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

[markrob]

Hi,

No. The feedback loop should be a pure velocity loop (perhaps with some loop compensation to improve stability). You want the closed loop response of the head to have a flat velocity response (as best you can). The RIAA is applied to your audio chain ahead of the feedback loop. The published RIAA curve assumes your recording is velocity based. For example, if the head were flat vs amplitude, the required RIAA curve would be much different . The magnetic pickup you are using is a velocity measuring device, so when you close the loop it will attempt to make it flat vs velocity. Since you may not achieve truly flat closed loop response, you may need to add additional corrective EQ ahead of the feedback loop to compensate.

If you tried to include the RIAA within the feedback loop you would just cause stability problems given all of the poles and zeros in the forward path. Even if by some miracle, the head was stable with an RIAA network in the feedback path, the head would NOT have the RIAA response.

Mark

[EpicenterBryan]

Well, I made some progress this weekend - not as much I hoped. Big game day about to start so here is the update...

I reconfigured one of my Design 888's for -10db, calibrated in's and out's and ran a full batch of tests with RightMark Audio Analyzer. Awesome program, and free!

I got the body modified and the phono cartridge mount made, and it actually fits in there! It clears for initial test, then touches when ready to glue in place.

IMG_3285.JPG

IMG_3283.JPG

I did an initial check on the preamp. To do this I ran non-processed white noise from Protools out to an iRIAA2 - Inverse RIAA Filter made by Hagerman Audio Labs.
http://hagerman-audio-labs.myshopify.com/products/iriaa2-inverse-riaa-filter.
It is a passive Inverse RIAA with a -40 / -60 output. Then this went into the preamp with gain set to 30 (50 is too much and distortion goes up). Then back into Protools and exported to Audacity.

IMG_3287.JPG

It looked too good, but Audacity would only display one channel at a time so I decided to do a more thorough test, this time with RightMark for both channels. The frequency response shown below is from that program.

One thing to note is that the spec on the iRIAA filter says "+/-0.5dB inverse RIAA accuracy" so it does shock me that the combination of the iRIAA and this preamp nets a response like this. Too good to be true?

fr.png

Thanks the update - are you ready for some football?

Bryan

[opcode66]

Souri = Sourisseau Ulrich = Vinyl Recorder

He makes the VR

And, yes, you don't want RIAA on feedback that is being negatively summed. You do want it on feedback that is being monitored, however. Which is being done by the Caruso preamp. And, of course, you want RIAA on the signal from a tonearm. However, you want it switchable on the tonearm and on the feedback monitor so you can run a full calibration (which is why this is switchable on the Caruso). Not sure if you know what I mean by full calibration???

In Neumann amp racks, they have one RIAA that is used on the feedback monitor loop or the pickup depending on which is being monitored. The rotary switch on the monitor card changes not only what is being monitored but also what is being sent through the one shared RIAA circuit. There is a switch on the front panel that will remove RIAA and IRIAA from everything when calibrating.

If you want to easily and accurately calibrate a system, it is nice to have the ability to monitor input, feedback and pickup as well as have the ability to turn off or on the IRIAA and RIAA. You monitor output goes through VU meters. You adjust input, feedback and drive levels by sending 1K and 5K tones in and making adjustments. Then doing cut and making final adjustments. Afterwards, everything should read 0VU when fed the same signal and your pickup should read 0VU when playing the signal cut by a calibrated system. It all starts with calibrating the pickups from a NAB or similar test tone disc.

This is why I'm building a custom made phone pre-amp into my personal caruso build. Without it, the calibration procedure would be more difficult and open to error. And, I wanted to implement the same RIAA that is used for feedback monitoring. I could have used the bit that is already on the board for feedback and make it switchable like Neumann did, but, I decided to just buy the 15$ worth of parts and build a duplicate path. Either way, monitoring the feedback should be the exact equivalent as monitoring the pickup since I'm using the same exact eq.

[EpicenterBryan]

Hey Mark and Todd,

I was confused. I think I should explain about that preamp. The jumpers I added enable the RIAA. When they are removed it's just a low level preamp with no EQ. So my plan is to use it with jumpers removed, invert the signal, and sum with the audio that already has the inverse RIAA applied then run the sum to the head via the power amp.

The testing I did today was with Inverese and normal riaa, because that iRIAA test board has -40 and -60 outouts so I could feed the phono preamp with an attenuated line level from my protools system. I was so interested in how it checked out today with the filter on, I kind of fotgot what I really wanted to useit for in the first place. A lowlevel preamp for the sensor (phono cartridge).

It must be age related.

I'll build up some attenuators and check it that way...

Bryan

[Bahndahn]

Thrilled by your work, all. I've been sitting back enjoying the masterclass the three of you have been hosting. Thank you for sharing.

Here is a good look at the VR cutterhead with the feedback option. Not only is this method of feedback simply a good idea, it offers Souri's customers to have their dynamic cutterheads upgraded to a feedback version without too much trouble.

Here is someones promotional video for their record cutting service where said design can be seen. I am not affiliated with this company:

[EpicenterBryan]

Tremendous progress the last few days. So much progress and discoveries, I'm going to need to spread it over several posts over several days before I forget the sequence.

Tonight, let's get to it... going back to that phono preamp board, I discovered I did not understand the full implications of adding jumpers to remove the RC networks from the feedback path.

This preamp uses a non-inverting op-amp configuration, so when the RIAA curve is active, it uses a baseline gain at the low frequencies, and applies gain reduction with the RC combinations in the feedback path to accomplish the RIAA curve. When I removed the RC combination (by making them jumper selectable), the flat gain increased by 20db. So I commandeered the resistor and switch setting that is normally 50db gain WITH RIAA, and changed it to a 40db gain when RIAA is OFF (my jumpers removed).

The goal was to take the cartridge output level of this Audio Technica AT3600 cartridge which has an output of 4.2mv @ 1khz - 5cm/s and convert it to a -10db signal level. So, based on the following, I made the change.

Vin: .0042v
Vout Desired: .3162v
R Feedback: 750K
R1 calculated: 10.095K (10K 1%).

Once the change was made I ran a frequency response comparison between my 888 interface and the preamp with my modifications and no EQ. You can see part of the preamp response is also biased by the +0.5db response of my interface at high frequencies. So I am happy with the response of this preamp.

Notice that the graphic says 30db gain... that was an error when I captured the data - that is with my 40db mod...

fr.png

The next step was to mount the cartridge on an adapter inside the head as shown in earlier posts, with no physical connection to the torque tube, using the preamp modified as discussed above with no drive signal to drivers in the head. The signal was around -60db. Most of the signal and noise was 60hz related, and associated harmonics.

The next test changed the entire focus of testing and provided the first round of discoveries:

While everything was running, I tapped on the head with the heavy end of a screwdriver. The signal jumped to -20db (a jump of 40db). At this point it was clear that I needed to isolate the pickup from the body of the head. So, I revised the mounting as shown below by using a piece of Sorbathane between the cartridge mount and the actual cartridge. After the modification, I did the same tapping - at or harder, and the signal only went to -32db. So it was clear that I could gain 12db or so of feedback usefulness if the feedback sensor was isolated from the head.

IMG_3288.JPG

This also involved cutting deeper into the housing than I planned because of the thickness of the Sorbathane. There were several rounds of fit checking, and yes, I sacrificed the LED lighting for now...

The next round of testing got a LOT more involved, and resulted in WAY too many changes than I can go into tonight (place a day of commercial interruptions here).

Tune in tomorrow night for the next installment.

Bryan

[EpicenterBryan]

The next check was to install the cartridge so that the stylus was not touching the torque tube. Then to connect the output of the cartridge to the preamp and see what the level was like when it was in place and there was no audio to the head.

Now, a few things about the next photos... Shown here is a Protools plugin that's called a noise meter. It's really just an extended range VU meter with the ability to apply filters. The filter was off. Also, the actual VU meters in this Protools system are calibrated such that 0 is actually -18db if the -18 meter level button is pressed in the plugin display. It's by Digidesign's instructions - That provides headroom before clipping of something like 24db as I recall... Anyway I believe the noise meter plug in shows actual level, while the VU meter plugin when it shows 0, it's actually -18db (or whatever the meter level button is selected). I will have to verify that but again, don't fuss too much about this right now.

OK, so the first photo below is of the signal level with the cartridge installed, not touching the torque tube, and with no audio driven to the head. The meter is consistent with tests I did on the preamp on it's own. So far so good.

no audio noise.JPG

The next check was with the drivers being driven, still no physical connection to the torque tube. Raw White noise was used - no eq. Ideally I would love to have had no change in level. That would mean that the cartridge was not outputting anything related to magnet coupling. But such was not the case (as we all kind of suspected would be the case), but how bad was it?

This is where I made a mistake in my setup - it's not catastrophic and the data is still useful but... I intended to drive the head with 1 Watt. But I connected the meter to the amp and not the other side of the RC network so the head was really getting 1/2 watt when I intended it to have 1 watt...

If I had thought this through, I should have taken data at several levels. I did not.
Anyway, here is what the level was:

Half W drive, no connection white noise.JPG

This is where it gets interesting. While this signal was being outputted, I took a wrench and inserted it inside the head near the side of the cartridge (but not touching it), and from about 15 feet way from the screen I could see the meter for one channel dropped. I don't know by how much but I knew that I was affecting the induced magmatic pickup by interrupting it with a metal object without touching anything. So I decided I might be able to deal with that.

Although the guts of the head was all glued in place as it was constructed, I came up with a scheme to insert some Mu-Metal in there to try to cut down the magnetic stuff going on. Ideally, this should have been done on the driver before stuff got installed but it is what it is. All in all, I put 5 layers of Mu-Metal in there, and although the cartridge had already been glued to the Sorbathane and mount, I also tossed a layer on the cartridge where I could. More than one layer on the exposed section of the cartridge wasn't practical from a space standpoint, and the cartridge still needed to be isolated by the Sorbathane so nothing was added on that mating surface.

IMG_3299.JPG

Another note - I used a screw driver to feel the tug (magnet effect) as I started to insert the Mu-Metal. By the time I put the first layer or two in there, I had to switch to a paper flip with half of it extended to hold onto like a divining rod. I have never used this Mu-Metal before and was shocked at how well it worked. The paper clip still pulled to the center of the drivers, and also oddly enough to the edge of the Mu-Metal layer intersections. Also to the push rods and especially to the ends of the push rods. Yes, I used spring steel for the push rods. I had researched Beryllium Copper (AKA Spring Copper) which is non-magnetic and it wasn't something I could buy in town while I was building this so I didn't go that way. I sure wish I had!

This is what the monstrosity looked like on the next test:

IMG_3301.JPG

CONTINUED ON NEXT POST

[EpicenterBryan]

CONTINUED FROM PREVIOUS POST

Referring the to previous post, here is the level under the same conditions once the Mu-Metal additions were made (again with 1/2 W drive, no physical connection to the torque tube blah blah):

Cart with shield Half W white noise.JPG

Notice the level dropped by 12 to 14db or so. The take away is that by applying the shield I'm confident that there is at least 12 to 14db of usable feedback that I am confident has no relationship to magnetic coupling (at least with 1/2 watt drive). Is it enough to fully compensate for the response of this head? Not a chance. Is it useful? Perhaps, but only if the two big resonances are dealt with ahead of the signal chain? Comments?

The weekend is approaching and I should have 2 days of glorious uninterrupted time to play. So I ask the crew out there for inputs...

Knowing that everything is now glued in place, is there anything you can think of I should do?

Should I add more Mu-Metal in layers as I have already done (no way to compare before and after)?
Should I add doughnuts of Mu-Metal to try to close the holes near the driver push rods (no way to compare before and after)?
Should I add Sorbathane to the body of the cartridge where possible (no way to compare before and after)?

Please chime in.

There is one thing I have already done that I have not posted yet, and that was a quick check of the frequency response. I can say that the response looked similar to my original post but did not include all the low frequency rumble from my lathe. Is it real? Probably not based on the 12 to 14db of usable signal I have seen by using the Mu-Metal.

What say you? Should the home team go for it, or punt this idea?

Bryan

[Stevie342000]

It will be interesting to see how this develops but I suspect it is dead in the water but we will see I am keeping an open mind.

How is the feedback done on Souri's head? As it is an after market fit or upgrade, I have never seen one of his heads with or without feedback in the flesh and watching a video on YouTube does not give much more information.

You may be going against the grain as it is the magnetic flux that you are or should be using to influence the feedback to get the head to cut more flat and to compensate for any peaks or troughs.

Which leaves me with a couple of questions:

1: Why did you not wind a feedback coil on the main coil former (this is what I meant by using the magnetic Flux)
2: Is all this extra sheilding not really dealing with the object of how to apply feedback.
3. This is useful even if it is a dead duck in the water as others may think of alternative methods or just not go down this route as it is a dead - as you may or may not demonstrate further down the line.

Still interesting thanks for sharing your developmental experiences as this relates back to 3 again, it helps others to save money and time when attempting to develop their own cutter heads mono or stereo with or without feedback.

[markrob]

Hi,

Good work on the experiments. One thought. Rather than use a meter to make your measurements, it would be good to see the spectrum via an RTA. This will tell you how the crosstalk energy is distributed. For example, you might find that the crosstalk is far down at low frequencies and much less at high frequencies. The other measurement that would be good to know is the spectral response with the pickup stylus physically coupled to the torque tube. If you do this with the drive conditions held constant between the two measurements, you can see how much room for feedback you have over the entire spectrum.

At some point, you will have to plot the open loop gain and phase to see if you can close the loop with no compensation.

Mark

[EpicenterBryan]

Here is a good look at the VR cutterhead with the feedback option. Not only is this method of feedback simply a good idea, it offers Souri's customers to have their dynamic cutterheads upgraded to a feedback version without too much trouble.

Wow Bennett, having watched that video it jogged my memory. I DID see a good photo of that and actually saved it in a folder called "Stereo head possibilities" on my desktop about 5 months ago! I don't know where this photo came from but I'll post it here, side by side with the non-feedback version which I also had saved...

Non feedback first:

00U0U_iSoPDyHwOTM_600x450.jpg

Feedback:

pure-phase1.jpg

WOW!!! If that doesn't look like a phono cartridge with the stylus end replaced with something.
The funny thing is I looked at using an "all-in-one" type cartridge where the head shell and cartridge are one piece like below but I decided it added mounting complexities in my situation and added cost since I had no clue if this would work or not so I opted for a $24 test solution with the Audio Technica cart for testing. Here is a shot of a typical all in one these days... (of many out there).

Numark Cartridges- CS 1 Phenomenal Output and Superb Tracking. [large view].gif

But the more I look at that photo, the more it looks like something else - perhaps a Bang and Olufson? Are there any "playback" guys out there who might be coaxed into chiming in?

Again, a fabulous spot on your end Bennett. You young pups need to chime in as much as possible! I for one appreciate your inputs! Man, when I saw the pedestal drop down, I was just thinking connection for heater wires as I had seen before. Then you mentioned feedback, and there it was in another photo. Way to go Bennett!

B

[EpicenterBryan]

Which leaves me with a couple of questions:

1: Why did you not wind a feedback coil on the main coil former (this is what I meant by using the magnetic Flux)
2: Is all this extra shielding not really dealing with the object of how to apply feedback.
3. This is useful even if it is a dead duck in the water as others may think of alternative methods or just not go down this route as it is a dead - as you may or may not demonstrate further down the line.

Hi Stevie342000,
1). I didn't build the drivers. I bought those off the shelf as "surface wave transducers" and have just used them as-is on this project. I am not smart enough, and my eyes are not good enough to be able to modify them to include feedback coils and be sure the additional feed back coils are magnetically isolated from the drive coils to make internal feedback possible. Todd (opcode66) has been working on feedback on his own coils / drivers. It sounds complicated and is too involved for me but I think he is on track...
2). The extra shielding was an attempt to modify what I have already built to have feedback. But this feedback attempt is mechanical - it is an attempt to measure the actual movement (as close as possible) at the cutting stylus position. The idea was to use a phono cartridge and have the playback stylus glued right where movement is generated to excite the cutting stylus. If a measurement could be made there with a phono cartridge, and used as feedback, that could be used to generate an "error in movement" signal related to the cutting stylus, to be used in a feedback system to correct for errors in frequency response.
3). I agree. I am a hobbyist and learning as I experiment. That is why I only post in the "Experimenter's forum" (usually). I do enjoy sharing what I learn along the way, and especially like inputs from people as we all learn, or share what we have discovered. As it turns out, most of the time what I "discover" other people have already discovered. So I'm not sure how useful my posts really are but they will continue...

This is a great place!

B

[Stevie342000]

Thanks Bryan for the clearer pictures of Souri's head with and without the feedback, it makes it all clearer. As does the refresher that you did not wind your own coils, I assumed wrongly that you had made your own (its been a long thread and memory fails us all over time).

I get what you are trying to achieve and if Souri did it then you should be able to achieve it as well, with the right amount of magnetic shielding, I am sure you are making extensive notes so you can replicate your head design in the future.

If I understand this right then you are trying to figure out how to:
1. shield the feedback set up (cartridge) from stray magnetic fields?
2. the best way to connect the stylus shank to the actual torque tube?
3. at the present time you are trying to determine how much an influence the magnetic field is on the feedback mechanism before you add the feedback into the system (by connecting the feedback set up to torque tube)?

Nevertheless it is all very interesting, yes I get it with the failing eyesight and potential hands like shovels when working on such a small accurate scale that this field of development requires.

We know from heads that were made originally there there is more one method to apply feedback, most are based around the Westrex Patents especially for stereo which is odd really when you consider that (it's debatable) Blumlein was hired by EMI to circumvent the Bell labs patent in the early 30s which he did so they had to stop paying a fee to AT & T and that when they revisited stereo disc recording they found Blumlein had got their first. However the Grampian heads are not true feedback types even though they use a form of feedback (developed in the early 1940s by the BBC in the UK) - Grampian were licensed to build them as Cecil Watts (MSS) was licensed to make the actual recorders in which he helped to develop as well.

Today of course it is possible to do it in the digital domain, not just in the physical real world, what you may have issues with is (if you try to replicate your head) the standard of performance from one head to another head of the same manufacture/design. Which will lead of course (hopefully) to further refinements.

Keep up the good work guys, it is much appreciated, there should come a point in the not too distant future when we will have the options through the hard work of others of building our own systems from the ground up through the designs you have all worked so hard on. A go to thread where a list of all the designs available would be of use, so new users can quickly look up the how too....

Back to the pictures of Souri's head in the non feedback head what are those two small coils on the paxolin board for? Are they not some kind of feedback mechanism to flatten the head response out?

[EpicenterBryan]

If you do this with the drive conditions held constant between the two measurements, you can see how much room for feedback you have over the entire spectrum.

Hi Mark,

I'll going to tackle that this weekend. Today I'll do tests with the stylus glued (it's already glued), tomorrow I'll break the stylus loose and run the same tests.
Here is the plan:
Test at 1/2W, 1W, 1.5W, 2W.
Single frequency 1K, 5K, 10K
White noise
I'll also run a full set of tests with RightMark.

Is there anything I'm forgetting?

Bryan

[Stevie342000]

Would a full frequency sweep with and without feedback be in order, or am I missing something?

[EpicenterBryan]

Would a full frequency sweep with and without feedback be in order, or am I missing something?

That was the goal of the white noise, but no big deal to do a sweep as well. 20hz-20Khz it is.
Thanks for the inputs.

Bryan

[Stevie342000]

No problem, it helps my understanding in what you are trying to achieve and to make sure I am following what you are doing.

[EpicenterBryan]

Wow, what an interesting weekend.

Saturday I prepared files for testing. First thing Sunday I did a few quick checks that changed everything. Before acquiring ANY data I decided to check a few things.

This cartridge has an output of 4.2mv @ 5cm/s with a 1Khz signal. So I decided to use the cartridge (with stylus glued to the torque tube) to see if I could set the amp to get exactly 4.2mv from the cartridge and determine what the power value was at the head as a baseline. Adjusting the signal to the head to get 4.2mv from the cartridge resulted in 0.755V. With the driver being 4.7 ohm, that works out to 0.1212W to produce a signal equal to 5cm/s with a 1Khz signal. Roughly 1/8th of a watt to get 5cm/s? Then I adjusted power at the head for exactly 1/2 watt (1.53V at the head). The cartridge outputted 8.54mv. And then I pumped 1W into the head (2.167V at the head), and the cartridge said 11.64mv. All these numbers looked bogus - but were they?

Let me explain why the numbers made no sense. Going back to an earlier post, where I actually cut a 1KHZ signal at 2 Watts (when I thought it was actually at 1 watts), using a different phono cart for playback the signal was measured as a staggering 9.54mv right at the cartridge. You can see from above why those numbers made no sense.

Now realize that the test I mentioned early on in this post was done by playing back an actual disc on a turntable. In the testing this weekend I was trying to measure the movement right at the torque tube bypassing the cutting process. Since the numbers made no sense, I broke the pickup stylus loose, and removed the cartridge to look further into it. I then built a test jig that I will discuss later so I could investigate this further.

Since the signal was so much larger than I expected, and realizing the pickup stylus is not exactly where the cutting stylus is (for obvious reasons), I decided to calculate the predicted movement at the location further down the torque tube. Below is the diagram with some dimensions and calculated numbers.

Let me explain the diagram so you can follow along and double check my calculations.
The Yellow line is the torque tube. D is where the drivers are connected, C is where the Cutting Stylus is located, and P is where the Pickup (phono cartridge) stylus is located. The blue line is the point where the V-spring and piano wire is located. The Pink lines represent the peaks (left and right) for lateral stylus movement, and defines the full movement angle Theta (in degrees) for what I'm trying to calculate below.

IMG_3323.JPG

So if I want to cut at location C with a velocity of 5cm/s @1khz the arc length (S) at location C (SC) should be 0.050mm for 1 cycle @1Khz. With that arc length, and the distance (SC), angle (Theta) can be calculated. To calculate the arc length at the pickup location which is 3.5mm further down the torque tube and using the same calculated angle results in a longer arc length in the same amount of time - thus the velocity calculated at that location is higher - being 5.875cm/s. Before I did the arc calculations, I used right angle triangles as a quick estimate using the old school SOH CAH TOA trick and the number still worked out to 5.88cm/s, so I'm confident my calculations are correct but feel free to check me.

OK, so why was the measured velocity so much higher than expected? Well, it could be that there is simply so much magnetically induced signal in there, or something else I don't understand like where the shielding diverts the field, or the steel push rods acting like magnetic conduits .. OR, could it also be because the head was not cutting as disc? No, could it be because the torque tube WAS actually moving further because there was no mechanical load?

So I decided to build this test jig...
More on this tomorrow.

IMG_3318.JPG

IMG_3321.JPG

[markrob]

Hi,

The basic calculation seems fine to me, but I'm not getting your basic numbers. The relationship between velocity, excursion, and frequency should be:

E = V/(2 x pi x f)

or

V = E x 2 x pi x f

Where V is the velocity and E is the excursion. Note that the 5 cm/sec reference is the rms value not the peak (rms is what your voltmeter reads as well). The peak 1.41 x RMS (Square root of 2). So for 5 cm/sec rms, you really have 7.07 cm/sec peak velocity. If you put that into the first formula, you get E = .001125 cm peak excursion. Peak to peak would be twice this value or +/-.00225 cm = +/-.0225mm.

Your next test looks like a good one as you will be able to remove any coupling given the distance from the driver.

[dubcutter89]

Hi Bryan,
Congrats on your project! Real DIY spirit!

About your measurements:
-I would recommend you to switch from (cm/s per Watt) to (cm/s per Volt) as this is what I found on most cutterhead papers and also personaly feel more comfortable with as it makes calculations independent from load (driver) impedance.

-You also try to measure sensitivity but your cutterhead has no flat response (and maybe isn't linear either). So your numbers are only valid for one frequency/level... Also 1KC may be to close to resonance to give realistic numbers, especially if comparing in air (no load) and on disk. (This is what I think is the problem in above post..)

-Your calculations on the geometry of driver/stylus/pickup are right (at first sight...) and show one thing. The ratio of excursion vs length of torque tube is so small that you don't have to worry about that (to much). Theoretical difference between 5 and 5.875 is only 1.xx dB... There is a ton of other things that will also influence the result (Pickup angle, shielding, couplings, secondary resonances...) and are probably to complex to analyse. Go ahead with your numbers and make the best out of it.

Lukas