I am still experimenting with different phono preamp topologies. This is the one I intend to build next (and after that -- a shunt feedback RIAA phono stage).
Brief description:
The highest "normal" signal from an MM cartridge that we should ever see is about 166mV RMS or 235mV peak (Self, 2018), although this value is presented as an outlier, and it is suggested that most records played with a normal hi-fi cartridge should only present levels up to 60mV RMS (~85mV peak). Finding the expected value of spurious signals (i.e. clicks and pops) was a bit more difficult, some searching (including various discussions here) suggested that it could be as high as 400mV peak (Clappers et al., 2020). The gains of the two stages were selected such that the first stage never clips before the second.
The 75us corner of the RIAA de-emphasis curve is implemented passively after the 1st stage using R19 and C8 in the hopes that this would be enough to prevent the high-frequency content of pops and clicks from overloading the 2nd stage. This and the other two poles were calculated using Peter Baxandall's equations (Baxandall & Lipshitz, 1981). The component values were selected to give reasonable values of gain resistors R2 and R4 (they are relatively small to minimise their noise contribution, but not so small as to strain the gain stages' ability to drive the feedback network). Another design objective was to avoid the use of many different capacitor values for the RIAA EQ. As presented, four capacitors of the same value are used (C10 is three 33nf capacitors in parallel). The values of RIAA components were tweaked a bit after simulating the circuit in LTSpice: for example, the best standard value for R4 is 390 rather than 360Ohm, but I reduced it to slightly increase the total gain. Theoretically this impacts the low frequency response, but in this case any aberrations are irrelevant since the circuit includes some infrasonic filtering.
The values of C7/R20 and C2/R3 were selected to give a useful degree of infrasonic attenuation without impinging on the musical content too much. We're down 1.4dB at 20Hz and 0.6dB at 30Hz, and between 42Hz and 200kHz we're within +/-0.25dB of the target curve (assuming that all components are bang on their nominal value). The degree of infrasonic attenuation ranges from 16dB at 4Hz to 53dB at 0.55Hz -- the region between 0.55Hz and 4Hz is considered the most problematic when it comes to artifacts produced by warped discs, etc (Self, 2018). While this is definitely not the best infrasonic filter out there, it is still much better than no infrasonic filter at all, and I did not want to include another gain stage.
R21 and C9 are the "Marcel damper" proposed by @MarcelvdG which can be useful if this phono preamp is implemented using opamps from the OPA165x family, or in general for filtering RFI collected by the tonearm cable.
I modeled this circuit using LT1115 because the model is already available in LTSpice, however it can be implemented using many different gain devices. I want to implement it using two different discrete opamps:
https://www.diyaudio.com/community/threads/discrete-opamp-with-jfet-cascode-input.427760/ - first stage
https://www.diyaudio.com/community/...e-or-going-crazy-with-discrete-opamps.427868/ - second stage
Bibliography
Baxandall, P. and Lipshitz, S., 1981. Comments on "On RIAA Equalization Networks" and Author's Reply. Journal of the Audio Engineering Society, 29(1/2), pp.47-53.
Clappers, M. et al., 2020. Overload margin in Phono pre-amps. Available at: https://www.diyaudio.com/community/threads/overload-margin-in-phono-pre-amp.356112/ (Accessed 2025/05/20.)
Self, D., 2018. Electronics for Vinyl. New York: Routledge.
Brief description:
The highest "normal" signal from an MM cartridge that we should ever see is about 166mV RMS or 235mV peak (Self, 2018), although this value is presented as an outlier, and it is suggested that most records played with a normal hi-fi cartridge should only present levels up to 60mV RMS (~85mV peak). Finding the expected value of spurious signals (i.e. clicks and pops) was a bit more difficult, some searching (including various discussions here) suggested that it could be as high as 400mV peak (Clappers et al., 2020). The gains of the two stages were selected such that the first stage never clips before the second.
The 75us corner of the RIAA de-emphasis curve is implemented passively after the 1st stage using R19 and C8 in the hopes that this would be enough to prevent the high-frequency content of pops and clicks from overloading the 2nd stage. This and the other two poles were calculated using Peter Baxandall's equations (Baxandall & Lipshitz, 1981). The component values were selected to give reasonable values of gain resistors R2 and R4 (they are relatively small to minimise their noise contribution, but not so small as to strain the gain stages' ability to drive the feedback network). Another design objective was to avoid the use of many different capacitor values for the RIAA EQ. As presented, four capacitors of the same value are used (C10 is three 33nf capacitors in parallel). The values of RIAA components were tweaked a bit after simulating the circuit in LTSpice: for example, the best standard value for R4 is 390 rather than 360Ohm, but I reduced it to slightly increase the total gain. Theoretically this impacts the low frequency response, but in this case any aberrations are irrelevant since the circuit includes some infrasonic filtering.
The values of C7/R20 and C2/R3 were selected to give a useful degree of infrasonic attenuation without impinging on the musical content too much. We're down 1.4dB at 20Hz and 0.6dB at 30Hz, and between 42Hz and 200kHz we're within +/-0.25dB of the target curve (assuming that all components are bang on their nominal value). The degree of infrasonic attenuation ranges from 16dB at 4Hz to 53dB at 0.55Hz -- the region between 0.55Hz and 4Hz is considered the most problematic when it comes to artifacts produced by warped discs, etc (Self, 2018). While this is definitely not the best infrasonic filter out there, it is still much better than no infrasonic filter at all, and I did not want to include another gain stage.
R21 and C9 are the "Marcel damper" proposed by @MarcelvdG which can be useful if this phono preamp is implemented using opamps from the OPA165x family, or in general for filtering RFI collected by the tonearm cable.
I modeled this circuit using LT1115 because the model is already available in LTSpice, however it can be implemented using many different gain devices. I want to implement it using two different discrete opamps:
https://www.diyaudio.com/community/threads/discrete-opamp-with-jfet-cascode-input.427760/ - first stage
https://www.diyaudio.com/community/...e-or-going-crazy-with-discrete-opamps.427868/ - second stage
Bibliography
Baxandall, P. and Lipshitz, S., 1981. Comments on "On RIAA Equalization Networks" and Author's Reply. Journal of the Audio Engineering Society, 29(1/2), pp.47-53.
Clappers, M. et al., 2020. Overload margin in Phono pre-amps. Available at: https://www.diyaudio.com/community/threads/overload-margin-in-phono-pre-amp.356112/ (Accessed 2025/05/20.)
Self, D., 2018. Electronics for Vinyl. New York: Routledge.
Last edited:
Perhaps it's not worth it bothering with discrete implementations after all. Maybe I will use something like this as the first stage.
I experimented with different opamp models in simulation, OPA1611 seems to give the least noise and distortion, but there is a hint of ringing in square waves and it seems to have the smallest phase margin. LM4562 and NE5534 are slightly worse (very slightly) but there are no indications of possible instability.
For the second stage I will use a rail-to-rail opamp to get the best possible overload margin.
Also since I'm using opamps anyway, it makes sense to remove the CR filters and use a separate 3rd order Butterworth high-pass filter instead.
I experimented with different opamp models in simulation, OPA1611 seems to give the least noise and distortion, but there is a hint of ringing in square waves and it seems to have the smallest phase margin. LM4562 and NE5534 are slightly worse (very slightly) but there are no indications of possible instability.
For the second stage I will use a rail-to-rail opamp to get the best possible overload margin.
Also since I'm using opamps anyway, it makes sense to remove the CR filters and use a separate 3rd order Butterworth high-pass filter instead.
It might be of interest to look at Erno Borbely's 2 stage. I built Joe Tritschler's "revisited" tube interpretation of it and like it a lot.
Look at NJR2068 very low noise and distortion, sounds great on my RIAA circuit better then my 2 stage.
Last edited:
Remove C1 and replace R1 with 5...8 volts LEDs chain from Q1Q3 bases to Q2 collector [
Last edited:
The simulation model has no realistic source impedance, the 500 mH or so of the moving-magnet cartridge is not included. As a result, you don't see the effect of the OPA1611's large equivalent input noise current (1.7 pA/√Hz according to its datasheet). The same holds for the LM4562 (1.6 pA/√Hz if you are lucky). The NE5534A (0.4 pA/√Hz) and FET op-amps (almost no noise current) are much better in that regard.
I thought that adding a JFET front-end would largely mitigate any issues with input current noise? With a 500mH inductance in series with the input, the differences in V(onoise) between different opamps are very small. Except that the LM4562 seems to have a higher 1/f noise than the others (using the model from TI's website).
Nick, I am a member of your Patreon but not of your YouTube channel, is this video also available to your Patreon subscribers?
You are absolutely right, I completely misinterpreted your post. I thought you wanted to use an op-amp instead of a discrete circuit, but you are going for a hybrid solution.
Thanks for the link, that was interesting (although to be honest I don't really appreciate the amount of vitriol and just general unpleasantness generated in various ex-USSR communities). By the way, I completely agree with you (and also with Douglas Self) that a single stage opamp-based phono preamplifier is not ideal because it has too much closed loop gain at low frequencies (not enough negative feedback = increased distortion) and not enough closed loop gain at high frequencies (too much negative feedback = potential instability).
It seems that the biasing scheme using LEDs between the bases of the cascode transistors and the CCS works only with high luminosity LEDs that require only very small currents for normal operation. I don't have any such LEDs but I have literally hundreds of standard green LEDs, perhaps I will bias it like this:
This first stage appears to be unconditionally stable with no frequency or phase response peaking and phase margin of ~72 degrees:
It seems that the biasing scheme using LEDs between the bases of the cascode transistors and the CCS works only with high luminosity LEDs that require only very small currents for normal operation. I don't have any such LEDs but I have literally hundreds of standard green LEDs, perhaps I will bias it like this:
This first stage appears to be unconditionally stable with no frequency or phase response peaking and phase margin of ~72 degrees:
Last edited:
This is well-known. The better phono pre-amps have two gain blocks.stratokaster83 said:
Ed
No, that's not quite true. In the circuit quoted in my video, LEDs are used to tie the bases (and therefore the emitters) of the cascode transistors to the sources of the field-effect transistors in order to compensate not only for the Miller capacitance but also for the input capacitance of the field-effect transistors of the input stage. Approximately the same compensation of input capacitance is used in the Accuphase С-37\C-47 phonopreamp [ https://www.patreon.com/posts/analiticheskii-73044002 ], but the implementation is much more complex.
The best phono-preamps have 3 gain stages and Output TransAdmittance stage with ActiPassive constant loop gain RIAA EQ [ https://www.patreon.com/posts/otsenite-tekst-90783025 ], [ https://imrad-com-ua.translate.goog..._sl=ru&_x_tr_tl=en&_x_tr_hl=ru&_x_tr_pto=wapp ]
Last edited:
This compensation is probably necessary for the LSK170 with its big fat die, or for many JFETs in parallel as in the Accuphase. But not really needed for the JFE2140.
In any case, I am surprised that the LED biasing in the circuit you quote even works -- it uses two LEDs and a 100K resistor in series, which gives what, a couple hundred microamps of current? Most LEDs, except high luminosity ones, would have really low Vf at this point. At least in my simulation with regular green LEDs I am struggling to find a proper balance between the forward voltage and the current flowing through the LED string.
Speaking about simulation: I wonder if it's possible to get a meaningful estimation of input capacitance in LTSpice. For example, by looking at reactance by taking the imaginary part of V(in)/I(Vsource)
Last edited:
Hi Nick,
Here's a cascode that I've been using in an MM phono stage:
My understanding was that a cascode on top of a pair of jfets ... eliminates the Miller capacitance that the jfet gain stage would've had, if there was no cascode and the Drains were merely connected to the DC rail by R1. (So the cascode stops the jfet gain stage from cap-loading the MM cart.)
At least, this was my conclusion after some LTspice simulation. Am I mistaken?
This reminds me of a discussion about clippers for electric guitar. It turned out that some manufacturer-supplied LED SPICE model had an insanely high emission coefficient, while measurements I did on a few LEDs I had lying around just resulted in emission coefficients between 1 and 2, just like ordinary silicon diodes.
See https://www.diyaudio.com/community/...nt-bulb-style-led-strings.407877/post-7570403 and the post after that (not the guitar thread, but the same measurements, measured emission coefficients 1.6 and 1.9). See also https://www.diyaudio.com/community/...id-state-e-guitar-preamps.391513/post-7225586 for the insane SPICE model emission coefficient (9.626).
Last edited:
Very interesting, thank you Marcel. I will have to experiment with the LEDs I have lying around.