Project "Groove Scribe"- A DYI stereo 45/45 head.

[EpicenterBryan]

That response is looking really nice. You seem to have a good handle on things. One odd thing. I think you might have the polarity flipped. The phase display starts out at -90 degrees at low frequencies and transitions to +90 after resonance. It should be just the opposite (+90 at low frequencies and -90 after system resonance.

What is the distance from the drive coil to the phono pickup? The extra phase lag could be due, in part, to the sound propagation delay from the driver to the pickup. It could also be due to the driver L/R time constant. One way around that (if it is the source of the lag) is to drive the head using a constant current source rather than a constant voltage. That can be done easily with a second feedback loop around the power amp with a small current sense resistor in the ground return of the head.

In any event, it looks like you should be able to close the loop beyond 8 Khz with a bit of lead lag compensation.

Rats! I didn't catch that.
I'll check again since I took it apart to remove that epoxy. I used super glue this time.
I'll also look at it with and without the rod support on the test jig.
The total rod length is 67mm, and the notch for the stylus is 44mm from the driver.

We had talked about the L/R time constant at one point, and adding a resistor in series (since I have power to burn). Right now I have a 10 ohm in series with the driver. But I also have a 2uF in parallel with the 10 ohm. That 10 ohm / 2uF gives me about a 3db boost at 10K, and about a 6db boost at 20K. But it also gives me a +30 degree phase shift from about 10Khz-20Khz. Down around 2.5Khz (where the new resonance is) it's about 12 degrees, but again in a positive direction. So it still seams beneficial. What do you think?

As far as the second feedback loop, and current sensing are we talking about doing something like this?

http://www.current-drive.info/9

I've used shunt resisters before to make current measurements but I'm note sure what would be involved in this application. Do you have time to sketch up something?

Also, earlier when I said 0.76 watts is 1.42cm/s at the driver rod, I forgot this represents only 1/2 the velocity on an actual groove because one driver only is cutting 1/2 of the mono sine wave. So if both drivers are running (out of phase) shouldn't the velocity of the groove actually be twice that? That sounds more reasonable, right?

Bryan

[EpicenterBryan]

The key will be to get the best compromise between stiffness x level of power.

Yep. That's why I made the offer to share this material. Take me up on it and play with this stuff. I'm more than willing to post my sketchup spring files so people can tweak them. Just share your changes if they work better. The thickness I chose was just to make sure the coil was at the exact same depth as in the stock driver once the carbon fiber disc was attached, and that one surface printed totally flat for best bonding on the printer since this stuff is hard to print. I'm not saying this nylon / carbon fiber pair is the solution, but the results are much different (and better) than everything else I've tried. There are just so many different things that could be tweaked, too many for any one of us to try them all. Who is up for this? So far, I have one taker.

Above this frequency, the feedback sum rather than subtract. This is phase shift the blame?

Yes. A practical single feedback loop can only handle around 135 degrees of phase shift before it can go unstable - the other 45 degrees are usually reserved as phase margin to insure stability for changing conditions. So the trick is to introduce some phase shift if needed (one way or the other) in the feedback loop or withing a nested feedback loop to compensate sometimes in only a region of frequencies. What ever is needed for the outer loop to stay well below 180 degrees of overall phase change. With only one main feedback loop, at 180 degrees phase shift, the signal is completely inverted and will add rather than subtract. That's what you are running into. That's what I have run into. So before trying to design something to band-aid the underlying problem, I wanted to try to get the drivers working better first. After all, I'll have to make another driver and who knows if it will be exactly the same!

Mark is the best one to explain this, but the closer we can get to the ideal +90 at low frequencies, and a switch over at the resonant frequency to -90 at high frequencies the better. Then some added phase shift (like the 30 degree positive I mentioned with the RC series components on the high end) and a little shift on the low end may get the overall with that 135 degree number I mentioned to be totally stable. Who knows what other things will come up when the drivers are mounted with the other stuff.

Bryan

[markrob]

Hi,

Your link on using current mode is right on. Use figure b as a guide to setting that up. I forgot that you already have some crude lead lag compensation going with the series resistor and bypass cap. It might be good to re-plot without this, just to see where you are.

I did some thumbnail calcs on the phase shift due to the distance from your driver to the pickup. You indicated that distance is about 44 mm. I forget what material you are using for the rod, bur lets use steel as an example. The speed of sound in a steel rod is in the range of 6000 m /sec.

http://www.engineeringtoolbox.com/sound-speed-solids-d_713.html

Very fast as compared with air (~343 m / sec). The formula for phase shift in degrees vs. delay time in seconds and frequency in Hz is:

Phase shift = -360 x F X t

The time delay in steel at a distance of 44 mm = 44 mm / 6000000 mm / sec = 7.3 us. That doesn't sound like much but...

At 20 Khz, that results in a phase shift of -52 degrees. That not insignificant. This is a very rough approximation as we don't know the speed of sound in the rod you are using. But I think its in the range.

Mark

[EpicenterBryan]

I did some thumbnail calcs on the phase shift due to the distance from your driver to the pickup. You indicated that distance is about 44 mm. I forget what material you are using for the rod, bur lets use steel as an example. The speed of sound in a steel rod is in the range of 6000 m /sec.

Mark, I think you just solved the issue.
The rods I'm using are Carbon Fiber. A quick check on the web found: Composite, graphite/epoxy: 3070 m/sec. So it's twice as bad as steel. It gets worse for Ciuens though because his rods (at least where his feedback is wound) is plastic. For ABS it's 2230 m/sec. I didn't find a listing for PLA and I'm not sure what he is using.

The reason I was using carbon fiber for the rods (at least on the test jig) was mainly for weight and also because they were non-magnetic and easy to find compared to Beryllium copper, Titanium, or even non-magnetic stainless.

So, if I really want to keep the mass down, use something non-magnetic while having a material with fast sound propagation I should be looking at Beryllium 12900 m/sec (dangerous to mess with), or Titanium 6100 m/sec. Aluminum is also an option at 6320 m/sec, but not a great option for connection right at at the torque tube because some flex is needed there.

So for Ciuens situation, he should probably use a non magnetic push rod as well, extend it all the way to the voice coil connection and wind the coil directly on the rod. Don't you think?

Bryan

[Bahndahn]

Interesting!!!!!

Wow.

For what its worth, Pyrex glass's sound velocity is 5640, its non-magnetic, and its density is low! [2.21 10^3 kg/m3]

[EpicenterBryan]

For what its worth, Pyrex glass's sound velocity is 5640

I've got some 3mm Magnesium on order to try. It's 5800 m/sec and less dense than Aluminum. In the final build, I'll try to drill a 1mm hole in the end to insert a final link segment of stainless music wire. So the goal will be to have the slight flex needed at the torque tube connection be the stainless music wire.

I guess I'll need to rethink at least part of the torque tube if the phono cartridge as feedback idea could ever work. But then again, if Ciuens gets his feedback working better with some changes to the rods, that's a better way to go anyway. Harder to do, but better.

Fun stuff.

Bryan

[Bahndahn]

I've got some 3mm Magnesium on order to try. It's 5800 m/sec and less dense than Aluminum. In the final build, I'll try to drill a 1mm hole in the end to insert a final link segment of stainless music wire. So the goal will be to have the slight flex needed at the torque tube connection be the stainless music wire.

Great to hear!

The recent progression of information on this thread has really shone light on just how many variables are necessary to factor in in this quest. I am continually challenged in understanding phase, but its slowly coming together.

Another factor that really blows my mind is the fact that the speed of sound in the pushrod material is a unique effect separate from the moving mass in the case of feedback, but it makes sense in that the feedback is referencing the original, un-actuated signal. As I am attempting to understand this: the delay introduced by the [pushrod] material at a given distance from the coil would not necessarily have an adverse affect on an open loop cutterhead. [?]

I understand that the feedback coil's phase - relative to the 'pure' signal, is one to be corrected in the analog domain [until someone presents a microprocessor option ]. It is however possible to correct the driver's response phase [frequency-specific phase problems] in pre-processing. Is frequency-specific phase shift in an open loop system audible? Considering all of the content at a frequency y is offset by x, no risk of cancellation occurs..? I'm not sure if that is true, this is a question.

To summarize, I guess I am asking: Is phase response in a driver only crucial [in this case] as we are attempting to implement a closed-loop design?

• • •

Moving the feedback coil as close to the drive coil is proving to be important. An Idea I am attempting to predict the feasibility of is to use the same magnet for the feedback coil. The FB coil could be wound on an [backwardly] extended portion of the pushrod such that it was nestled at the center of the large neodymium a millimeter or so away. This apparatus could be surrounded in a magnetic isolation cylinder to block out the interference. I don't understand magnetic interaction well enough to know if this would work, but I think its worth throwing out there for 'feedback'.

[Ciuens]

Hello guys,
During those days did some testing with feedback, I will share with you my results.
First, I made double magnetic isolation with copper plates, as I show in pictures.

IMG_3898.JPG

IMG_3922.JPG

Before placing the coil on the rod, I decided to test the magnetic isolation. All tests were performed using pink noise and the same level of power. The first image shows the coil with 6mm away from the voice coil and without any isolation or coupling.

IMG_3909.JPG

The second image was taken with the voice coil distance of 6mm, without isolation, but attached to rod.

IMG_3910.JPG

Continue...

[Ciuens]

Continuing ..
The third image was taken with the voice coil distance of 6mm, with a double magnetic insulation but without being attached to rod.

IMG_3914.JPG

The fourth image was made with coupled coil, double magnetic insulation

IMG_3915.JPG

The difference between isolate and not isolate the feedback coil is very large, especially in the higher frequencies.
I think I got isolate enough to get a good response feedback, what do you think?
Notice the small neodymium magnets installed next to the feedback coils. This did much to increase the gain of the coil sensitivity.
After this check, I did another test to measure the phase shift. I used a great program called Smart 7. The result can be seen in the image.

IMG_3955.JPG

With this configuration, the main resonance is at 330Hz, exactly where it changes phase. I noticed that was reversed again in 8khz.
I do not know if the measurements are correct enough, but I think this is the way.
I plan on making this phase correction electronically, someone would have some circuit for this? Does the professional systems use some kind of phase correction?

Ciuens

[EpicenterBryan]

As I am attempting to understand this: the delay introduced by the [pushrod] material at a given distance from the coil would not necessarily have an adverse affect on an open loop cutterhead. [?]

That's correct in theory. However, I ran across an interesting graphic last night which I can't find tonight (for a post I was going to do). It had to do with using software to generate and apply an Inverse RIAA curve to a file, and then using an actual RIAA preamp (analog version) to decode the same. It compared the digital encode / Analog decode to a full analog implementation.

The mystery graphic I mentioned showed the resulting waveforms. The full analog encode / decode waveform was correct. The Digital encode with no phase change / Analog decode of the same waveform was goofed up. Yes, phase matters. Especially when the resulting waveform is being used in feedback to correct the original signal.

To summarize, I guess I am asking: Is phase response in a driver only crucial [in this case] as we are attempting to implement a closed-loop design?

Closed loop is a goal in my opinion. Going as far as we all have to make open loop possible with these drivers, crazy home brew mods, EQ and all - only to call it quits as it's good enough? A passing grade is still passing.

I think this whole Digital encode / Analog decode thing is where Ciuens may be running into problems - and there is a long post to digest from Ciuens and it is getting late...

Let me say this - Lots of software implementations of digital filters as I read (wish I could remember the name of the algorithm), including Audacity - phase is not implemented in the output. That means if you are using Audacity (and most others) to apply IRIAA digitally to a file (or to generate a noise source with IRIAA applied), the results when payed back (decoded from record or from a feedback source) is correct in frequency BUT NOT with respect to phase. I think this is a big issue with several guys out there who are messing with feedback including Ciuens.

This really did not occur to me until late last night. Fortunately, a long time ago I modified a phono preamp to eliminate the RIAA curve and I've been using non-processed noise in testing the drivers. Now, this is where Wayne's differential preamp boards come into play and why I'm thrilled he came out with them.

So imagine that I need to swap polarity of the phono cartridge I'm using for measuring. I could insert a cable swap adapter but all the sudden the shield is on the inside of the cable run and the signal is un- hielded on the outside... Big noise issues. So being able to run twisted pair to a differential phono input has great advantages especially while prototyping stuff... The Caruso boards have differential low level feedback inputs and can decode the RIAA to sum to the output and all that, but have fixed input impedance that doesn't match what is needed for a phono cartridge like what I'm messing with for feedback.

Anyway, there is a great post from Ciuens after this post that I'll dive into tomorrow. It's getting late.
After a quick look at his phase flips, I have some questions about what the source is and such.

Bryan

[Bahndahn]

Thank you, Bryan. I have some intellectual digestion to engage in.

Here is a paper that I found tonight that appears to be in the right direction of helping this understanding; for those who may benefit from it.

Loudspeaker Phase Measurements, Transient Response And Audible Quality (Brüel & Kjaer).pdf

[markrob]

Hi,

The type of digital filter used determines if the phase response matches the analog version of the filter. If the filter was implemented using a linear phase FIR approach, the phase response will not match. However, that is not the only way to implement the filter. In the case of my RIAA plugin, I used a IIR filter (a recursive form of digital filtering) and used the Matched Z Transform to map the poles and zeros from the S plane to the Z plane. Using this approach, the gain and phase match the analog version of the filter in the usable range of operation. The phase response matters here because you want to the responses to cancel out, resulting in flat gain and phase response, when you pass through the encode and decode process. The downside to using the Matched Z Transform is that it is subject to aliasing distortion. To get around this, you need to run the sample rate much higher than normal (96 Khz and above in the case of my plugin). Most texts on digital filter design do not recommend this type of filter for that reason. However if you are trying to match the response of an analog network, it works well as long as you understand the limits.

Mark

[markrob]

Hi,

I've attached a pdf of my MathCAD sheet showing the comparison of the analog and digital filter responses for the RIAA plugin. I used 96 Khz to present the best results. I can re-plot at 44.1 Khz if there is any interest. You can see the gain and phase response of each filter and the error comparison between the two.

Mark

[juba bc]

Ciuens hey , sorry to ask ... but the magnets feedback without isolation from each other could not cause interference or read errors on feedback ?

[EpicenterBryan]

First, I made double magnetic isolation with copper plates, as I show in pictures.

Hey Ciuens,
Copper is fine for radio frequency (RF) shielding, but for magnetic fields below about 100 kHz, you want to go back to your Mu-Metal idea. I don't think that copper is doing anything for you. I think you may be getting better isolation because your feedback coils are further from the driver and picking up less magnetic field from the driver.

But, that increased distance may be an issue now that we know about the propagation delay.

Read this section from Wikipedia:

https://en.wikipedia.org/wiki/Electromagnetic_shielding#Magnetic_shielding

Also, are you using raw pink / white noise? No riaa or iraa involved at all right?

Bryan

[EpicenterBryan]

If the filter was implemented using a linear phase FIR approach, the phase response will not match. However, that is not the only way to implement the filter. In the case of my RIAA plugin, I used a IIR filter (a recursive form of digital filtering) and used the Matched Z Transform to map the poles and zeros from the S plane to the Z plane.

I think a lot of people are using Audacity because it's free. I found an interesting thread on the Audacity forum from 2014. A guy asked if there were plans to add IIR based EQ curves to Audacity. The answer was no. But here was another interesting thing the Moderator said after the poster discussed the advantages of IIR as opposed to the FIR approach Audacity uses:

If phase shift really is important, then clamp your head in vice when listening to your music otherwise you'll ruin the experience. The wavelength of audible sound ranges from 17 m to 17 mm. That means that if your head moves by 8.5 mm then the phase of the highest audio frequencies are shifted by 180 degrees, while the phase shift for the lowest frequencies is only 0.18 degrees.

http://forum.audacityteam.org/viewtopic.php?f=28&t=77894

Bryan

[chris-zwarg]

He is right in a sense that the phase-shift caused by a filter usually makes a negligible difference for the listener, provided that it is identical in both stereo channels (or else the stereo soundstage breaks down) and does not change over time (or you get "flanging" artefacts). The ear/brain instantly adapts to almost any stationary phase-shift as long as it isn't so extreme to produce audible echo. However, (ab)using several phase-shifting filters in a row can smear the impulse response of your system badly enough to make a noticeable difference, which is why digital declicking/denoising of analogue disc recordings works better on a "flat" transfer than on one first RIAA-filtered and then inverse-filtered to get it "flat" again!

In the case discussed here (feedback to dampen the mechanical resonances of a cutterhead), exact phase response is absolutely _crucial_ however: While a feedback signal at -180° phase ("negative feedback") compared to the input dampens the spurious resonances most effectively, a positive (0° phase) feedback would increase them and eventually make the construction oscillate on its own (like the acoustic feedback encountered with microphones or e-guitars placed close to speakers).

So Audacity's answer was both correct and ignorant, depending on what you are planning to do with the processed signal.....

Chris

[EpicenterBryan]

You can see the gain and phase response of each filter and the error comparison between the two.

That's impressive Mark. Very little error with your path.
The moral of the story for everyone out there who is using software to apply an iRIAA curve to your file before cutting, you might want to dive into what method the software is using and if it actually applies phase changes like an analog iRIAA filter does.

I'm starting to think we should recommend people use an analog iRIAA encooder, or contact their software vendor to verify it uses an IIR filter approach.

I will be using Caruso at some point when I do more test cuts, but I also ordered Wayne's boards to play with. Like I need another project! But man, I love those boards.

Bryan

[EpicenterBryan]

So Audacity's answer was both correct and ignorant, depending on what you are planning to do with the processed signal.....

I got you Chris.
We are using so many great tools out there in unusual ways - I don't expect developers to envision odd applications like this. But when a user brings it up and the developer acknowledges it, then shrugs it of as not mattering? Just add a button next to it that says "Emulate Analog" and use the other algorithm if selected. So it takes an extra minute to process.

That's all.

Bryan

[chris-zwarg]

Yes Bryan, you're spot-on. Probably no software programmer ever dreamt of the things we are doing. Even with dedicated audio-processing and -restoration software I have found it time and again over the years that some functions work much more effectively if you employ them in a way they are definitely NOT designed for!

That said, there are several freeware or shareware "VST" EQ plugins that allow one to choose between linear-phase and quasi-analog (phase-lagging) filtering. Or, if you still have a 32-bit Windows machine going, try to turn up a copy of good old "Cool Edit Pro" before it became "Adobe Audition" - its "FFT Filter" function that allows you to freely draw your curve has a very "analog-ish" phase response, and does a rather good iRIAA on A/D transfers made through an analog RIAA preamp. The inevitable clicks and crackles on a vinyl transfer make that especially obvious, since a mismatched RIAA/iRIAA phase-shift drastically changes their shape, rendering them unrecognizable for "declicker" algorithms like CEDAR. No idea whether more recent versions of "Adobe Audition" retain that implementation or use linear-phase filtering.

Chris