As long as the output of the driving stage behaves resistively, you can just take its Thévenin equivalent, see that the stage's output resistance is right in series with the first resistor of the network and reduce the first resistor to correct for that. The fact that the voltage at the output of the driving stage becomes frequency-dependent doesn't matter.
Add an AC coupling capacitor with a value less than 100 times the sum of the RIAA correction capacitors, and things get a lot more complicated...
Don't those complications express themselves as a high-pass filter on the network? If so, that can be put to good use, since we're talking about vinyl record players here. What's the use of amplifying 20Hz if records don't go that low anyway? I figure if response is flat down to 40Hz, but is rolled off 1dB or so at 20Hz, that's definitely not a catastrophe for most setups.
Incidentally, it looks like a -0.5dB loss at 20Hz in that graph. Is that really a horrible problem?
Another way around this would be to use an input cascode to get the input capacitance down (D3a-triode has horribly high input C) with that driving a source follower buffer to get the source impedance down. Yes, the cascode is much less linear than a trioded D3a, but we're talking dozens of millivolts of input signal here, so linearity will not be the biggest issue (noise will be).
What would be lost by using a series R on the order of 100k ohms? Insertion loss, certainly, so there would be a noise penalty. What else?
Incidentally, it looks like a -0.5dB loss at 20Hz in that graph. Is that really a horrible problem?
Another way around this would be to use an input cascode to get the input capacitance down (D3a-triode has horribly high input C) with that driving a source follower buffer to get the source impedance down. Yes, the cascode is much less linear than a trioded D3a, but we're talking dozens of millivolts of input signal here, so linearity will not be the biggest issue (noise will be).
Last edited:
Some people put the AC coupling after the RIAA network. This requires that high voltage rated capacitors are used for the RIAA network, but that would take care of the shift in the first RIAA pole, correct?
I haven't done any calculations for that configuration, but I expect the pole will still shift to some extent. You will have to have a grid leak resistor after the AC coupling capacitor, so there is again an extra RC branch connected to the RIAA correction circuit.
In any case, you can get very close to correct passive RIAA correction without even needing awkward capacitance values (only awkward resistances) using the network and calculation from chapter 3.1 of the attachment of https://www.diyaudio.com/community/...a-calculator-formula-help.401437/post-7439558
In any case, you can get very close to correct passive RIAA correction without even needing awkward capacitance values (only awkward resistances) using the network and calculation from chapter 3.1 of the attachment of https://www.diyaudio.com/community/...a-calculator-formula-help.401437/post-7439558
Yes, that spreadsheet is nice. Also, John Broskie published an app called TCJ RIAA that works well. I don't think it's available any more, though.
With the RC after the RIAA EQ network, the reactance of the series C shouldn't interact with the EQ capacitances. It will introduce a parallel impedance, but using 1M ohms to keep that high enough should keep it from being a problem. That allows use of a reasonable value of series C, perhaps 0.22uF, or 1uF if you're worried about the extreme low frequencies.
For vacuum tube phono preamps, I've used a sort of 'standard' network with capacitors of 0.01uF, 3300pF and a cap in parallel with the 3300pF to adjust for 2nd stage input C. I adjust the value of Rs to get the first pole close to where it needs to be in simulation, then build and test. A friend built a preamp I'd chosen the RC values for and measured it on a fancier apparatus than I have. It came out very well, nice and flat. We were pleased.
With the RC after the RIAA EQ network, the reactance of the series C shouldn't interact with the EQ capacitances. It will introduce a parallel impedance, but using 1M ohms to keep that high enough should keep it from being a problem. That allows use of a reasonable value of series C, perhaps 0.22uF, or 1uF if you're worried about the extreme low frequencies.
For vacuum tube phono preamps, I've used a sort of 'standard' network with capacitors of 0.01uF, 3300pF and a cap in parallel with the 3300pF to adjust for 2nd stage input C. I adjust the value of Rs to get the first pole close to where it needs to be in simulation, then build and test. A friend built a preamp I'd chosen the RC values for and measured it on a fancier apparatus than I have. It came out very well, nice and flat. We were pleased.
Would not this be true no matter where in the signal path the high pass is located?
All good fortune,
Chris
Not when there is a unilateral amplifying stage between the RIAA correction network and the AC coupling. (Completely unilateral amplifying stages don't exist, but you can get close enough to unilaterality.)
For example, output of the input stage DC coupled to the RIAA correction network, output of the correction network DC coupled to a cathode follower, cathode follower output AC coupled to the next amplifying stage. That would be a rather silly design, though.
For example, output of the input stage DC coupled to the RIAA correction network, output of the correction network DC coupled to a cathode follower, cathode follower output AC coupled to the next amplifying stage. That would be a rather silly design, though.
I still don't understand the distinction, but because it comes from you I believe it. I'll keep hitting my head on the desk and it'll eventually come to me.
Much thanks, as always,
Chris (all strain, no brain)
Much thanks, as always,
Chris (all strain, no brain)
Can I sandwich the RIAA network between TWO cathode followers DC coupled from one to the other? Example: input pentode (or cascode) stage cap coupled to cathode follower to RIAA network DC coupled to second cathode follower and cap coupled to second pentode (or cascode or triode) gain stage and then DC or AC coupled to final cathode follower. There will be 5 tubes in series with 3 cathode followers. Those 6AN8/6BL8/6U8/7199 pentode/triode combo tubes can put to good use. Just thinking out loud...
I don't know if this helps, but I'll give you a numerical example.
Suppose you have a first stage with a resistive output impedance of 25 kohm (R0), you want to AC couple it via a 220 nF capacitor (C0) and a 470 kohm resistor to ground (R4) to a RIAA correction network using 22 nF (C1) and 2.2 nF (C2) capacitors, using the correction network
where R3B is actually 0.
When you calculate the RIAA correction values neglecting the reactance of the AC coupling capacitor, you find
R2 = 14454.54545454... ohm
R1 + (R0 * R4)/(R0 + R4) ~= 119183.428858293 ohm, hence R1 ~= 95446.0551209197 ohm
R3A ~= 23982.2825587722 ohm
When you use my more accurate approximation, you find
R2 = 14454.54545454... ohm
R1 ~= 109171.187490245 ohm
R3A ~= 23894.8717329408 ohm
When I run the first set of values, including the 220 nF AC coupling capacitor, through a pole-zero extraction program, it finds:
Poles:
-8.334 rad/s
-346.478 rad/s
-13.334 krad/s
Zeros:
-3.145 krad/s
5.116 frad/s
When I run the second set of values, including the 220 nF AC coupling capacitor, through a pole-zero extraction program, it finds:
Poles:
-8.332 rad/s
-313.815 rad/s
-13.333 krad/s
Zeros:
-3.145 krad/s
-5.24 frad/s
Ideally, there should be one subsonic pole from the AC coupling and two RIAA correction poles, one at -1/taup1 ~= -314.46540880503 rad/s and one at -1/taup2 = -13.333333... krad/s. The second RIAA correction pole is pretty close in both cases, but the first is off, especially with the first set of values.
Ideally, there should be one zero at -1/tauz ~= -3.1446540880503 krad/s, the zero of the RIAA correction, and one zero at 0, that's from the AC coupling. That's precisely where they are, but the program makes small round-off errors.
Comparing the results to ideal RIAA correction, the error is about -0.75 dB at 20 Hz with respect to 1 kHz for the first set of values, about -0.44 dB at 50 Hz. For the second set of values, it is -0.005 dB at 20 Hz with respect to 1 kHz, +0.005 dB at 50 Hz. The subsonic -3 dB frequency is about 1.65 Hz for the first set of values and about 1.33 Hz for the second, so neither of them has much inherent subsonic filtering.
I am using a flat front end ksc1815 which drives riaa op amp. It reduces gain in the op amp 6 dB while providing a match to the cart
82db signal to noise
Most common with MC. Brings MC performance to Vm15 type 2 it may roll off at 19 khz but i can't hear there anyway
82db signal to noise
Most common with MC. Brings MC performance to Vm15 type 2 it may roll off at 19 khz but i can't hear there anyway
So, within this simplified (so stage gain can be ignored for RIAA accuracy purposes) simple loss ("passive") version of the OP's question, if we want to include the IEC 20Hz high pass at this obviously desirable location we really need to include it in the calculations for the other inflection points. Drat!
Might as well just use the superior anode follower second stage with RIAA around it - not especially more difficult to calculate if open loop gain variation is controlled with cathode degeneration etc. Seems like all roads lead to LTSpice.
Thanks for going to so much trouble, very kind of you,
Chris
Might as well just use the superior anode follower second stage with RIAA around it - not especially more difficult to calculate if open loop gain variation is controlled with cathode degeneration etc. Seems like all roads lead to LTSpice.
Thanks for going to so much trouble, very kind of you,
Chris
If you would want to use the passive network, include the IEC high-pass and want that to also be accurate, then you would either have to solve the equations I didn't solve because I found them too complicated, or resort to some iterative technique - such as tweaking things in LTSpice, or tweaking component values, calculate the coefficients of the characteristic polynomial and keep changing component values until the coefficients are what you want.
Then again, as far as I know, the IEC standard with the extra high-pass has been retracted, and the extra high-pass was not compensated for during recording anyway, so the exact cut-off frequency doesn't matter much.
Then again, as far as I know, the IEC standard with the extra high-pass has been retracted, and the extra high-pass was not compensated for during recording anyway, so the exact cut-off frequency doesn't matter much.
And at that we're counting angels on heads of pins, with system transfer dominated at sub-20Hz by the two pole resonance of tonearm effective moving mass x cartridge compliance and its damping. Is it correct that that resonance is unilateral (in the sense of not interfering with each other's transfer functions)? If so, maybe the optimum choice would be to include a serious high pass filter into the mix at this point. It wouldn't be the easy way, but it'd be The Cowboy Way. Gotta love Riders in the Sky.
A happy and safe weekend to all y'all,
Chris
A happy and safe weekend to all y'all,
Chris
Last edited:
Supposing that we could include an opamp-based ("ideal") second order servo around the second stage anode follower with classic RIAA in the feedback path in parallel with the servo, or maybe even some Linkwitz-Transformed modification to the servo based on measured response (because nothing else is trustworthy in the phono cartridge, tonearm world) what simplifications of the model are possible? Let's assume a brute-force (very large c0) to keep things somewhat manageable.
* I'm not even certain that a second order servo is possible (stable). It's not a 101 topic, so even the simplest model is beyond me in this respect. There's undoubtedly an elegant answer. Hint, hint.
** My spies tell me that a measuring system including test vinyl discs and appropriate software is in some serious progress by a member and very dedicated truth seeker. If I knew any more I'd spill my guts, but it's all future-soon-crossed-fingers. The rumor is that the software is the slowest part - "I'm Shocked, Shocked! to find gambling in here!
Your winnings, sir." As they say, hardware eventually breaks, software eventually works.
All good fortune,
Chris
I'm defeated by editing involving asterisks and parenthesis's, but can't remove 'em, so what the hey. Software - shocking!
* I'm not even certain that a second order servo is possible (stable). It's not a 101 topic, so even the simplest model is beyond me in this respect. There's undoubtedly an elegant answer. Hint, hint.
** My spies tell me that a measuring system including test vinyl discs and appropriate software is in some serious progress by a member and very dedicated truth seeker. If I knew any more I'd spill my guts, but it's all future-soon-crossed-fingers. The rumor is that the software is the slowest part - "I'm Shocked, Shocked! to find gambling in here!
Your winnings, sir." As they say, hardware eventually breaks, software eventually works.
All good fortune,
Chris
I'm defeated by editing involving asterisks and parenthesis's, but can't remove 'em, so what the hey. Software - shocking!
Last edited:
Unfortunately DC servo loops around RIAA correction stages lead to several complications. One of them is (again) shifted RIAA correction poles, especially the first pole. I know of three solutions:
1. Ensuring the DC loop has very little loop gain left at 50 Hz, say magnitude of the loop gain below 0.01. For loop of which the loop gain drops at a second-order rate, that means a cut-off frequency of the order of 5 Hz (give or take a factor √2) or less.
2. Precorrecting the poles so they end up at the right place after the loop is closed. This again much complicates the calculations (or requires some iterative technique).
3. Putting a zero in the feedback network on top of each RIAA pole. When you work out this idea, you end up with a reverse RIAA network in the DC servo circuit.
1. Ensuring the DC loop has very little loop gain left at 50 Hz, say magnitude of the loop gain below 0.01. For loop of which the loop gain drops at a second-order rate, that means a cut-off frequency of the order of 5 Hz (give or take a factor √2) or less.
2. Precorrecting the poles so they end up at the right place after the loop is closed. This again much complicates the calculations (or requires some iterative technique).
3. Putting a zero in the feedback network on top of each RIAA pole. When you work out this idea, you end up with a reverse RIAA network in the DC servo circuit.
Much thanks for the cool thoughts. In the mid-1980s Yamaha (in their model C-4) did a phono stage with synthesized 47K input resistor, like you do with a flat first stage / flat feedback, but with conventional long loop opamp-style RIAA and an inverse RIAA and its own opamp from output back to the bottom of the synthesizing resistor.. Maybe this could be combined with DC servo, just in case things aren't complicated enough yet.
It wouldn't be the easy way, but would it be The Cowboy Way?
As always, Thank you,
Chris
It wouldn't be the easy way, but would it be The Cowboy Way?
As always, Thank you,
Chris
You could put a capacitor in series with the feedback resistor of the op-amp, add an AC coupling capacitor between the cartridge and the input and design the whole thing to have a response that is a combination of RIAA correction and a second-order Butterworth high-pass response.
For a 47 kohm input resistance (moving-magnet) and a 20 Hz cut-off frequency, the coupling capacitor has to be 120 nF. Its reactance will increase the sensitivity to hum currents injected into the grid of the input valve. No idea how big a deal that will be.
For a 47 kohm input resistance (moving-magnet) and a 20 Hz cut-off frequency, the coupling capacitor has to be 120 nF. Its reactance will increase the sensitivity to hum currents injected into the grid of the input valve. No idea how big a deal that will be.