it. The weight is under 1gramm. It seems to be not realy usable. *fck!*

[Fela Borbone]

Hello,
Hey, nice project! So the main resonance is 12k? Thats way too much for a moving coil... are you sure that the search coil picks up only the motion or also the field emmited by the drive? If it makes no damage, try to stop the motion and measure what happens...

[Fatrecco]

Sorry I forgot to name the diagrams. the first of the spektral (fft) diagram shows the frequency response taken from the feedbackcoil. Without mixing feedback to the main-signal. The second below shows it with mixed Feedback.

The main resonance is around 830Hz - 840Hz. Second(?) resonance is about 12kHz. Without any Feedback. The 360°(?) Phaseshift is a little above these resonance. The Amp makes some problems at 500-600Hz. Bass boost. I have to check out what the amp produces for shifts When I try to stop the pushrod motion then the signal goes very low! Without the feedback magnet the signal is only there when I drive the coils very much.

Note that I have only a few positive feedback testet. Negative feedback brings same problems cause the second resonance.


Test with white noise:

3response.JPG

[EpicenterBryan]

There are several things that are causing you problems right now.

The number one thing you need to do is put a resistor and non polarized capacitor in parallel. Then put that combination in series with your driver. Not knowing details on your amp and impedance or DC resistance of your drivers, you can just start with a 2.2uf (non polarized) , and 4.7ohm resistor. What that does is provides a frequency dependent voltage divider that boosts HF. Actually, at lower frequencies it attenuates but as the frequency increases the 2.2uf cap starts shorting out the 4.7 ohm resistor and since the combination is in series with the driver, HF is boosted (really it is less attenuated). The useful side effects are two fold: First, it causes a positive phase shift for higher frequencies which you really need for you feedback. It also makes your amplifier operate in more of a current amp mode than voltage mode. The load from your driver is more frequency dependent than you think, so this RC combination helps keep the load at the amplifier more consistent VS frequency.

The next issue you are running into is your measurement gear is not able to directly extract the phase information. It's going to be hard for you to understand what's going on. But here is the quick version, since I don't have much time left today...

Really, you need to keep your thinking into +90 degrees, and -90 degrees. That's 180 degrees change. A feedback loop needs a stability margin so the useful range may be 20 or more degrees less... Once you go past 180 the polarity changes exactly like you are seeing at 12Khz. So if the bass is +90, and it's 0 at resonance, and drops below 0 as frequency increases - all the sudden the feedback starts adding rather than subtracting at -91 degrees and your feedback adds and go unstable (probably way sooner).

Anyway - where do these phase changes come from? You nailed it on several issues. First, the materials used. If your paper cones are not stiff enough, and mass on the end of them they will compress and stretch as you try to accelerate the mass. The effect is more noticeable at high frequencies. Also as you noted, even if the structure is rigid and low mass there is a propagation delay through materials. At high frequencies it adds up. Keeping the feedback coils close to the driving coils is more important than you might think.

HOWEVER, that brings up the next big issue. Even if you can get the pickup coils closer, the magnet structure and shielding becomes a big deal. Most pro feedback pick up coils have two opposing windings inside opposing magnetic structures so that the "hum-bucking style" coils resist magnetic coupling induced into the coils from the drivers coils.

That's about all the time I have for tonight.

I want to make sure that I'm an not discouraging you in the feedback quest. I have not given up either - I just have lots of other projects going on. The down side about taking a break from anything for too long is that you forget how it works and where you left off. That's nearly where I am now. I hope I did not lead you astray, and hope you figure it out and share...

Bryan

[Fatrecco]

Hi Bryan thanks for your feedback (< 1)° . I read and tried to understand your OsoMiso and Groovescribe threads. Very helpful. I wonder that you got such a good phase correlation in contrast of my.


I know about these capacitor and on my real (selfmade) lathe I had such of 2,2µF capacitors and parallel there where two 12V 5W light bulbs. But here at this desktop test I used no serial high pass. Mayby I should. My driver are 4 Ohm impedance. The amp is very poor, I got larger (better) amps. But the AMP produces no phase shift. Checked with my 2 channel oscilloscope and a sine sweep signal.

The -90° phase shift is at 2kHz ): after that it goes to -180°. Then it goes to > -360° I know electrotechnical +90°,0°,-90°. But in reality the signal is over one periode in the past.

I think you have right by the fact, that something is wrong with the driver.

1. The feedback coils are far away from the Driver. There is a time delay by the sound velocity of the materials. These part should by a constant value independent from the frequency. Or I'm wrong ? This could by located by a diagram, I have to make in future with turbo pascal. (:

2. The paper cone can be not stiff enough. I can make another 2k epoxi layer on the cone and test. But I don't want to trash the driver.
The Mass of the whole cone is under 0.5g. Maybe the coil was not realy fixed at the pushrod.

drivercone.jpg

3. You observed the elektromagnetic field from the driver coil. The method you write, I know this from guitar pickups. Problem is you need an opposite magnetic field so that the voltages whould not substract each other. I think this is very difficult to mount it so near at the driver.

Now I disassembled it to make some parts new.

I need more time time time.

MfG

[EpicenterBryan]

I wonder that you got such a good phase correlation in contrast of my.

There were several things going on. First, I had the RC network installed between the amp and drivers. Also, I was using the Caruso preamp boards that have it's own differential FB amp inputs and it's own mixer that has lead / lag and iRIAA built in. Initially, I used values like specified for a Caruso head. The Caruso boards also have a LF boost intended for the feedback loop. The upper part of the iRIAA can be switched out, but the lead path at HF is intended to help add positive phase at higher frequencies for the feedback loop as well.

In the end, I added analog parametric equalizers installed between the Caruso board output and the Power amp inputs. In effect, I added another place within the feedback path where I could experiment with injecting positive phase to keep the loop stable. Ideally, what worked could be incorporated into the mixer and the parametric EQ could be eliminated. But the Groove Scribe head was just too big to be practical. And it was never clear how much of the FB phase issues were related to physical constriction (FB coils far away from driver) and how much was related to poor shielding and magnetic coupling.

The next big difference is that my cones were attached directly to the coils. Many off the shelf drivers have a metal cap attached on top of the coils, and a rubber surround or spider attached on-top of that. If your cones are not directly attached to coils or at least to the metal cap very close to the coils and or are glued to the rubber surround and not directly to the coil former, your cones will suffer extra delay in movement. The phenomenon is observed as a "decoupling" of feedback at HF, where the negative phase is much larger than expected.

Going back to Groove Scribe, you see my cones were attached directly to the driver coil former. And I even went as far as eliminating the springs in favor of my own diaphragms / driver centering devices.

What is the ideal combination of off the shelf DIY and totally custom? Groove Scribe was as close to custom as I ever want to mess with these days. It worked, sounded really decent, was too big, was fragile and hand made. After I blew it up doing a YouTube demo and rebuilt it, it was vastly improved and was a great learning tool.

I'm kind of over it these days. One of my sisters passed away a few months ago. She was only 4 years older than me. So I'm kind of re-evaluating what battles I want to keep tackling. FB on a DIY head is kind of at the bottom of my list right now.

Things for you to think about...

Bryan

[Fatrecco]

Thanks for input. I maka a new transducer with another glue. At the beginning of making my heads, I make some cones out of alu from a beer can. Maybe I should try this again. One speaker I destroyed to look where the coil is situated. Not in the deepening at the diphragm, a little bit more in the inner diameter. And I see a thin coil and it seems it is out of 17 windings (wtf).

You gave me lots of information. I've to try and find out how rigid my mechanics are. When I get some new results I will post it. In the future I want to make my own transducer. I think I posted some pictures in the past from the build of the metal parts. Since nearly one year I made no more record . But this year ...



MfG