I am assuming it is R18, R24 of 1M and C14, C25 of 2.2uF that filter out those frequencies? I was trying to find information on these elements but no luck so far.
The DC servo is a lowpass filter with a cutoff frequency of (in your case) 1/(2*pi*2.2e-6*1e6) = 72 mHz. This is a bit more than two decades below 20 Hz, so figure the filter will attenuate a bit over 40 dB at 20 Hz. With the LM3886 that could be enough that the DC servo doesn't add any measurable contribution within the audio band. Whether it is enough will likely need to be determined by measurement as THD tends to not simulate reliably.
I originally had a 1st order DC servo similar to yours in the Modulus-86 Rev. 1.0. I found the output of the DC servo degraded the THD at low frequency and had to push the servo pole down to 12 mHz to push the THD added by the DC servo below the noise floor. This resulted in a rather long settling time (minutes!) so for Modulus-86 Rev. 2.0 I designed a 3rd order DC servo (-60 dB/dec filter slope). I found multiple feedback filters handy for this. TI has a good application note on the topic (https://www.ti.com/lit/an/sloa049b/sloa049b.pdf).
Granted, with the Modulus-86 I apply error correction to the LM3886 so I get much lower THD. This causes things like the DC servo to pop up as a contributor to THD. It's entirely possible that you don't need to worry about this. I'm just bringing it up as something to look for as it can be a trap if you allow it to be.
Tom
I originally had a 1st order DC servo similar to yours in the Modulus-86 Rev. 1.0. I found the output of the DC servo degraded the THD at low frequency and had to push the servo pole down to 12 mHz to push the THD added by the DC servo below the noise floor. This resulted in a rather long settling time (minutes!) so for Modulus-86 Rev. 2.0 I designed a 3rd order DC servo (-60 dB/dec filter slope). I found multiple feedback filters handy for this. TI has a good application note on the topic (https://www.ti.com/lit/an/sloa049b/sloa049b.pdf).
Granted, with the Modulus-86 I apply error correction to the LM3886 so I get much lower THD. This causes things like the DC servo to pop up as a contributor to THD. It's entirely possible that you don't need to worry about this. I'm just bringing it up as something to look for as it can be a trap if you allow it to be.
Tom
In the official AN-1192 application note TI recommends using R = 2.2MOhm and C=0.47uF. That results in cutoff freq of 153 MHz.
I am thinking I should use C = 1uF and R=1MOhm to get similar cutoff freq (159MHz)?
Looks like you pushed the cutoff freq in the other direction to 12MHz, even further from the 153MHz? So the official note provides value that will add audible noise? Does it change if you apply feedback to inverted input of the LM3886?
I am thinking I should use C = 1uF and R=1MOhm to get similar cutoff freq (159MHz)?
Looks like you pushed the cutoff freq in the other direction to 12MHz, even further from the 153MHz? So the official note provides value that will add audible noise? Does it change if you apply feedback to inverted input of the LM3886?
Last edited:
MHz = Megahertz (1E6 Hz)
mHz = Millihertz (1E-3 Hz)
There is a difference.
Official app note or not I'm just letting you know of a potential pitfall. I know of it because I've measured it. I'm not saying it is guaranteed to be a problem for you. I'm only suggesting that you take a look once you've built the circuit. That's all.
I wouldn't use a slow start circuit for the regulators.
Tom
mHz = Millihertz (1E-3 Hz)
There is a difference.
Official app note or not I'm just letting you know of a potential pitfall. I know of it because I've measured it. I'm not saying it is guaranteed to be a problem for you. I'm only suggesting that you take a look once you've built the circuit. That's all.
I wouldn't use a slow start circuit for the regulators.
Tom
As far as gain, I am thinking about input voltage of 0.75 V for unbalanced input and 1.4V for the balanced input. I am assuming these are the official voltages for +4 and -10 db levels.
For 4ohms speakers gain calculation:
Av = sqrt(68*4) / 0.75 = 16.49 / 0.75 = 21.98
For 8ohm speakers:
Av = sqrt(50*8) / 0.75 = 20 / 0.75 = 26.66
Gain is set by R14/R10 and R23/R27:
Values I am considering is R10, R27 = 1KOhm and R14, R23 = 22KOhm or 27KOhm
For 4ohms speakers gain calculation:
Av = sqrt(68*4) / 0.75 = 16.49 / 0.75 = 21.98
For 8ohm speakers:
Av = sqrt(50*8) / 0.75 = 20 / 0.75 = 26.66
Gain is set by R14/R10 and R23/R27:
Values I am considering is R10, R27 = 1KOhm and R14, R23 = 22KOhm or 27KOhm
Last edited:
You might want to decrease R32=R33 to 120 Ω so you get the required minimum current flowing in the LM317/LM337. You'll then have to tweak R30=R31 to get the desired output voltage. Even though the LM4562 and TL072 will cause the load current to increase beyond the 10 mA minimum, it's generally better to run a bit of current in the LM317/337. You'll get better transient performance that way.
You don't need D1. Just set R13 = 15-20 kΩ and you'll be fine. That said, you can always short it out later if you want.
Tom
You don't need D1. Just set R13 = 15-20 kΩ and you'll be fine. That said, you can always short it out later if you want.
Tom
That would be dBu not dB. Or so I'm assuming anyway. dB without a reference is meaningless unless we're talking unit-less ratios such as gain.
+4 dBu = 1.23 V RMS
-10 dBu = 245 mV RMS
Handy-dandy calculator: http://www.sengpielaudio.com/calculator-db-volt.htm
Common gains are:
+23 dB (= 14.1 V/V) for balanced
+29 dB (= 28.2 V/V) for unbalanced
+26 dB (= 20 V/V) used to be very common but I get the sense that 23 and 29 dB have taken over. I usually design for 20 dB (= 10 V/V) for better gain structure.
Tom
It looks like the gain of your circuit is 22/1 = 22 V/V (= 26.8 dB) for single-ended input and 23.8 dB for differential input. The amp is inverting, so technically it'll have -22 V/V gain.
Also beware that R9 will likely be the dominant noise contributor in the circuit when the single-ended input is used.
Tom
Also beware that R9 will likely be the dominant noise contributor in the circuit when the single-ended input is used.
Tom
Should I increase gain to 33 V/V? R14, R23 = 33KOhm?
The goal is to set the gain so that amp will output full power at corresponding input levels.
-10 dBu = 245 mV RMS
For 8ohm speakers:
Av = sqrt(50*8) / 0.245 = 20 / 0.245 = 81 ???
+4 dBu = 1.23 V RMS
For 8ohm speakers:
Av = sqrt(50*8) / 1.23 = 20 / 1.23 = 16 ???
OR
+4 dBu = 1.23 V RMS
For 8ohm speakers:
Av = sqrt(50*8) / (1.23/2) = 20 / 0.615 = 32 ???
I am confused
According to this calculator https://circuitdigest.com/calculators/lm317-resistor-voltage-calculator
R30, R31 should be 1.32 KOhms for 15V
I sketched some below. I don't know how to calculate those values. Used website here - https://www.embedded.com/analyzing-circuit-sensitivity-for-analog-circuit-design/
Maybe just will end up using what I have already with R = 1MOhm and C = 2.2 uF.
But just for fun, hope the schematics below is something that can help improve order (slope) at which frequency is filtered out.
Ok, so far this is what I am thinking as far as gain:
Official formula:
Av >= sqrt(Po*Rl)/Vin
where Po - power out, Rl - resistance in load, Vin - input voltage.
With +-35V,
in 8 ohm load I want to get around 60W - that is from the signal of +4dbu or 1.23V. Since I half the input signal, it will be 1.23/2 = 0.615
Av = sqrt(60*8)/0.615 = 21.908/0.615 = 35.62
for 4 ohm load and 100W
Av = sqrt(100*4)/0.615 = 20/0.615 = 32.52
Therefore I am thinking of a gain of 33. Is that correct?
So feedback resistor should be 33Kohm?
I don't know anything about zener diode based voltage regulators. I am doing my reading on it still. I just know that Lm317, lm337 is a good way to regulate voltage, as well as 7815, 7915.
If there is a way to achieve the same result with less parts - I will go for it. I just need to do more research on zener based circuits...
It is just a shunt regulator. The Zener sets the voltage. The resistor determines the maximum current that can be drawn. You need enough current such that the Zener does not drop out of conduction.
In the sim we have 21 milliamp in the resistor and 5 milliamp in Zener. The remaining 16 milliamps (21-5) are what the opamps draw. If you use an opamp that draws more current then use a lower value resistor.
A good start is to say we'll use a 15 volt 1.3 watt Zener per rail.
1.3 watt means that we can allow 86 milliamps to flow (I= W/V). For a 35 volt supply the resistor needs to be (35-15)/0.086 which is 232 ohm. Power dissipation in the resistor is (20*20)/232 which is 1.7 watt. That's quite high but read on*
That gives us a 15 volt rail and if we allow a minimum of 5 milliamps in the Zener then we can draw 81 milliamps from the rail (86ma - 5ma). The more current we draw the less flows in the Zener and its dissipation drops.
*If you know the opamp currents then you can scale the resistor upwards which reduces dissipation.
In the sim we have 21 milliamp in the resistor and 5 milliamp in Zener. The remaining 16 milliamps (21-5) are what the opamps draw. If you use an opamp that draws more current then use a lower value resistor.
A good start is to say we'll use a 15 volt 1.3 watt Zener per rail.
1.3 watt means that we can allow 86 milliamps to flow (I= W/V). For a 35 volt supply the resistor needs to be (35-15)/0.086 which is 232 ohm. Power dissipation in the resistor is (20*20)/232 which is 1.7 watt. That's quite high but read on*
That gives us a 15 volt rail and if we allow a minimum of 5 milliamps in the Zener then we can draw 81 milliamps from the rail (86ma - 5ma). The more current we draw the less flows in the Zener and its dissipation drops.
*If you know the opamp currents then you can scale the resistor upwards which reduces dissipation.