I've been taking a look at an alternative approach to RIAA equalisation that has been patented in the last few years (https://patents.google.com/patent/US20170126186A1/en) by Graham Slee. This basically equalises the pickup's constant velocity frequency response by putting an integrator as the first stage providing the 3180us pole (50Hz). This effectively provides a constant loop gain for that stage. The remaining 318us (500Hz) and 75us (2122Hz) time constants can easily be dealt with by a simple passive filter. My take of an MM phono using this approach is attached. I've include a secondary LP filter to address the non-inverting amp's roll-off at unity gain, a 20 Hz rumble filter and a simple flat gain stage with potential gain adjustments to allow for differing cartridge output levels. Overload margin is in excess of 30dB for a 5mV MM cartridge.
I've modelled the input stage with an OPA1656, a very low current noise/low distortion FET op-amp. Other suitable choices would be OPA1641 / OPA1642 or similar. As an aside, the use of a FET device as the input stage will also provide good RFI proofing.
This has been built and is my current phono amp in use on a regular basis. On a subjective basis, its clean, precise and provides an open sound stage. As far as measurements go, I've run a quick and dirty frequency response check using an HP3312A function generator, inverse-RIAA network and an analogue 'scope. This confirms the flat response across the 20Hz to 20KHz range. I also confirmed the overload margin.
I've modelled the input stage with an OPA1656, a very low current noise/low distortion FET op-amp. Other suitable choices would be OPA1641 / OPA1642 or similar. As an aside, the use of a FET device as the input stage will also provide good RFI proofing.
This has been built and is my current phono amp in use on a regular basis. On a subjective basis, its clean, precise and provides an open sound stage. As far as measurements go, I've run a quick and dirty frequency response check using an HP3312A function generator, inverse-RIAA network and an analogue 'scope. This confirms the flat response across the 20Hz to 20KHz range. I also confirmed the overload margin.
Hi!
Nice design, it reminds me John Broskie's ideas showed at https://tubecad.com/2022/03/21/60dB MC Hybrid Active-Eq Phono Preamp Large.png
https://tubecad.com/2022/03/blog0555.htm
cheers
P
Nice design, it reminds me John Broskie's ideas showed at https://tubecad.com/2022/03/21/60dB MC Hybrid Active-Eq Phono Preamp Large.png
https://tubecad.com/2022/03/blog0555.htm
cheers
P
My take on breaking down the RIAA filtering into stages endevoured to avoid any one stage having a high gain (in order to get the best from each opamp) - each pole and zero is done actively. https://www.diyaudio.com/community/threads/can-an-opamp-mm-phono-preamp-be-blamefree.401679/ - add an ultrasonic roofing filter, rumber filter and electronically cooled 47k load resistor for good measure!
Your high gain in the first stage is forcing the need for C11 (avoid large DC offset) I think - having less gain in the first stage has several advantages - more headroom at some frequencies, no need for C11, more linear (or at least less demanding on the first amp).
I disproprotionally like separating out each pole and zero to avoid the complicated algebra for determining RC values!
You only have 2 poles in the rumble filter, yet its a demanding task, as you want to get good attenuation from 15 Hz down without trespassing on the bottom octave - I chose 5 pole Butterworth which is perhaps overkill, but I suggest more than 2 poles is worth it. Not sure you need output capacitor after a high-pass filter, the DC-offset will be tiny.
Given the phasing out of the NE5534A the OPA1656 seems a good choice, especially as its a dual package.
Your high gain in the first stage is forcing the need for C11 (avoid large DC offset) I think - having less gain in the first stage has several advantages - more headroom at some frequencies, no need for C11, more linear (or at least less demanding on the first amp).
I disproprotionally like separating out each pole and zero to avoid the complicated algebra for determining RC values!
You only have 2 poles in the rumble filter, yet its a demanding task, as you want to get good attenuation from 15 Hz down without trespassing on the bottom octave - I chose 5 pole Butterworth which is perhaps overkill, but I suggest more than 2 poles is worth it. Not sure you need output capacitor after a high-pass filter, the DC-offset will be tiny.
Given the phasing out of the NE5534A the OPA1656 seems a good choice, especially as its a dual package.
Last edited:
And excellent designs they are too, each with its own twist to a common theme - achieving RIAA eq with as good a result as can be attained with current tech.
Hi Mark,
Thanks for your constructive feedback. I'll try to address elements of this in stages.
The NE5534 is still an excellent op-amp (despite many decrier's) and by design was optimised for use in MM phono stages. However, it doesn't have sufficient open loop gain for the approach taken in my design. My choice of an OPA1656 for stage one was based on a number of things:
I've selected resistors / capacitances for the three RIAA time constants which are readily available as low tolerance COG ceramic / film capacitors and low tolerance thin film chip resistors.
The 3180us time constant is covered in a single stage. This stage has a 6dB/octave roll-off from 50Hz. At 20KHz, loop gain is still in excess of 40dB
C2, C3, R4, R5 and R6 form a pole/zero filter effecting the 318us and 75us time constants. Again, these use easily obtainable chip resistors and capacitors. The formula for calculating the values to implement each pole and necessary levels of attenuation before and after each of the pole / zero is a simple one. Also, the resultant values aren't adversely affected by loading from stages before and after them.
R5 /C3 are the secondary filter to compensate for the high frequency unity gain consequence of a non-inverting gain stage (stage one).
Following this secondary filter is a three pole 20Hz rumble filter then a flat gain stage which is simply there to bring the final output up to a nominal line level.
While I'm showing LME49860 (one half of an LM4562) devices in the LTSpice diagram, there's no reason not to use OPA1656 or OPA1641/2 in their place.
Thanks for your constructive feedback. I'll try to address elements of this in stages.
The NE5534 is still an excellent op-amp (despite many decrier's) and by design was optimised for use in MM phono stages. However, it doesn't have sufficient open loop gain for the approach taken in my design. My choice of an OPA1656 for stage one was based on a number of things:
- Low input noise voltage / extremely low current noise
- Rail to rail output swing
- 150dB open loop gain
- 50MHz gain bandwidth
- 20V/us slew rate
- FET 's inherent immunity to RFI
I've selected resistors / capacitances for the three RIAA time constants which are readily available as low tolerance COG ceramic / film capacitors and low tolerance thin film chip resistors.
The 3180us time constant is covered in a single stage. This stage has a 6dB/octave roll-off from 50Hz. At 20KHz, loop gain is still in excess of 40dB
C2, C3, R4, R5 and R6 form a pole/zero filter effecting the 318us and 75us time constants. Again, these use easily obtainable chip resistors and capacitors. The formula for calculating the values to implement each pole and necessary levels of attenuation before and after each of the pole / zero is a simple one. Also, the resultant values aren't adversely affected by loading from stages before and after them.
R5 /C3 are the secondary filter to compensate for the high frequency unity gain consequence of a non-inverting gain stage (stage one).
Following this secondary filter is a three pole 20Hz rumble filter then a flat gain stage which is simply there to bring the final output up to a nominal line level.
While I'm showing LME49860 (one half of an LM4562) devices in the LTSpice diagram, there's no reason not to use OPA1656 or OPA1641/2 in their place.
Last edited:
Best S/N and best dynamic range performance will always come from the classic single op-amp model with long loop feedback (as long as everything is linear - and if it's not, you can't fix it that way). I really don't understand these elaborations - do they even really eliminate interactions of the RC poles/zeros? Until I've heard MarcelvdG's ruling, I'll have to hold up agreement with the whole premise. And even then, why?
All good fortune,
Chris
All good fortune,
Chris
Having rerun my simulation, I see that R18 (only there to provide discharge path for C9) and C9 can also be eliminated with no detrimental effects.
Chris,
I'm all for the KISS approach to things.
The classic single op-amp model uses the Lipshitz eq network to achieve the resultant RIAA equalisation. The calculations are anything but straightforward due to interactions between the various elements. Using Bonsai's RIAA calculator (others are around on the internet) makes life easier for this but we still have resistor and capacitor values which can only be met by series / parallel combinations of standard values.
Just FYI, a passive EQ network driven by a transconductance amplifier can match that and not suffer from the +1 error.
Ed
Very interesting circuit.
But I don't really understand, what is very special from the approach described under
https://patents.google.com/patent/US20170126186A1/en
P.S.: As I know, Graham Slee wasn't a good friend of the NE5534 due a not ear-friendly distortion signature.
He prefer OP-Amps with only one voltage gain stage (folded cascode) like AD817 and AD825/AD826 - go to
https://www.hifisystemcomponents.com/forum/era-gold-mk-v-earlier-version_topic4768.html
(Posted: 13 Jul 2019 at 9:52pm)
and
https://www.tnt-audio.com/ampli/elevator_e.html
https://www.audioasylum.com/cgi/vt.mpl?f=vinyl&m=863876
Once you connect a real world MM cart to a phono amplifier, the phono amp noise for the most part is dominated by the cart + load resistor noise. On MM inputs, my X-Altra MC/MM pre has a 1.6 nV/rt Hz spot noise spec vs 0.9 nV/rt Hz on Denis Colin’s LP797 preamp. But, connect a cart and thd difference in measured noise is not nearly 6 dB, but <0.2 dB and the all active X-Altra has >30 dB overload margin vs IIRC leads than half that on the active/passive LP797.
Nice design btw Geoff 🙂👍
Nice design btw Geoff 🙂👍
Hi,
I remember an older paper by JLH in which he discussed different equalizing strategies.
He came to the conclusion, that a linear input stage followed by a passive 75us network, followed by a actively eq'd 3180/318us gain stage fulfilled the requirements almost perfect.
I'm using a variation of this approach for almost 40years by now with great success.
My variation of this theme utilizes a INA as input stage, actively servoed to implement a first 7950us HP.
This allows for highest flexibility regarding the range of pickups, from LO-MCs to HO-MMs, and their connection, either balanced or un-balanced.
And while it may not be the very very best regarding noise specs, the noise performance is still top notch and way above the requirements of Vinyl playback ... hence only of academical interest.
Splitting the equalization into two or three separates allows to optimize amplitude response and noise performance to highest levels and leaves room to implement variations of the RIAA response like the infamous 'Neutrik'.
Moving along the signal path subsonic filtering stages are added as either passive HPs or a dedicated active stage ahead of a output driver/buffer.
Since the second active stage can add about 20dB of gain @1kHz, there's plenty of overload margin left for the first linear gain stage.
A minor 'advantage' of the wide-input-range stage is that all attached carts are amplified with the exact same sonic fingerprint, making it a favourate for pickup reviews, as no additional stepup devices or different dedicated input stages for MM or MC with differing sonic characters need to be involved.
While it sounds complicated it is still applied KISS principle .... just offering great flexibility to the user.
jauu
Calvin
I remember an older paper by JLH in which he discussed different equalizing strategies.
He came to the conclusion, that a linear input stage followed by a passive 75us network, followed by a actively eq'd 3180/318us gain stage fulfilled the requirements almost perfect.
I'm using a variation of this approach for almost 40years by now with great success.
My variation of this theme utilizes a INA as input stage, actively servoed to implement a first 7950us HP.
This allows for highest flexibility regarding the range of pickups, from LO-MCs to HO-MMs, and their connection, either balanced or un-balanced.
And while it may not be the very very best regarding noise specs, the noise performance is still top notch and way above the requirements of Vinyl playback ... hence only of academical interest.
Splitting the equalization into two or three separates allows to optimize amplitude response and noise performance to highest levels and leaves room to implement variations of the RIAA response like the infamous 'Neutrik'.
Moving along the signal path subsonic filtering stages are added as either passive HPs or a dedicated active stage ahead of a output driver/buffer.
Since the second active stage can add about 20dB of gain @1kHz, there's plenty of overload margin left for the first linear gain stage.
A minor 'advantage' of the wide-input-range stage is that all attached carts are amplified with the exact same sonic fingerprint, making it a favourate for pickup reviews, as no additional stepup devices or different dedicated input stages for MM or MC with differing sonic characters need to be involved.
While it sounds complicated it is still applied KISS principle .... just offering great flexibility to the user.
jauu
Calvin
Also, if the passive EQ network is instead brought back to transconductance amp's inverting input, Miller C compensation can usually be eliminated, however the +1 error creeps back. Either would be difficult to do with conventional monolythic 8-legger opamps, requiring access to the junction of VAS output and follower input, and any remaining internal Miller C would compromise the trans-amp's output impedance. Discrete designs or even vacuum valve designs could benefit.
All good fortune,
Chris
Chris - Yes, my statement was in the context of discrete transistor pre-amplifiers.
Geoff - I think your design is fine. I did not see any problems (that is a complement coming from me). 🙂
Ed
Geoff - I think your design is fine. I did not see any problems (that is a complement coming from me). 🙂
Ed
Great discussion. You all seem to be discussing performance with typical cartridge response. It has been suggested in other threads that what really endears any phono stage to the listener in the long term is how it behaves in recovery after large overloads, say dirt or a scratch on the LP, where the cartridge will output a signal several times in magnitude whatever it says on the datasheet. Would that not be a big weakness of FB RIAA?
T. Holman's work in the late 1970s showed that even signal levels (rarely, but possible) on commercial records could approach +30dB ref the 0VU of 5cm/s each groove wall. This can be accommodated by an ordinary 8-legger with +/- 15VDC supplies if gain is optimally distributed and total gain is limited to a 1KHz output at 0VU of about 300mV, where +30VU is almost 10Vrms. Rocks and gouges can still clip.
If constrained to +/- 15VDC supplies and desiring 300mV output at 0VU, anything other than a single long loop feedback, or a transconductance input stage with RIAA loading, will have less dynamic range capacity, ei will clip sooner. Nobody wants to hear this, because it's not in fashion.
Of course, nobody said we all have to work with +/- 15VDC supplies. Ordinary vacuum valves work with 300VDC easily, and can move the clipping point with big rocks and scratches forward, where it's maybe after a volume control.
All good fortune,
Chris
If constrained to +/- 15VDC supplies and desiring 300mV output at 0VU, anything other than a single long loop feedback, or a transconductance input stage with RIAA loading, will have less dynamic range capacity, ei will clip sooner. Nobody wants to hear this, because it's not in fashion.
Of course, nobody said we all have to work with +/- 15VDC supplies. Ordinary vacuum valves work with 300VDC easily, and can move the clipping point with big rocks and scratches forward, where it's maybe after a volume control.
All good fortune,
Chris
My experience is that big pops come from boulder-sized rocks stuck in the grooves and deep gouges. At that point, I discard the record.
Program material can't be too much above 5cm/s because the record might skip.
Clean records are nearly silent. Cleaning also eliminates the static that attracts dust. The static elimination is permanent... I am not sure why.
Having over 20dB of headroom is great but not something that one wants to use in practice.
Ed
Program material can't be too much above 5cm/s because the record might skip.
Clean records are nearly silent. Cleaning also eliminates the static that attracts dust. The static elimination is permanent... I am not sure why.
Having over 20dB of headroom is great but not something that one wants to use in practice.
Ed
Hot 12" 45 pressing would say otherwise - the groove velocity is upto 70cm/s, so tangential stylus velocities can easily be 30 or 40cm/s 60cm/s is reckoned to be about the max plausible. See here if you're sceptical: https://pspatialaudio.com/max_velo.htm
So 60cm/s is about 22dB more than 5cm/s, so you really need 20+dB headroom above that nominal 5cm/s value.
Not even fresh lacquer cuts are free of obvious surface noise. With a good quiet phono preamp this should be very obvious (try headphones). We automatically filter out surface noise when listening to vinyl by habit, but its definitely there and definitely audible.