I've just finished the build for the first pcb of my modular phono preamp project.
It does RIAA compensation and basic amplification suitable for an MM input.
I intend to design an MC stage pcb, an input connector pcb, an output connector and power pcb (perhaps with battery option), a rumble filter pcb. They will slot together with dupont-style edge connectors, allowing various configurations.
The two channels are independent in this pcb, roughly mirror-image, allowing a star-ground to be used at the power supply end.
SMT is used where possible to keep size down. PPS capacitors seem to be the best compromise for low distortion and SMT availability - here a single value, 10nF, is needed.
The RIAA circuit topology is slightly unusual in that each of the poles and zeroes is an explicit separate RC pair. Care has to be taken to keep the opamps stable into capacitive loads in this topology (the 33pF capacitors and 470 ohm resistors).
The nominal component values give exactly 75µs, 318µs and 3180µs time-constants. Alas I could only source 2% PPS capacitors, but 0.1% thin-film resistors are used throughout.
Where DC-blocking is needed I used back-to-back electrolytics of high value with a centre bias to keep them formed over time. The 22M resistors were a guess for this purpose - the time constant is measured in hours so in theory won't have any impact audibly.
There is no input DC-blocking capacitor, as my calculations show the tiny bias currents of opamps are not going to have any effect on a MM pickup (despite what a lot of people glibly state). 500nA in a 1k5 winding gives an offset of a fraction of a mV, and a few mA-turns of magneto motive force, insignificant compared to signal levels, especially when the inefficiency of the MM cartridge as a motor is considered. I think I'd need to see some hard evidence to the contrary, such as a plot of THD v. offset current.
It does RIAA compensation and basic amplification suitable for an MM input.
I intend to design an MC stage pcb, an input connector pcb, an output connector and power pcb (perhaps with battery option), a rumble filter pcb. They will slot together with dupont-style edge connectors, allowing various configurations.
The two channels are independent in this pcb, roughly mirror-image, allowing a star-ground to be used at the power supply end.
SMT is used where possible to keep size down. PPS capacitors seem to be the best compromise for low distortion and SMT availability - here a single value, 10nF, is needed.
The RIAA circuit topology is slightly unusual in that each of the poles and zeroes is an explicit separate RC pair. Care has to be taken to keep the opamps stable into capacitive loads in this topology (the 33pF capacitors and 470 ohm resistors).
The nominal component values give exactly 75µs, 318µs and 3180µs time-constants. Alas I could only source 2% PPS capacitors, but 0.1% thin-film resistors are used throughout.
Where DC-blocking is needed I used back-to-back electrolytics of high value with a centre bias to keep them formed over time. The 22M resistors were a guess for this purpose - the time constant is measured in hours so in theory won't have any impact audibly.
There is no input DC-blocking capacitor, as my calculations show the tiny bias currents of opamps are not going to have any effect on a MM pickup (despite what a lot of people glibly state). 500nA in a 1k5 winding gives an offset of a fraction of a mV, and a few mA-turns of magneto motive force, insignificant compared to signal levels, especially when the inefficiency of the MM cartridge as a motor is considered. I think I'd need to see some hard evidence to the contrary, such as a plot of THD v. offset current.
ENJOY it now 
You don't need to explain yourself.That is what most people do out of fear of being judged for their different experience.Yet you won't have any problem with a 4.7uf/16v Nichicon Muse bipolar either as they were already measured for less than -120 db THD when a dc bias less than 400mv...
You don't need to explain yourself.That is what most people do out of fear of being judged for their different experience.Yet you won't have any problem with a 4.7uf/16v Nichicon Muse bipolar either as they were already measured for less than -120 db THD when a dc bias less than 400mv...Looks good Mark, have you listened to it yet? measurements? nice that we can make cheap high quality pcbs for our protos. looking forward to your MC implementation.
I remember one of my first phono amps I built, using a NE5533 dual as shown in the Signetics app note.
What do you have for RFI filtering at the input? I see that Self does not use any either in the Elektor 2012 phono design.
One thought you could combine MC/MM on one pcb since your are isolating channel grounding.
I am not sure of the advantages or finding a way to demonstrate that some of your design choices have merit for but hey why not experiment.
Do not think I have ever seen a audio design where they bias an ecap such as R39,41
I remember one of my first phono amps I built, using a NE5533 dual as shown in the Signetics app note.
What do you have for RFI filtering at the input? I see that Self does not use any either in the Elektor 2012 phono design.
One thought you could combine MC/MM on one pcb since your are isolating channel grounding.
I am not sure of the advantages or finding a way to demonstrate that some of your design choices have merit for but hey why not experiment.
Do not think I have ever seen a audio design where they bias an ecap such as R39,41
The whole point of modular is not to combine functions on one pcb!
The unusual filter topology was just conceit really - its not efficient in opamps, but does make the component calculations trivial. It would be more rational in a general purpose equalization unit as each pole and zero can be individually switched/controlled. Of course the idea is limited to real poles and zeroes only.
The unusual filter topology was just conceit really - its not efficient in opamps, but does make the component calculations trivial. It would be more rational in a general purpose equalization unit as each pole and zero can be individually switched/controlled. Of course the idea is limited to real poles and zeroes only.
Testing with two tone signal into inverse-RIAA network (-45dB or so at 1kHz), via modular preamp (+30dB at 1kHz) to home-made FFT engine running from an ADS8885 18 bit SAR ADC.
44kHz and 4.8kHz tones. The 0dB level is about 2.8V rms IIRC
The output level is thus about 25mV per peak, corrsponding to about 2mV at RIAA amp input around 4kHz. The peaks are about 82 dB above the IM3 products at the 3rd harmonic frequencies.
The rise in noise floor at low frequencies is mainly due to the greater attenuation of the inverse RIAA network at low frequencies. Since a flat-top window was used this plot's noise floor is rather higher than a Hann window would give, I forgot to do a Hann plot.
Anyway I'm happy the preamp is performing OK from this - the FFT system is very much a work in progress I should point out.
The signal source is a lowly WM8524G, but I'm working on one based on the PCM1796 with better performance.
44kHz and 4.8kHz tones. The 0dB level is about 2.8V rms IIRC
The output level is thus about 25mV per peak, corrsponding to about 2mV at RIAA amp input around 4kHz. The peaks are about 82 dB above the IM3 products at the 3rd harmonic frequencies.
The rise in noise floor at low frequencies is mainly due to the greater attenuation of the inverse RIAA network at low frequencies. Since a flat-top window was used this plot's noise floor is rather higher than a Hann window would give, I forgot to do a Hann plot.
Anyway I'm happy the preamp is performing OK from this - the FFT system is very much a work in progress I should point out.
The signal source is a lowly WM8524G, but I'm working on one based on the PCM1796 with better performance.
An update on the modular RIAA amp. I added a rumble filter module and input and output modules (the latter with PCB connectors for two 9V batteries) to make a complete working system.
The rumble filter is 10 pole Salen Key Butterworth:
The full suite of boards:
And the flat style SMT pin header connectors:
The rumble filter is 10 pole Salen Key Butterworth:
The full suite of boards:
And the flat style SMT pin header connectors:
I've put together a simulation of the rumble filter to show its response, and this version of the schematic has readable resistor values (!)
[BTW I meant 5 pole above, not 10 pole, ie 30dB per octave]
This is a 3rd-order Salen Key highpass followed by a 2nd-order Salen Key high pass.
Which is a nice 30dB down by 10Hz with a -1dB point of about 23.5Hz and -3dB at 20Hz - possibly slight overkill but only 2 opamps and the caps are all identical (I used PPS 100nF 2% (Panasonic ECHU) for the surface mount board).
[BTW I meant 5 pole above, not 10 pole, ie 30dB per octave]
This is a 3rd-order Salen Key highpass followed by a 2nd-order Salen Key high pass.
Which is a nice 30dB down by 10Hz with a -1dB point of about 23.5Hz and -3dB at 20Hz - possibly slight overkill but only 2 opamps and the caps are all identical (I used PPS 100nF 2% (Panasonic ECHU) for the surface mount board).
I was tinkering with my distortion-measuring Python script and thought to do a THD test on my modular preamp via an inverse RIAA network through a 3rd-gen Scarlett Solo, getting a quite pleasing 0.00068% figure (including rumble filter, so 5 opamps in the signal path)
The software takes in a .wav file and iterates to a best-fit model sinusoid to match the data (in amplitude, frequency and phase), then subtracts it off to get the residual, and measures the correlation of the residual against harmonics of the model to get the red peaks and thus can measure THD and SNR separately.
I pre-processed with a 200Hz high-pass filter to knock off most of the mains hum components (wasn't measuring in a screened box, just on desktop) - otherwise this dominates the SNR (but doesn't affect the THD measurement).
[ The SNR ratio is fairly meaningless as I failed to record the actual level ]
I also measured the current consumption at 36mA per rail, which is as expected with 5 opamp stages per channel. Next step is boxing it up.
The software takes in a .wav file and iterates to a best-fit model sinusoid to match the data (in amplitude, frequency and phase), then subtracts it off to get the residual, and measures the correlation of the residual against harmonics of the model to get the red peaks and thus can measure THD and SNR separately.
I pre-processed with a 200Hz high-pass filter to knock off most of the mains hum components (wasn't measuring in a screened box, just on desktop) - otherwise this dominates the SNR (but doesn't affect the THD measurement).
[ The SNR ratio is fairly meaningless as I failed to record the actual level ]
I also measured the current consumption at 36mA per rail, which is as expected with 5 opamp stages per channel. Next step is boxing it up.
I recently measured the performance of this preamp, and also realized the images have gone away so first I'll repost the images:
RIAA preamp:
And the rumble filter:
The schematic of the RIAA preamp section:
So now the measurements (with RIAA plus rumble filter, power +/-9V), 5mV input yields:
And plotting THD and THD+N against input level:
And frequency response weighted by RIAA inverse curve:
RIAA preamp:
And the rumble filter:
The schematic of the RIAA preamp section:
So now the measurements (with RIAA plus rumble filter, power +/-9V), 5mV input yields:
And plotting THD and THD+N against input level:
And frequency response weighted by RIAA inverse curve: