Hello everyone. I am electronics engineering student of Kaunas University of Technology.
I present one of my first audio related designs and hope to receive at least some tips what could be done better. The design is my course work of Analog Devices course.
Ir is OPAMP based RIAA corrector with balanced IN, selectable 40/50/60dB gain and +/-0,09dB error from ideal RIAA curve (no IEC). Resistor values from E96, cap values from E12.
I don't have any idea how will it work, but I hope to put it together during upcoming holidays.
I used DIN5 connectors both for balanced in and unbalanced out.
Here are the schematics and PCB layers.
I give it to your trial. Thank You for your opinions and ideas.
I present one of my first audio related designs and hope to receive at least some tips what could be done better. The design is my course work of Analog Devices course.
Ir is OPAMP based RIAA corrector with balanced IN, selectable 40/50/60dB gain and +/-0,09dB error from ideal RIAA curve (no IEC). Resistor values from E96, cap values from E12.
I don't have any idea how will it work, but I hope to put it together during upcoming holidays.
I used DIN5 connectors both for balanced in and unbalanced out.
Here are the schematics and PCB layers.
I give it to your trial. Thank You for your opinions and ideas.
You should post a circuit description or theory of operation. To me, there seems like some extra stuff in there...but it may be valuable extra stuff, so I'd benefit from hearing your reasoning behind your design choices.
One little thing...I would think that you would choose R6=R19=R25=R35, but the first two and last two have slightly different values...is that a typo, or is there a design intent lurking there?
One little thing...I would think that you would choose R6=R19=R25=R35, but the first two and last two have slightly different values...is that a typo, or is there a design intent lurking there?
I thought that - at first. But no, its correct, otherwise the gain looking into the -ve input would be (69.1*70.1) and that into the +ve just 70.1. The opamp feeding the -ve input needs to compensate for the slight gain difference between the two inputs. You meant R36 rather than R16 didn't you?
Last edited:
Here is a corrected schematic.
One again thanks for pointing my mistypes out. The nominal Rin for my Korvet gzm-128 cartridge is 47kOhm. I could have put a pair of resistors and made an exact 47k, but I believe this is not necessary.
OK, the schematic. I will explain using designators of upper circuit.
I have put a DIN5 input connector with non inverted input in the regular place 3 and 5 and inverted input in place of "output" pins 1 and 4.
The C4, C8 caps are for opamp bias current blocking from cartridge. The 100uF is hopefully enough to not introduce any distortion.
Then goes the two-opamp differential amplifier. The gain of the stage is set to about 70 times (about 37dB). High gain with 1% resistors gives high CMRR. The RIAA stage gives another 23dB resulting in 60dB total 1kHz gain.
Because I want to have a selectable gain I have put a resistive divider between stages. This gives 40/50/60dB selectable gain.
The RIAA corrector is nothing really special. It's transfer characteristic can be seen in attached file. w3,4,5 are RIAA defined corner frequencies, w2 is due to C1 cap and w1,6 exist due to feedback type used. w1,2 does not affect the 20..20'000Hz band notably, because with 100uF cap it is far away from this band, but w6 zero does so it is compensated out by additional pole of RC circuit with C7 and the equivalent resistance of R7,R9,R10,R13,R14,R15 which is about 536R in all three jumper positions (the output resistance of the divider in series with R10 or R14). And finally a Zobel network on the output node to load the cable. I could not find the output resistance of NE5532 so I have put a Zobel seen in some other designs. The output connector is also a standard wired DIN5.
One again thanks for pointing my mistypes out. The nominal Rin for my Korvet gzm-128 cartridge is 47kOhm. I could have put a pair of resistors and made an exact 47k, but I believe this is not necessary.
OK, the schematic. I will explain using designators of upper circuit.
I have put a DIN5 input connector with non inverted input in the regular place 3 and 5 and inverted input in place of "output" pins 1 and 4.
The C4, C8 caps are for opamp bias current blocking from cartridge. The 100uF is hopefully enough to not introduce any distortion.
Then goes the two-opamp differential amplifier. The gain of the stage is set to about 70 times (about 37dB). High gain with 1% resistors gives high CMRR. The RIAA stage gives another 23dB resulting in 60dB total 1kHz gain.
Because I want to have a selectable gain I have put a resistive divider between stages. This gives 40/50/60dB selectable gain.
The RIAA corrector is nothing really special. It's transfer characteristic can be seen in attached file. w3,4,5 are RIAA defined corner frequencies, w2 is due to C1 cap and w1,6 exist due to feedback type used. w1,2 does not affect the 20..20'000Hz band notably, because with 100uF cap it is far away from this band, but w6 zero does so it is compensated out by additional pole of RC circuit with C7 and the equivalent resistance of R7,R9,R10,R13,R14,R15 which is about 536R in all three jumper positions (the output resistance of the divider in series with R10 or R14). And finally a Zobel network on the output node to load the cable. I could not find the output resistance of NE5532 so I have put a Zobel seen in some other designs. The output connector is also a standard wired DIN5.
Attachments
I have forgotten another last minute modification:
the final values are in the attached schematic.
I have increased the R11 and R29 to 1k to lower the loading of U1B/U2B and the resulting THD figures.
the final values are in the attached schematic.
I have increased the R11 and R29 to 1k to lower the loading of U1B/U2B and the resulting THD figures.
Hi,
if I were to design a Phono stage I´d ask myself if a differential input is needed. MC-Pickups usually come as true symmetrical devices and here a differential-in OP-amp could make sense. It does not so for MM-Pickups of whom most are desymmetrized in that one of the pins is connected to GND. Here a ´asymmetrical´ classical OP-Amp stage is the right choice.
Keep in mind also that the noise figures of such Instrumentation amps are 3dB higher because of two noisy inputs (+3dB, because of uncorrelated adding of the noise). So for MC-pickups where noise is a serious issue You need OP-Amps (or INAs) with extremely low specified voltage noise and low 1/f frequency. The NE5532 would be a bit too noisy for my taste.
A very low noise would be the AD797. A true differential-in amp could be made up from three OP-amps of which two work as linear gain stages and the third works as differential to single ended converter. You can find this structure integrated in one IC like the INA103 and INA163 of BurrBrown.
The big advantage of this 3-OP-amp topology is that You can adjust gain over a wide range by changing just one resistor and still achieve high CMRR (which is what we aim for primarily in differential amplification.)
A two-OP-amp topology doesn´t offer this flexibility, but it can be tweaked with a variabel resistance Rvar connected between the two inverting OP-amp inputs of U1A and U1B. The gain-formula then changes fom Vo=(V2-V1)x(1+R4/R11) to Vo=(V2-V1)x(1+R4/R11+2xR4/Rvar). Still though the two input signals are not treated equal. The delay introduced by U1B leads to a decreasing CMRR with increasing frequencies.
The electrolytics C4/C8 may be omitted with if the OP-Amps feature low input-bias-currents and low input-bias-offset-currents. Connect C6 directly to R6/R19. MC-pickups have a much lower source impedance than MMs. This requires considerably lower input resistors R6/R19 in the range of 100 to 1.000Ohms. You might add switchable resistors in parallel to C6.
Because of the high gain values of this stage You have to expect considerable values of output offset voltage. Either a active DC-servo or a DC-blocking cap should prevent DC from entering the following stages.
The intention of the resistor network/P1/C7 is not really clear to me. Is it intended as variable attenuator with lowpass character to set the 40/50/60dB of gain? If You want to change gain, change it in the first stage as suggested above with a resistor between the inverting inputs of the two OP-amps. You can´t afford to waste gain in an attenuator, simply because of noise issues. Instead, just connect a resistor from U3As input to GND as a a bias-path for the OP-input. The shematics around U3A look ok so far, but the gain tends against 1 with increasing frequencies and not against 0 as it should be. Yo now have the choice of different solutions. Either change U3 to an inverting gain stage, or You should add a RC-lowpass filter either in front of U3A or behind U3A. The filter put at the output of U3A has the disadvantage that it is load dependant while put between U1As output and U3As input impedances are fixed and predictable. If the crossover frequency is then designed for 75µs (2120Hz) You can omit with one cap in U3As feedback network (which then only functions as 50/500Hz equalizer). This makes calculations easier, is more precise and allows for a correct RIAA-equalization towards the upper frequency range.
To the original RIAA equalization (1928) was later (1963) a highpass cutoff of 20Hz added. C1 could and maybe should be designed to achieve the 20Hz crossover frequency.
You may have a read in the thread "Schematic for Pro-ject phono box" from post #62 on (schematics in #73) till 78.
jauu
Calvin
ps. the postion of R12/20 were correct in the first schematic and are correct in the third.
if I were to design a Phono stage I´d ask myself if a differential input is needed. MC-Pickups usually come as true symmetrical devices and here a differential-in OP-amp could make sense. It does not so for MM-Pickups of whom most are desymmetrized in that one of the pins is connected to GND. Here a ´asymmetrical´ classical OP-Amp stage is the right choice.
Keep in mind also that the noise figures of such Instrumentation amps are 3dB higher because of two noisy inputs (+3dB, because of uncorrelated adding of the noise). So for MC-pickups where noise is a serious issue You need OP-Amps (or INAs) with extremely low specified voltage noise and low 1/f frequency. The NE5532 would be a bit too noisy for my taste.
A very low noise would be the AD797. A true differential-in amp could be made up from three OP-amps of which two work as linear gain stages and the third works as differential to single ended converter. You can find this structure integrated in one IC like the INA103 and INA163 of BurrBrown.
The big advantage of this 3-OP-amp topology is that You can adjust gain over a wide range by changing just one resistor and still achieve high CMRR (which is what we aim for primarily in differential amplification.)
A two-OP-amp topology doesn´t offer this flexibility, but it can be tweaked with a variabel resistance Rvar connected between the two inverting OP-amp inputs of U1A and U1B. The gain-formula then changes fom Vo=(V2-V1)x(1+R4/R11) to Vo=(V2-V1)x(1+R4/R11+2xR4/Rvar). Still though the two input signals are not treated equal. The delay introduced by U1B leads to a decreasing CMRR with increasing frequencies.
The electrolytics C4/C8 may be omitted with if the OP-Amps feature low input-bias-currents and low input-bias-offset-currents. Connect C6 directly to R6/R19. MC-pickups have a much lower source impedance than MMs. This requires considerably lower input resistors R6/R19 in the range of 100 to 1.000Ohms. You might add switchable resistors in parallel to C6.
Because of the high gain values of this stage You have to expect considerable values of output offset voltage. Either a active DC-servo or a DC-blocking cap should prevent DC from entering the following stages.
The intention of the resistor network/P1/C7 is not really clear to me. Is it intended as variable attenuator with lowpass character to set the 40/50/60dB of gain? If You want to change gain, change it in the first stage as suggested above with a resistor between the inverting inputs of the two OP-amps. You can´t afford to waste gain in an attenuator, simply because of noise issues. Instead, just connect a resistor from U3As input to GND as a a bias-path for the OP-input. The shematics around U3A look ok so far, but the gain tends against 1 with increasing frequencies and not against 0 as it should be. Yo now have the choice of different solutions. Either change U3 to an inverting gain stage, or You should add a RC-lowpass filter either in front of U3A or behind U3A. The filter put at the output of U3A has the disadvantage that it is load dependant while put between U1As output and U3As input impedances are fixed and predictable. If the crossover frequency is then designed for 75µs (2120Hz) You can omit with one cap in U3As feedback network (which then only functions as 50/500Hz equalizer). This makes calculations easier, is more precise and allows for a correct RIAA-equalization towards the upper frequency range.
To the original RIAA equalization (1928) was later (1963) a highpass cutoff of 20Hz added. C1 could and maybe should be designed to achieve the 20Hz crossover frequency.
You may have a read in the thread "Schematic for Pro-ject phono box" from post #62 on (schematics in #73) till 78.
jauu
Calvin
ps. the postion of R12/20 were correct in the first schematic and are correct in the third.
Last edited:
Hi,
thought about the input differential amp again.
If we name the output of U1B U1b and the input voltage V1 than and the same for U1A (U1b and V2) then following equations apply.
U1b=(1+R12/R20)xV1.
By superposition OP-amp U1A gives:
U1a=-(R4/R11)xU1b + (1+R4/R11)xV2
This leads to:
U1a=(1+R4/R11)xV2 - (R4R12/R11R20)Vx1
If now R12/R20 = R11/R4 then
U1a=Ax(V2-V1)
with a gain A of:
A=1+R4/R11
So far so good, but this means that there´s nearly no gain for V1 (slightly >1) so U1B works rather like a buffer. Apart from that it jsut adds noise
jauu
Calvin
thought about the input differential amp again.
If we name the output of U1B U1b and the input voltage V1 than and the same for U1A (U1b and V2) then following equations apply.
U1b=(1+R12/R20)xV1.
By superposition OP-amp U1A gives:
U1a=-(R4/R11)xU1b + (1+R4/R11)xV2
This leads to:
U1a=(1+R4/R11)xV2 - (R4R12/R11R20)Vx1
If now R12/R20 = R11/R4 then
U1a=Ax(V2-V1)
with a gain A of:
A=1+R4/R11
So far so good, but this means that there´s nearly no gain for V1 (slightly >1) so U1B works rather like a buffer. Apart from that it jsut adds noise
jauu
Calvin
The DC servo is a good idea. A high order high-pass filter is another one.
The non-inverting stage is 14dB quieter that inverting one. network/p2/V7 is the RC-lowpass behing U3A with gain selection. The idea o changing gain in the first stage with a single resistor is new to me.
Now when you have mentioned separating the 75us time constant from others I do not understand why so many authors do it my way (or I do it their way). Why would anyone bother using this type of feedback when you can put a shelving filter + 1st order low-pass?
I intentionally leave out the IEC amendment.
Hi,
if You feed the MM-Pickup directly into the inverting stage than indeed the inverting stage is a lot noisier if You take a measurement bandwidth of 20kHz. Because of the lowpass-character of the amplitude response the effective Bandwidth is much lower than 20kHz, probabely something around 2kHz. So in praxis the audible differences between a noninverting and the inverting stage are small. In our case the source impedance the inverting stage would see would be small and so the noise contribution would be snmaller than with a MM-Pickup as source.
The inverting stage shows some desirable properties. Its much less affected by overload conditions, its common mode input voltage remains very small and it follows the equalization curve without the need of an additional RC-filter stage.
One reason why it became less popular in modern times is probabely the noise penalty. To get really good nouse figures, many manufacturers offered measurements with shortcircuited inputs (Yamaha for example) instead of reallife source impedances. Under such circumstances the noninverting stages appeared very superior to the inverting stages, though in practise they weren´t.
So a first linear gain stage -maybe with variable gain- with either active or passive DC-blocking and followed by a inverting gain-stage using a parallel RIAA-Feedback structure is a truely fine solution regarding performance, flexibility and low parts count.
jauu
Calvin
if You feed the MM-Pickup directly into the inverting stage than indeed the inverting stage is a lot noisier if You take a measurement bandwidth of 20kHz. Because of the lowpass-character of the amplitude response the effective Bandwidth is much lower than 20kHz, probabely something around 2kHz. So in praxis the audible differences between a noninverting and the inverting stage are small. In our case the source impedance the inverting stage would see would be small and so the noise contribution would be snmaller than with a MM-Pickup as source.
The inverting stage shows some desirable properties. Its much less affected by overload conditions, its common mode input voltage remains very small and it follows the equalization curve without the need of an additional RC-filter stage.
One reason why it became less popular in modern times is probabely the noise penalty. To get really good nouse figures, many manufacturers offered measurements with shortcircuited inputs (Yamaha for example) instead of reallife source impedances. Under such circumstances the noninverting stages appeared very superior to the inverting stages, though in practise they weren´t.
So a first linear gain stage -maybe with variable gain- with either active or passive DC-blocking and followed by a inverting gain-stage using a parallel RIAA-Feedback structure is a truely fine solution regarding performance, flexibility and low parts count.
jauu
Calvin
Thank you for your posts. I shall revise my schematic according to what is said and repost it. Thank You.
I have replaced RIAA stage with inverting one.
I see following To-Do:
insert DC blocking. Active or passive? Won't DC servo induce additional noise compared to simple DC blocking cap? Is it worth it?
I see following To-Do:
insert DC blocking. Active or passive? Won't DC servo induce additional noise compared to simple DC blocking cap? Is it worth it?
- Status
- Not open for further replies.