Hi,
I'm looking at various solid state phono preamop designs out there, trying to learn from them and coming up with my own design at some point. (I know, there are a lot of working designs out there but for this one I want to have something that I understand from the ground up)
I'm still in the very early stages thinking about the general features and structure of the amp before going into designing the details.
One thing I know I want to build is the ability for the amp to be adapted to a wide range of MM and MC cartridges because I like to experiment with those a lot (I don't like step-up transformers).
So the overall gain needs to be adaptable and I have seen a number of approaches how that is implemented:
- Simple have one gain stage that can provide 40-65 dB by variying the amount of NFB on the OpAmp
- Use two stages (one before and one after RIAA) and make the gain in the second stage variable so it can provide additional boost in case of the MC cardridge
- Have a Pre-Pre-Amp stage that gets inserted before the main MM/RIAA stage to provide around 20-25 dB of Gain
I have a bit of a hard time figuring out the advantages and disadvantages of the various topology choices, so maybe some here would be able to share some wisdom on the matter.
What I like about the idea of the Pre-Pre-Amp topology is that that stage can be much more customized toward the needs of the lower impedance MC system, beyond simply changing the terminating resistor to load the cartridge apropriatly... may discrete transistors would make more sense in that stage than an op-amp
However, there are probably many more things to consider.
Cheers,
Lars
I'm looking at various solid state phono preamop designs out there, trying to learn from them and coming up with my own design at some point. (I know, there are a lot of working designs out there but for this one I want to have something that I understand from the ground up)
I'm still in the very early stages thinking about the general features and structure of the amp before going into designing the details.
One thing I know I want to build is the ability for the amp to be adapted to a wide range of MM and MC cartridges because I like to experiment with those a lot (I don't like step-up transformers).
So the overall gain needs to be adaptable and I have seen a number of approaches how that is implemented:
- Simple have one gain stage that can provide 40-65 dB by variying the amount of NFB on the OpAmp
- Use two stages (one before and one after RIAA) and make the gain in the second stage variable so it can provide additional boost in case of the MC cardridge
- Have a Pre-Pre-Amp stage that gets inserted before the main MM/RIAA stage to provide around 20-25 dB of Gain
I have a bit of a hard time figuring out the advantages and disadvantages of the various topology choices, so maybe some here would be able to share some wisdom on the matter.
What I like about the idea of the Pre-Pre-Amp topology is that that stage can be much more customized toward the needs of the lower impedance MC system, beyond simply changing the terminating resistor to load the cartridge apropriatly... may discrete transistors would make more sense in that stage than an op-amp
However, there are probably many more things to consider.
Cheers,
Lars
An advantage of the pre-preamp approach is that you can noise optimize the pre-preamp for MC cartridge impedances and the rest for MM cartridge impedances. For MM cartridges, having a low equivalent input noise current is far more important than for MC cartridges and for MC cartridges, having a low equivalent input noise voltage is far more important than for MM cartridges.
That said, if you don't care about the noise in between records, record surface noise will dominate unless you completely mess up the noise optimization. The only design I know of that is so poor that its noise is of the same order as record surface noise is the moving-magnet version of the Elektor Supra 2.0 with four LT1028's per input.
By the way, threads like this are usually placed in the Analogue source subforum, rather than Analog line level.
That said, if you don't care about the noise in between records, record surface noise will dominate unless you completely mess up the noise optimization. The only design I know of that is so poor that its noise is of the same order as record surface noise is the moving-magnet version of the Elektor Supra 2.0 with four LT1028's per input.
By the way, threads like this are usually placed in the Analogue source subforum, rather than Analog line level.
Last edited:
That's what I was leaning to and I think I like that idea very much. I also think it can be implemented well in terms of the routing/switching that is required to select between MM/MC. No swicthing required before the signal from the MC cartridge and the initial stage. Only a selector before the MM/RIAA stage to feed it either from the MC section or from the MM inputs.
I think I have read that it was the other way round and equivivalent current noise was more important for MC due to its low impedance nature. Could you give a few more details? I'd like to really understand that part.
That is super interesting. The supra 2.0 was indeed one of the designs I have been studying in the past weeks in order to learn what's out there. The idea I found the indea of parallel gain devices to limit noise very nice. It's similar to what is done in some image captuturing techniques (e.g. deep-sky astrophotography) where you take multiple images of the same subject and compute the average of those exposures to reduce the noise.
What went wrong in case of the Supra 2.0?
Apologies. If a moderator cares to move this over there I'd be very greateful.
Thanks!
Lars
It's quite simple as long as you only look at one frequency. The noise of a noisy amplifier can be modelled with two equivalent input noise sources, a noise voltage source in series with the input and a noise current source shunted across the input. When you multiply the equivalent input noise current with the source impedance, you know how much equivalent input noise voltage would have the same impact (just Ohm's law for impedances, voltage is current times impedance). The higher the impedance, the more sensitive you are for equivalent input noise current.
What complicates noise optimization of phono amplifiers, particularly moving magnet, is the frequency-dependence of the source impedance. A typical 1 kohm, 500 mH moving-magnet cartridge will be 1 kohm at low frequencies and almost 63 kohm at 20 kHz. What impedance do you have to take into account?
A long time ago I've shown that when the equivalent input noise current and voltage are white and uncorrelated, optimizing the noise at 3852 Hz is the same as optimizing the RIAA- and A-weighted noise over the whole band. Using RIAA and ITU-R 468 weighting, that frequency becomes 5179 Hz. A 1 kohm, 500 mH cartridge has an impedance of about 12 kohm at 3852 Hz, 16 kohm at 5179 Hz.
When you connect n > 1 similar amplifiers in parallel and average their output signals, you indeed reduce the effect of their equivalent input noise voltages. However, the equivalent input noise current sources end up in parallel, so their noise currents are summed and affect all n amplifiers equally. As the equivalent input noise current sources are independent, the RMS value of their sum increases with the square root of n. So all in all, the effect of the equivalent input noise voltage of the amplifiers is reduced by a factor of sqrt(n) and the effect of the equivalent input noise current of the amplifiers is increased by a factor of sqrt(n).
The LT1028 has an equivalent input noise voltage of about 0.85 nV/sqrt(Hz) and an equivalent input noise current of about 3.4 pA/sqrt(Hz). The datasheet says 1 pA/sqrt(Hz), but when you carefully read the small print, you will see that that only applies when the positive and negative inputs of the op-amp are driven from exactly the same impedance, which almost never happens in a real-life application (you are not likely to put a dummy cartridge in the feedback network).
3.4 pA/sqrt(Hz) times 12 kohm is 40.8 nV/sqrt(Hz), so when you use a single LT1028, the noise current already dominates by far over the 0.85 nV/sqrt(Hz). Use four of them, and it becomes 6.8 pA/sqrt(Hz) and 0.425 nV/sqrt(Hz), and 6.8 pA/sqrt(Hz) times 12 kohm is even 81.6 nV/sqrt(Hz).
A similar approach does make sense for moving coil, because of its much lower impedance, or with FET op-amps, as those hardly have any equivalent input noise current.
What complicates noise optimization of phono amplifiers, particularly moving magnet, is the frequency-dependence of the source impedance. A typical 1 kohm, 500 mH moving-magnet cartridge will be 1 kohm at low frequencies and almost 63 kohm at 20 kHz. What impedance do you have to take into account?
A long time ago I've shown that when the equivalent input noise current and voltage are white and uncorrelated, optimizing the noise at 3852 Hz is the same as optimizing the RIAA- and A-weighted noise over the whole band. Using RIAA and ITU-R 468 weighting, that frequency becomes 5179 Hz. A 1 kohm, 500 mH cartridge has an impedance of about 12 kohm at 3852 Hz, 16 kohm at 5179 Hz.
When you connect n > 1 similar amplifiers in parallel and average their output signals, you indeed reduce the effect of their equivalent input noise voltages. However, the equivalent input noise current sources end up in parallel, so their noise currents are summed and affect all n amplifiers equally. As the equivalent input noise current sources are independent, the RMS value of their sum increases with the square root of n. So all in all, the effect of the equivalent input noise voltage of the amplifiers is reduced by a factor of sqrt(n) and the effect of the equivalent input noise current of the amplifiers is increased by a factor of sqrt(n).
The LT1028 has an equivalent input noise voltage of about 0.85 nV/sqrt(Hz) and an equivalent input noise current of about 3.4 pA/sqrt(Hz). The datasheet says 1 pA/sqrt(Hz), but when you carefully read the small print, you will see that that only applies when the positive and negative inputs of the op-amp are driven from exactly the same impedance, which almost never happens in a real-life application (you are not likely to put a dummy cartridge in the feedback network).
3.4 pA/sqrt(Hz) times 12 kohm is 40.8 nV/sqrt(Hz), so when you use a single LT1028, the noise current already dominates by far over the 0.85 nV/sqrt(Hz). Use four of them, and it becomes 6.8 pA/sqrt(Hz) and 0.425 nV/sqrt(Hz), and 6.8 pA/sqrt(Hz) times 12 kohm is even 81.6 nV/sqrt(Hz).
A similar approach does make sense for moving coil, because of its much lower impedance, or with FET op-amps, as those hardly have any equivalent input noise current.
You are awesome. Thanks for breaking it down. This is detailed enough for me to work it out for a few examples on my own and that should hopefully make it click. I don't yet have the intuitive grip on all those concepts but once I have broken it down to a level where I can run the math I can get there after a few iterations 🙂
A pragmatic approach is to use two different amplifiers (with different input device types), one for MC and the other for MM.
For MM you care about equivalent input current noise and you don't care about equivalent input voltage noise as much -- so use an amplifier with FETs as the input devices. Such as the ADA4625
For MC you care about equivalent input voltage noise and you don't care about equivalent input current noise as much -- so use an amplifier with BJTs as the input devices. Such as the ADA4898. Or an all discrete MC "flat amp" using the legendary Zetex ZTX951 PNPs as input devices. Possibly several of them in parallel for even lower noise.
Now arrange the MM amplifier to include RIAA de-emphasis, and arrange the MC amplifier to be flat frequency response with "flat gain" of 20-24dB (10x to 15x). Use high quality relays with gold plated contacts to let the MM amplifier's input connect to either (A) the rear panel RCA input jack for MM cartridte; or else (B) the output of the MC amplifier. Done.
For MM you care about equivalent input current noise and you don't care about equivalent input voltage noise as much -- so use an amplifier with FETs as the input devices. Such as the ADA4625
For MC you care about equivalent input voltage noise and you don't care about equivalent input current noise as much -- so use an amplifier with BJTs as the input devices. Such as the ADA4898. Or an all discrete MC "flat amp" using the legendary Zetex ZTX951 PNPs as input devices. Possibly several of them in parallel for even lower noise.
Now arrange the MM amplifier to include RIAA de-emphasis, and arrange the MC amplifier to be flat frequency response with "flat gain" of 20-24dB (10x to 15x). Use high quality relays with gold plated contacts to let the MM amplifier's input connect to either (A) the rear panel RCA input jack for MM cartridte; or else (B) the output of the MC amplifier. Done.
Great, thanks for those pointers, Mark.
While following up on those, I came across by Andrew Russell's X-Altra -> You Can DIY! X-Altra MC-MM RIAA EQ Preamp - Part 1 | audioXpress that's a wonderful project I think...
While following up on those, I came across by Andrew Russell's X-Altra -> You Can DIY! X-Altra MC-MM RIAA EQ Preamp - Part 1 | audioXpress that's a wonderful project I think...