I did study the Bastiaans when you generously referenced it in the Stylus Friction paper discussion. It's really separate from modeling of the generator and its load, and I shouldn't have included anything about the stylus/groove wall interface in this thread, only muddying the waters. We didn't agree then, and probably won't now, about that stuff, and we're probably both wrong (because the issue is too complex) but it was inappropriate to include it here, even tangentially. For that I must apologize.
In my simple mind, a comparison between a high impedance MM/MI generator and a very low impedance MC generator is equivalent (or maybe I should say analogous) to the topic at hand. All other things being equal, in an imaginary situation where only the generator's response is to be considered, the MC generator + load can be considered "flat" and perfect, and the MM/MI generator + load is flawed by its messy LCR-ness. In the bad old days, geometric and stylus ETM x vinyl compliance were, more or less effectively, massaged towards something almost "flat" by the LCR loading of contemporary cartridges.
Today folk seem to be concentrating on the generator + loading side of the total response, moving towards an ideal of duplicating that of a MC generator, which is almost indepedent of (reasonable) loading. This leaves the geometric and stylus/groove wall response intact, and that was my poorly expressed intent above.
Much thanks, as always, and great good fortune in the new year,
Chris
In my simple mind, a comparison between a high impedance MM/MI generator and a very low impedance MC generator is equivalent (or maybe I should say analogous) to the topic at hand. All other things being equal, in an imaginary situation where only the generator's response is to be considered, the MC generator + load can be considered "flat" and perfect, and the MM/MI generator + load is flawed by its messy LCR-ness. In the bad old days, geometric and stylus ETM x vinyl compliance were, more or less effectively, massaged towards something almost "flat" by the LCR loading of contemporary cartridges.
Today folk seem to be concentrating on the generator + loading side of the total response, moving towards an ideal of duplicating that of a MC generator, which is almost indepedent of (reasonable) loading. This leaves the geometric and stylus/groove wall response intact, and that was my poorly expressed intent above.
Much thanks, as always, and great good fortune in the new year,
Chris
Chris,
Don't apologize for this, it's o.k. to know what was included and what not.
Bastiaans was referred to because you mentioned:
Could we agree that all of these models ignore geometric losses arising from non-zero stylus dimension in the plane of travel and stylus ETM x vinyl compliance resonance, and so will not conform to real playback? (Or at least be critically dependent on the stylus/groove wall interface, so not general).
I did not create a cantilever model out of the blue, but started with the TF as the result of subtracting the Generator's response from the overall recorded response.
For a number of Carts their response after subtracting the generator's response from the overall response came out as in Fig 1.
Figure 1
The first two obvious functions being part of this mechanical response are Fres and Fcutoff , to be calculated in 1) and 2) below,
using all relevant parameters like ETM, tip diameter, position on the disk, speed, etc etc.
From there the indentation 3) had to modeled as the still missing part to fully match the TF in the above figure.
1) Fres being a function of ETM and vinyl compliance is the first part in the Cantilevers model.
fr = (0.632/πSqrt(m)*(E0²FvR)^1/6
m being the equivalent tip mass.
Eo the Young modulus, 3.76e9 N/m² for vinyl
Fv the the stylus force, usually 2 gram
R is the tip radius touching the track wall.
2) And the second part concerning the cutoff frequency is
fc = 1.51(V/π)*(E0/FvR)^1/3
V is the track speed.
fc is inversely proportional to Fv^1/3 , the higher the stylus force, the lower fc
Q of fc is at 0.88
3) Losses arising in the plane of travel was the third part in the model, called Indentation and is responsible for the level sag for most (MM) carts around 8Khz.
The model for the indentation was the still missing and remaining part after having modelled Fres and Fcutoff to match the response as in the Fig 1.
The generic model as a result of all the above looked like:
Figure 2
Since 1), 2) and 3) together produce the exact TF of the one in figure1 for a specific cartridge there is no way that geometric losses are ignored and that the mechanical model will not conform to real playback just because they were extracted from real playback.
Hope to have made this part of the model clearer as in the original posting.
Hans
Again, I must apologize for diverting the discussion away from the generator + load intended topic. And thank you Hans for taking the time to describe the workings of the Cantilever Assembly model. It's very much appreciated.
All good fortune,
Chris
+
All good fortune,
Chris
+
@Bob Cordell ,
Bob, as an attempt to make things more obvious, I have made in the image below three individual graphs, resp the Cantilevers transfer curve in isolation in Red, the Generator's response terminated with 47K//100pF in Blue and the Generator terminated with 17k42//100pF while corrected with a 8Khz zero in Green.
It's much clearer to see now what the effect is of this 8Khz pole with 17K42//100pF plus added correction in compare to the standard 47K termination.
What can be seen is that the BW has increased considerable by almost eliminating the Generators L-C resonance, exactly as targeted with the VinylTrak option in adding this 8Khz damping.
When looking at these figures, some incorrectness of Eddy currents compensation in the Generator that was questioned by you will hardly sort significant effect in the above graphs with these large differences.
However, while the Cantilever assembly has a 2dB peak at 20Khz, similar to many other completely different MM carts that I tested, the result of combining Cantilever plus Generator results in the image below, where the increased Generator's level around 20Khz, leads to an overall 2dB increase in level at the HF side.
For a VinylTrak damping option, one will need a Cart to start with having no visible Cantilever peak at around 20Khz, but even then the response may be disappointing because of the elevated Generator's response after damping.
Hans
P.S. As before I have included the .asc files
Bob, as an attempt to make things more obvious, I have made in the image below three individual graphs, resp the Cantilevers transfer curve in isolation in Red, the Generator's response terminated with 47K//100pF in Blue and the Generator terminated with 17k42//100pF while corrected with a 8Khz zero in Green.
It's much clearer to see now what the effect is of this 8Khz pole with 17K42//100pF plus added correction in compare to the standard 47K termination.
What can be seen is that the BW has increased considerable by almost eliminating the Generators L-C resonance, exactly as targeted with the VinylTrak option in adding this 8Khz damping.
When looking at these figures, some incorrectness of Eddy currents compensation in the Generator that was questioned by you will hardly sort significant effect in the above graphs with these large differences.
However, while the Cantilever assembly has a 2dB peak at 20Khz, similar to many other completely different MM carts that I tested, the result of combining Cantilever plus Generator results in the image below, where the increased Generator's level around 20Khz, leads to an overall 2dB increase in level at the HF side.
For a VinylTrak damping option, one will need a Cart to start with having no visible Cantilever peak at around 20Khz, but even then the response may be disappointing because of the elevated Generator's response after damping.
Hans
P.S. As before I have included the .asc files
Hi Hans.
I was disappointed that you did not address the concerns I raised regarding cartridge modeling in light of eddy current effects, e.g., the frequency response of magnetic flux loss due to eddy current. Instead, I did some of my own simulations using the models you supplied earlier for the AT150 cartridge with the ATN152 cantilever in Post 154 on 1/1/2025.
Your biggest criticism of my damped loading approach was that I evaluated its performance using a very simple cartridge model - one that did not take into account the frequency dependence of cartridge inductance (and impedance) due to eddy current effects. You are correct in expressing concerns over that, but I sought to investigate how big those errors are as a result of the simplification of the cartridge model. I did that by looking at differences in frequency response between your cartridge model and my simple one for both conventional and damped cartridge loading. In other words, how well does damped cartridge loading work when your more complex cartridge model is used?
I simulated the cartridge and cantilever models you kindly provided in your earlier post #154. First with conventional loading and then with damped loading.
My model is obviously over-simplified and that is why you asserted that damped loading is not a good technique, since I based its operation on that simple model. So I simulated with both your model and my simple model for both normal loading and damped loading to see how big is the difference. In other words, how much error is created by using my simple model versus your more complex model that includes loss resistances and which was derived from actual measured cartridge impedance.
In looking at conventional loading, your model used 47k and 100 pF loading. It was fairly flat (down only 0.4 dB at 20 kHz), but 100 pF loading is a bit unrealistic, since most turntables contribute 100 pF of loading with wiring and interconnect before the signal ever gets to the preamp. For your loading to yield a flat response, a preamp with input capacitance set to zero would be needed. Most phono preamps do not have selectable capacitance loading; most have a default fixed 100 pF of input capacitance. A practical cartridge must be able to operate reasonably flat with default loading, which will usually be 200 pF when wiring and default preamplifier input capacitance is in place.
While your model is flat with only 100 pF total load, it is up 0.5 dB at 17 kHz with total loading increased to just 150 pF. At 200 pF total, it is up by 1.2 dB at 16 kHz. As expected, frequency response with conventional loading is quite sensitive to amount of capacitance loading. Thus, I question your model for this cartridge because of its required small load capacitance to achieve a flat response.
This also leads me to ask what was the total capacitive load when you did the pink noise playback measurements on this and other cartridges; was the total cartridge loading in those experiments really only 100 pf, even when turntable interconnect and preamp input capacitance is considered? If so, how did you get the total capacitance to be that low?
The equivalent simple model of mine that you supplied with the same conventional 100-pF loading was up 0.2 dB at 12 kHz and down 0.3 dB at 20 kHz. Compared to your more complex model, mine was at most different by 0.3 dB. These differences are fairly small, suggesting that my simple model is not as far off as one might think.
An advantage of the damped loading approach is a significant reduction in sensitivity to different capacitive loading. However, in fairness, this comes at the expense of needing to adjust the cartridge loading resistance to be appropriate for the effective inductance of the particular cartridge. A second advantage of the damped loading approach is that it makes the generator arrangement more like a first-order system, without a second-order system resonance that some would say impairs achievable sound quality. Of course, we are still left with the unavoidable pesky resonance of the mechanical cantilever system.
I next simulated both your complex model and my simple model using damped cartridge loading. The frequency response differences between using your model and mine when using damped cartridge loading were fairly small. I did make a small 10 pF reduction in the inductance used for my model to account for smaller effective inductance at the frequencies of interest due to differences with your more complex model.
There was no difference between your model and mine at 10 kHz. Mine was up by 0.16 dB at 6 kHz and down by 0.35 dB at 15 kHz and 0.7 dB at 20 kHz. The differences were thus less than 0.7 dB up to 20 kHz, suggesting that the use of the simpler model in evaluating the viability and accuracy of the damped cartridge loading approach was not much of a problem; not enough to justify your criticism of damped loading and how I evaluated it.
Just as a matter of interest, I did simulate your model of the mechanical cantilever system frequency response for this cartridge. It contributed a maximum rise of +2 dB at 20 kHz and a null at 45 kHz. When included with your conventionally-loaded generator model, overall response with 100-pf loading had a net maximum rise of +1 dB at 16 kHz.
Cheers,
Bob
Bob,
Just a point of clarification. I believe the 'AT150' cartridge Hans is modeling is the Audio Technica AT150(MLX). The datasheet says Cload should be 100-200pF. It's been well known that many negative perceptions of this cartridge were likely due to the loading capacitance (arm wires, interconnect, input capacitance of preamp) being higher than specified -- it's much more sensitive than many other MM cartridges such as the Shure V15 series.
Just a point of clarification. I believe the 'AT150' cartridge Hans is modeling is the Audio Technica AT150(MLX). The datasheet says Cload should be 100-200pF. It's been well known that many negative perceptions of this cartridge were likely due to the loading capacitance (arm wires, interconnect, input capacitance of preamp) being higher than specified -- it's much more sensitive than many other MM cartridges such as the Shure V15 series.
These forum discussions tend to go round and round, this is "better", no - this is "better", and nobody ever defines their personal "better".
Is "better" a flat magnitude response, and if so, does that include the Cantilever Assembly response and geometric losses (which vary with radius from record center)? Does it include optimum or practically acheivable loading, which and with what tolerance? What flatness of magnitude is worth what total Q? All important design issues, seldom specified.
Is "better" a final rolloff single pole response even if at some expense to magnitude flatness, or a two pole response with possibly flatter magnitude response in-band, both overlaid with the Cantilever Assemply two poles? Geometric losses are small order, less than a ZR pole, and vary with radius, so are difficult to include.
And is "better" something available to casual DIYer's or would it require heroic measures, rare test equipment, etc.?
Just hoping that this very informative discussion stays on track, and all good fortune,
Chris
Is "better" a flat magnitude response, and if so, does that include the Cantilever Assembly response and geometric losses (which vary with radius from record center)? Does it include optimum or practically acheivable loading, which and with what tolerance? What flatness of magnitude is worth what total Q? All important design issues, seldom specified.
Is "better" a final rolloff single pole response even if at some expense to magnitude flatness, or a two pole response with possibly flatter magnitude response in-band, both overlaid with the Cantilever Assemply two poles? Geometric losses are small order, less than a ZR pole, and vary with radius, so are difficult to include.
And is "better" something available to casual DIYer's or would it require heroic measures, rare test equipment, etc.?
Just hoping that this very informative discussion stays on track, and all good fortune,
Chris
Sorry to say, but that disappointment makes two of us. The least I had expected was a spectrum plot of a recording without and with 8Khz damping.
But instead the eddy current thing keeps popping up, where you never told whether your expectations where them to be more or less and what effect this would sort.
Taking the Lorenz reciprocity theorem for fully correct and therefore adequate to address your concerns, I just could create a replacement diagram from the complex impedance without having to worry about the automatically included eddy current contribution, but from your side no comment on this.
You disagreed with my conclusions that damped loading seems to be only beneficial in very special cases and that simple models promising an extended FR don't hold in real life.
But you didn't make clear what you disagreed with the shown results. Was it that the expected FR extension didn't happen or was it because of the elevated level at the high side of the spectrum.
Flattening beyond 20Khz is nice on paper, but will not automatically lead to better sound reproduction, however elevated lift between 10Khz and 20Khz may be annoying.
And that the level will lift in this area is a simply physics, because with damping the generator will increase HF level by extended FR, thereby automatically lifting the Cart's FR between 10Khz and 20Khz because the Cantilever's FR won't change.
When you look at Fig 11 in the original article, you will notice that the graphs for this Cart where made with resp. 158pF and 478pF.
The conclusion that the Cantilever response is not dependent on termination load, I could therefore just as well use a 100pF load in the simulations in previous reply to you, just because this Cart asks for a low capacitance letting the FR come out as good as possible.
I can agree that damped loading reduces the influence of the termination cap, but at the same time it is exactly this cap in combination with the standard 47K envisaged by the manufacturer that is THE component to tune in getting an optimal FR.
So decreasing it's contribution makes it even more difficult to find an optimum.
You refer to simulations you did, but unfortunately without any image or .asc file to understand what exactly you did.
A last remark were I think this discussion could come to an end:
When having two models A and B for resp Cantilever and Generator, it could very well be when only using just one termination, that A has a bit too much and B too little getting a compound result that exactly conforms to the recorded FR for that termination,
But when using a set of very different termination loads and also completely different Carts, all giving results between recording and model within +/- 0.1dB, the chance that A and B are incorrect is mathematical impossible.
That's another way of telling that eddy current influences are fully taken care of in the models.
Hans
Last edited:
What Bob tried to show is :
I also missed graphs in Bob's replies. A picture speaks a thousand words.
The discussion seems to have become somewhat of a cat and mouse play.
What I would like to see is a simulation of a cart with 47k and optimized Cload loading vs 8 kHz damped optimized Cload loading of that same cart using the same model. Once for the simple model and once for Hans' model of that same cart.
I also missed graphs in Bob's replies. A picture speaks a thousand words.
The discussion seems to have become somewhat of a cat and mouse play.
What I would like to see is a simulation of a cart with 47k and optimized Cload loading vs 8 kHz damped optimized Cload loading of that same cart using the same model. Once for the simple model and once for Hans' model of that same cart.
Peaking in the upper octave has been a known issue with MM carts for decades. The problem is it differs from cart to cart and from setup to setup given cable and amplifier input capacitance. There is very little music energy above 12 kHz no matter what the source and vinyl is no different (see Tomlinson Holman et al). It’s doubtful this anomaly has a serious effect on sound quality, but in the spirit of making things perfect, I’d advocate for a switchable HF damping network located at the phono amp input offering 0 to -4 dB in -1 dB steps at 20 kHz. You’d need a test record to set it correctly, but if not available, just adjust it by ear. As I mentioned earlier, at the moment I just change the cart loading, which provides partial mitigation.
Just to clarify, does this mean that magnitude response from playing an actual record with actual cartridges can be predicted to within .1dB, or does it mean that the playback model can predict another model of the recording's response? I'm puzzled because that level of tolerance is super precise for any practical measurement, and the issue of verifying the recording's response arises.
Another way to ask the question is: how could one verify the recording's response without some playback mechanism? Is there an electron microscope calibrated to measure excursion, or what? That's some NASA level measurements, to an innocent eye onlooker.
Much thanks as always,
Chris
I've got a mid 1980s era Yamaha preamp (C4?) around somewhere that has switch selectable R and C inputs for MM and even some R choices for MC, all on the front panel. These wars were fought in the late 1960s, 1970s by well funded companies with research budgets and smart people who did nothing else. Sadly, little of the fundamental work survived the CD revolution - some AES papers, some lore, some references in old dusty books, but little to no fundamental work because it was all proprietary, in-house, and when the old guys left, it was gone.
We can't duplicate a Shure type V stylus cantilever anymore (beryllium clad) or a Bang & Olufson MC1 stylus cantilever (diamond with a hole drilled down the middle). But within the audio range, we'd already gotten close enough to right to have modern folk still discussing ways to at least duplicate, and hopefully improve on their work of half a Century ago.
To your point, something has to be adjustable for MM cartridges, to accommodate the variety of real world conditions, (if you can still hear up that high.) That's not me, but I might still want to make a decent transfer for other folk, of vinyl never released on CD. Maybe Scott W. (RIP) had it right - A/D asap and deal with the issues in DSP. The question of a verifyable reference recording remains.
Much thanks, as always, and all good fortune,
Chris
Hi Chris,
For each and every Cart individually, you have to take the long road of measuring the complex impedance up to 1Mhz, then translate this into a replacement model in LTSpice having exactly this same complex impedance over this BW.
Then make a recording from a validated test disk using an exactly measured R and C termination, preferably measured downwards from the headshell, thereby including all wiring, connectors, interlinks and even the preamp’s input capacity.
Then subtract the Generator’s response (made from the replacement diagram plus the exact same termination as used while recording) from the overall recorded response and there you have the response of the Cantilever’s part.
The next step is now to find in LTSpice the replacement diagram for the Cantilever assembly’s response.
This all takes at least a day get both replacement diagrams.
Now load the Cart with a completely different terminination say T2 and make a new recording.
With the replacement model consisting of Cantilever and Generator, simulate the response with the same T2 and you will see that both responses, recorded and simulated will be within +/- 0.1 dB, probably the result of having made all previous steps with great accuracy.
So each Cart has to be processed idividually in this same way to get a model that can predict the Cart’s response within +/- 0.1dB.
While each Cart has different parameters you can only use each specific model for a single Cart.
But my objective was to prove that the Cantilever has no interaction with the termination, and this could indeed be proven as a general outcome being valid for all Cart’s.
So a very low termination does not influence the Cart’s behavior and DC flowing through the Cart does not put the Cantilever under any stress.
Hope this answers your question.
Hans
P.S. to adjust a MM Cart, my view is to keep the 47k to start with and tune the capacity to get a flat response.
You will need a test disk for measuring the response.
For each and every Cart individually, you have to take the long road of measuring the complex impedance up to 1Mhz, then translate this into a replacement model in LTSpice having exactly this same complex impedance over this BW.
Then make a recording from a validated test disk using an exactly measured R and C termination, preferably measured downwards from the headshell, thereby including all wiring, connectors, interlinks and even the preamp’s input capacity.
Then subtract the Generator’s response (made from the replacement diagram plus the exact same termination as used while recording) from the overall recorded response and there you have the response of the Cantilever’s part.
The next step is now to find in LTSpice the replacement diagram for the Cantilever assembly’s response.
This all takes at least a day get both replacement diagrams.
Now load the Cart with a completely different terminination say T2 and make a new recording.
With the replacement model consisting of Cantilever and Generator, simulate the response with the same T2 and you will see that both responses, recorded and simulated will be within +/- 0.1 dB, probably the result of having made all previous steps with great accuracy.
So each Cart has to be processed idividually in this same way to get a model that can predict the Cart’s response within +/- 0.1dB.
While each Cart has different parameters you can only use each specific model for a single Cart.
But my objective was to prove that the Cantilever has no interaction with the termination, and this could indeed be proven as a general outcome being valid for all Cart’s.
So a very low termination does not influence the Cart’s behavior and DC flowing through the Cart does not put the Cantilever under any stress.
Hope this answers your question.
Hans
P.S. to adjust a MM Cart, my view is to keep the 47k to start with and tune the capacity to get a flat response.
You will need a test disk for measuring the response.
Last edited:
I have an AT440ML cartridge that sounds right with only the cable capacitance and 47K.
Record companies vary in how much high-frequency EQ they apply during mastering. Some companies (Arista) put out bright-sounding records.
Ed
Record companies vary in how much high-frequency EQ they apply during mastering. Some companies (Arista) put out bright-sounding records.
Ed
Should I then interpret this to mean that all measurements are considered relative to a chosen test disk, and not absolute (relative to an abstract idealized "perfect" disc)? This may seem obvious to you, but I'm very keen to understand.
Much thanks, as always, and all good fortune,
Chris
Much thanks, as always, and all good fortune,
Chris
Last edited:
Yes. I think that getting flat response from records involves kludges on both the recording and playback sides. I can hear the variations among record companies.
Ed
Ed
This would of course be perfectly valid for developing the model, and my question only relates to the general application of the numbers discovered.
Much thanks,
Chris
There is so much variation on the recording, cutting the master and the playback and we are still within 2dB at the top end of the upper octave. Should we even be agonising over this?