My building partner itishifi.com is already excited about this version and is working on parts order and exploring PCB layout. We'll use Jan's PS regulators (external, or course) and standard dual op-amp through hole construction. We'll post the board layout when finished if anyone's interested.
All good fortune,
Chris
ps: Pavel may mean response to overload. A single stage RIAA has the best possible overload margin, so brute force may win here.
All good fortune,
Chris
ps: Pavel may mean response to overload. A single stage RIAA has the best possible overload margin, so brute force may win here.
Exactly. Mistracking, erroneous jump from one vinyl track to another due to inability of cartridge to follow track modulation or due to surface damage, may be simulated by high impulse of certain width or measured with a real vinyl record. Then it is about overload recovery and step response of the high pass filter system - return to zero, which may not be as trivial as low pass filter response.
One could argue that a single stage RIAA has the highest possible overload margin across the audio band, so could only be stressed into overload by such terrible disk artifacts as to ruin the experience anyway.
Would a high pass filter following a coupled-to-DC RIAA stage behave differently, assuming all overloading is within the passband? I'm not smart enough to know either way, or even guess.
All good fortune,
Chris
Would a high pass filter following a coupled-to-DC RIAA stage behave differently, assuming all overloading is within the passband? I'm not smart enough to know either way, or even guess.
All good fortune,
Chris
Last edited:
All I can answer to @PMA's question is:
Like @Chris Hornbeck already wrote, a single-stage active RIAA correction amplifier has less chances to be driven into clipping than one with the RIAA correction split over several stages, as the only output stage that can be driven into clipping is the one with full RIAA correction. That's assuming practical input signals will not be steep enough to cause input stage clipping a.k.a. slew-rate limiting.
As long as the amplifier is not driven into clipping, the speed at which it recovers from a disturbance will be determined by the real parts of the closed-loop poles. Neglecting a very low pole that is supposed to be more or less covered by a zero and is therefore supposed to be more or less unobservable (*), the in absolute value smallest real part will be that of the complex pole pair of the third-order Butterworth filter. For a 16 Hz cut-off frequency, they have a real part of -16π rad/s, just like the pole of an 8 Hz first-order filter. I therefore expect the recovery when not clipping to be similar to that of a conventional single-stage active RIAA circuit with first-order roll-off below 8 Hz.
When the amplifier is driven far outside its normal operating region, I expect it to recover faster than a conventional design. Like I wrote in post #1, I came up with this structure trying to find ways to make a single-supply RIAA amplifier settle faster at power-on.
I haven't built it and I spend so much time at work doing simulations that I don't want to also run them at home, but if I understand it correctly, Chris is going to build and try the circuit.
(*): This only applies to the second-order version and the third-order version when it is extended with an output AC coupling network with 122.6 ms time constant.
Like @Chris Hornbeck already wrote, a single-stage active RIAA correction amplifier has less chances to be driven into clipping than one with the RIAA correction split over several stages, as the only output stage that can be driven into clipping is the one with full RIAA correction. That's assuming practical input signals will not be steep enough to cause input stage clipping a.k.a. slew-rate limiting.
As long as the amplifier is not driven into clipping, the speed at which it recovers from a disturbance will be determined by the real parts of the closed-loop poles. Neglecting a very low pole that is supposed to be more or less covered by a zero and is therefore supposed to be more or less unobservable (*), the in absolute value smallest real part will be that of the complex pole pair of the third-order Butterworth filter. For a 16 Hz cut-off frequency, they have a real part of -16π rad/s, just like the pole of an 8 Hz first-order filter. I therefore expect the recovery when not clipping to be similar to that of a conventional single-stage active RIAA circuit with first-order roll-off below 8 Hz.
When the amplifier is driven far outside its normal operating region, I expect it to recover faster than a conventional design. Like I wrote in post #1, I came up with this structure trying to find ways to make a single-supply RIAA amplifier settle faster at power-on.
I haven't built it and I spend so much time at work doing simulations that I don't want to also run them at home, but if I understand it correctly, Chris is going to build and try the circuit.
(*): This only applies to the second-order version and the third-order version when it is extended with an output AC coupling network with 122.6 ms time constant.
Last edited:
I agree if you place the HF pole in the first stage whatever technique is used can give decent overload performance. But, you then usually have a noise penalty arising in the second stage whatever technique - I am not sure about the noise implications of a transconductance first stage though because I have not investigated it.
That said, I see far too many RIAA designs that amplify the signal off the cart by a flat 35-40 dB and then passive EQ followed by a buffer/amp stage. Stereophile is littered with expensive phono amps offering 9-14 dB O/L margins with less than stellar noise performance and these are the designs I mostly comment on.
It's rare to see a genuinely new topology in something as old as phonograph reproduction. Yet here it is. Koolsville.
I appreciate you gentlepersons very much,
Chris
I appreciate you gentlepersons very much,
Chris
Those other passive EQ designs attenuate the signal.
I do transconductance amplifier -> passive EQ (shunt) -> voltage amplifier. At no point does the signal get attenuated. Most of the gain is in the transconductance amplifier.
ETA: Headroom is 23dB, including the input buffer in the integrated amplifier.
Ed
Last edited:
I don't know how you implemented your transconductance amplifier, but assuming it is an ordinary class-A negative-feedback amplifier, the bias currents limit how far you can drive it and the larger you make the bias currents, the more noise of the bias current sources gets added to the signal. In my experience, you can find perfectly acceptable compromise values for phono preamplifiers, but it gets problematic when you want a really large dynamic range.
As an example of what I mean, see the current followers in this post: https://www.diyaudio.com/community/threads/tone-and-loudness-controls.405853/post-7567174
As an example of what I mean, see the current followers in this post: https://www.diyaudio.com/community/threads/tone-and-loudness-controls.405853/post-7567174
Marcel is right, but the output stage current is much higher than the input stage current.
Ed
I couldn't find anything useful about the Paradise Phono preamplifier, but if your transconductance amplifier is conceptually like this, you have to be careful with the noise of the n current sources.
As the noise of the current sources can't flow through the input pins, it has to come out through the outputs.
If you have some clever scheme to ensure it is purely common-mode noise at the outputs while the rest of the circuit only looks at the differential current, or that the noise only comes out of pin ioutn while the rest of the circuit only cares about ioutp, then there is no issue. I had no such scheme in the loudness control circuit I linked to, so the current sources dominated the noise unless I used lots of degeneration in the current sources, requiring a higher supply voltage than intended.
As said, this was problematic for a line level circuit with a high dynamic range target. It may well be a non-issue for a phono preamplifier.
As the noise of the current sources can't flow through the input pins, it has to come out through the outputs.
If you have some clever scheme to ensure it is purely common-mode noise at the outputs while the rest of the circuit only looks at the differential current, or that the noise only comes out of pin ioutn while the rest of the circuit only cares about ioutp, then there is no issue. I had no such scheme in the loudness control circuit I linked to, so the current sources dominated the noise unless I used lots of degeneration in the current sources, requiring a higher supply voltage than intended.
As said, this was problematic for a line level circuit with a high dynamic range target. It may well be a non-issue for a phono preamplifier.
Marcel - I think the biasing noise is hiding behind the cartridge noise.
With the volume set to maximum, I hear:
Ed
With the volume set to maximum, I hear:
- Input shorted: very faint noise if I put my ear against the speaker
- Cartridge connected: faint noise up to about 3 feet from the speaker
- No cartridge (just 47K resistor): a lot of noise
Ed
@EdGr , “cartridge noise” term is misleading. It is not only about thermal noise of (R + jwL)//47k impedance, but much more about additional noise resulting from phono preamp input current noise.
https://www.audiosciencereview.com/...-mm-phono-preamps-properly.60594/post-2224815
https://www.audiosciencereview.com/...-mm-phono-preamps-properly.60594/post-2224815
It is. Below please see Shure M35X "cartridge noise" measured with various op-amps. The cartridge is loaded with 47k//220pF. When the op-amp voltage noise is below 5nV/rt(Hz), it is only and only op-amp input current noise that defines total noise over 20Hz - 20kHz band.
Are we miscommunicating because the terms 'current source noise' and 'equivalent input noise current' are too similar? The current source noise I wrote about in post #75 would contribute to the equivalent input noise voltage, (ideally) not to the equivalent input noise current.
Regarding the 47 kohm, I think it should be accounted for as an equivalent input noise current term of the amplifier. It is physically inside the amplifier and the amplifier designer can decide to use a normal resistor, an electrically cold resistance or a non-standard termination impedance (like @Nick Sukhov would do). Use the Norton equivalent of a noisy resistor and you see that it contributes to the equivalent input noise current.
Regarding the 47 kohm, I think it should be accounted for as an equivalent input noise current term of the amplifier. It is physically inside the amplifier and the amplifier designer can decide to use a normal resistor, an electrically cold resistance or a non-standard termination impedance (like @Nick Sukhov would do). Use the Norton equivalent of a noisy resistor and you see that it contributes to the equivalent input noise current.
Last edited: