Ok Andrew,
Yeah, I have read those threads. I know what are r8, r3 and the gain I got.
The servo thing is new for me and I´m trying to understand cause I dont want to copy a working one. That´s why I need some help.
I could change the opamp injection by modifing R4 value, right. But, what about R5 and R6?
Now I have seen your edit on post #30. Good explanation! That´s what I needed.
Will increase R4 value on one channel and get back with the new readings.
Thanks!
Yeah, I have read those threads. I know what are r8, r3 and the gain I got.
The servo thing is new for me and I´m trying to understand cause I dont want to copy a working one. That´s why I need some help.
I could change the opamp injection by modifing R4 value, right. But, what about R5 and R6?
Now I have seen your edit on post #30. Good explanation! That´s what I needed.
Will increase R4 value on one channel and get back with the new readings.
Thanks!
Last edited:
R5 is the R of the first RC filter and it is copied into R6 to minimise the imbalance of the opamp inputs.
The filter does not need to change.
The integrator acts as a second filter.
These two reduce the audio content to minuscule amounts and is confirmed in the LTspice simulation.
You have the opportunity to add a third filter at the output of the servo. This further reduces the audio, but more so the noise and distortion added by the servo.
Split R4 roughly in two and add a capacitor from junction to signal ground.
This gives a 3pole effect and can become unstable, so test thoroughly. LTspice can predict instabilities if you are good at asking the right questions. I'm not, so don't ask me.
The filter does not need to change.
The integrator acts as a second filter.
These two reduce the audio content to minuscule amounts and is confirmed in the LTspice simulation.
You have the opportunity to add a third filter at the output of the servo. This further reduces the audio, but more so the noise and distortion added by the servo.
Split R4 roughly in two and add a capacitor from junction to signal ground.
This gives a 3pole effect and can become unstable, so test thoroughly. LTspice can predict instabilities if you are good at asking the right questions. I'm not, so don't ask me.
My Rf is 50k, so I had 500k injection resistors. (10times Rf as manufacturer recommended at the AN-1192 sheet, page 21)
I doubled those values with 1M resistors.
Now I get around 1,2 volts and 0,6 volts opamp outputs.
Still reading the same offset as before. 0,3mV and 0,8mV.
You think those values are better? Should I increase even more injection resistor?
I want the servo to be reliable, even at the expense of some audio quality loss.
Thanks!
I doubled those values with 1M resistors.
Now I get around 1,2 volts and 0,6 volts opamp outputs.
Still reading the same offset as before. 0,3mV and 0,8mV.
You think those values are better? Should I increase even more injection resistor?
I want the servo to be reliable, even at the expense of some audio quality loss.
Thanks!
Last edited:
Rf=50k seems like the upper leg NFB
and 1k0 lower leg gives you the 50times gain you mentioned earlier.
Is that right?
Using 510k for the injection resistor (on the output of the servo opamp) with the 1k0 NFBlower gives a reduction of 1k/(510k+1k) = 0.001957.
If the maximum output of the servo is 12V, then the maximum correction at the -IN node becomes 12V*0.001957 = 0.0235V = 23.5mV
That is quite a low correction limit. I would not go any higher in value. Maybe too high already.
A very low correction range (here +-23.5mVpk) means that sometimes the opamp goes to full rail output and becomes unable to apply any further correction. That's what I was describing in post30
Changing the injection by doubling it and the output offsets remaining the same as before, tells me that the offset at the outputs have become dependant on the servo opamp parameters.
and 1k0 lower leg gives you the 50times gain you mentioned earlier.
Is that right?
Using 510k for the injection resistor (on the output of the servo opamp) with the 1k0 NFBlower gives a reduction of 1k/(510k+1k) = 0.001957.
If the maximum output of the servo is 12V, then the maximum correction at the -IN node becomes 12V*0.001957 = 0.0235V = 23.5mV
That is quite a low correction limit. I would not go any higher in value. Maybe too high already.
A very low correction range (here +-23.5mVpk) means that sometimes the opamp goes to full rail output and becomes unable to apply any further correction. That's what I was describing in post30
Changing the injection by doubling it and the output offsets remaining the same as before, tells me that the offset at the outputs have become dependant on the servo opamp parameters.
Last edited:
I have. If the DC servo stop band attenuation isn't low enough, you'll see a degradation in THD at low frequency.
I have not seen issues with DC servo settling and turn on/off transients. The settling time of the servo tends to be long enough that severe clicks and pops don't happen.
I have not seen any impact of DC servos on HF stability of amplifiers. It is possible to get some peaking in the LF end if you aren't careful with the gain of the DC servo path (so set R4 accordingly).
No. R8 really is necessary. The input of U2 does not draw any current (assuming ideal opamp), so without R8, the voltage across R3 would be zero and you'd have a unity gain buffer (and an unstable amp as the LM3886 is not unity gain stable).
Tom
Thanks for your "input".
I still see U2 output as ground as far as AC feedback signal is concerned, DC as far as the servo is concerned.
U2 has 100% feedback as far as AC is concerned through C4. It's output will draw current to correct for any deviation = any AC.
Therefore that point is "ground" from an AC point of view.
Connected to this ground is R4 therefore it would act as the feedback divider in conjunction with R3.
Removing R8 and making R4 the original R8 value I think should work?
I think I'll have to dig out one of my Sperry PCBs modify it accordingly and give it a go regardless....It might well be I'm overlooking something obvious here.... Fire extinguisher at the ready!!
Last edited:
Hi again,
glo, I think that if you remove R8, you won´t control the gain, it would depend on opamp output. Anyway, let me know if you try your experiment.
Btw, I´ve been playing and tweaking the amp those days.
Tried it at a friend´s living room and the input was too sensitive for his laptop (Andrew, you were right, 34db was too much) and you could hear the hiss. not an option.
Lowered feedback resistor to 27k, and R1 to 27k too. Lost some volume but now is silent. No hum, and no hiss.
As you suggested, lowered opamp injection resistor. From 500k to 100k and now get an offset of 0,2-0,3mV.
The opamp output voltage is even lower when the amp is steady, and if I touch the input cables or disconnect input jumper, the max output I´ve seen is around 1,5-2 volts. You think is right or should I lower it even more.
The amp sound is crystal clear but... not enough lows. Tried against other amps and yeah, this amp is so detailed, but very light on bass.
Im using 24v toroid (34,5v dc output) 2x10.000uf per rail at the ps. (unregulated, snubberized ps) and 1000uf, 33uf and 0,1uf onboard and 180pf capacitor between pins 9-10.
Zobel network has a 1uH in parallel with 10R resistor and 5ohm resistor in series with 220nf.
My speakers are rated at 8 ohms. 95db sens.
What do you think I could change to improve bass response? You think I have wrong values there?
Hope you can help.
Thanks!
glo, I think that if you remove R8, you won´t control the gain, it would depend on opamp output. Anyway, let me know if you try your experiment.
Btw, I´ve been playing and tweaking the amp those days.
Tried it at a friend´s living room and the input was too sensitive for his laptop (Andrew, you were right, 34db was too much) and you could hear the hiss. not an option.
Lowered feedback resistor to 27k, and R1 to 27k too. Lost some volume but now is silent. No hum, and no hiss.
As you suggested, lowered opamp injection resistor. From 500k to 100k and now get an offset of 0,2-0,3mV.
The opamp output voltage is even lower when the amp is steady, and if I touch the input cables or disconnect input jumper, the max output I´ve seen is around 1,5-2 volts. You think is right or should I lower it even more.
The amp sound is crystal clear but... not enough lows. Tried against other amps and yeah, this amp is so detailed, but very light on bass.
Im using 24v toroid (34,5v dc output) 2x10.000uf per rail at the ps. (unregulated, snubberized ps) and 1000uf, 33uf and 0,1uf onboard and 180pf capacitor between pins 9-10.
Zobel network has a 1uH in parallel with 10R resistor and 5ohm resistor in series with 220nf.
My speakers are rated at 8 ohms. 95db sens.
What do you think I could change to improve bass response? You think I have wrong values there?
Hope you can help.
Thanks!
Last edited:
Do you have a computer sound card? If so, all you need is some software. Holm Impulse (PC) or FuzzMeasure (Mac) will work. You'll need to divide the amp output with a voltage divider before feeding it to the sound card. I'd probably go with something like a 30:1 divider, so about 290 Ω (or 287 Ω if you prefer 1% resistor values) in series with 10 Ω.
Sweep the frequency and measure the amplifier output. It should be pretty darn flat (say ±0.5 dB) within the range of 20 Hz to 20 kHz.
Alternatively, if you have an AC voltmeter that covers the entire audio range (many don't) you can go to WavTones and generate some test tones for 20, 50, 100, 200, 500, 1k, 2k, 5k, 10k, and 20 kHz. Make the tones five seconds long (max supported by the free version of WavTones). Play them on repeat using your media player and adjust the volume such that you have 1.0 V AC going into the amp. Load the amp output with an 8 Ω power resistor. Measure the output voltage for the different tones.
I have no idea what "full" means to you. Bass lift? Higher THD? Frequency dependent THD? Without this knowledge and a measurement of how your amp is behaving, I can't make suggestions for component changes.
Tom
Did some measures as follows:
Oscillator triangle wave output set at 250mv output.
Put 8ohm 50w resistor at the amp output and conected the voltmeter and scope to it.
Tested at 20, 200, 2kHz and 20kHz.
The output voltage at the resistor is 7,10 volts.
Drop the oscillator frequency and the voltage keeps constant. At low freq the voltmeter could drop around 6,95volts. I think this is related to the voltmeter refresh rate.
On the scope, the output wave looks pretty good. No distortion or amplitude drop.
I think its working as it should, but Im missing more bass.
The sound is too bright.
After those readings, what do you think I could do to "tune" it?
Thanks!!
Oscillator triangle wave output set at 250mv output.
Put 8ohm 50w resistor at the amp output and conected the voltmeter and scope to it.
Tested at 20, 200, 2kHz and 20kHz.
The output voltage at the resistor is 7,10 volts.
Drop the oscillator frequency and the voltage keeps constant. At low freq the voltmeter could drop around 6,95volts. I think this is related to the voltmeter refresh rate.
On the scope, the output wave looks pretty good. No distortion or amplitude drop.
I think its working as it should, but Im missing more bass.
The sound is too bright.
After those readings, what do you think I could do to "tune" it?
Thanks!!
Yes,
I have no input capacitor so the circuit only has two filters.
The first would be r2/c1 but is a low pass filter, so it should not hurt lows, right?
The other one is the thiele network, and not sure if it affect lows too.
I have read that using big decoupling caps helps there, but Im already using 1000uf onboard caps.
Btw, power supply is very close to the amps so used very short power cables.
Still lost
I have no input capacitor so the circuit only has two filters.
The first would be r2/c1 but is a low pass filter, so it should not hurt lows, right?
The other one is the thiele network, and not sure if it affect lows too.
I have read that using big decoupling caps helps there, but Im already using 1000uf onboard caps.
Btw, power supply is very close to the amps so used very short power cables.
Still lost
Why triangle? You'll get a more reliable measurement with a sine wave - especially with cheaper AC voltmeters.
That seems likely.
You'll only see gross distortion (>> 1 %) on a scope. Audible distortion shows up in the spectrum measurements way before you notice anything on an oscilloscope.
Is it possible that your "reference" amp has a bass boost or has higher THD at lower frequencies than your DIY amp? That would cause the "reference" amp to sound warm (deeper and boosted bass), which tends to sell well in hifi stores. Once you're used to such bass, you might not like the precise bass that you can get from a DIY amp.
You could repeat the amplitude vs frequency measurement for your "reference" amp.
Tom
Tom,
To be honest, my oscillators needs some calibration (and Im a bit lazy).
The sine is not perfectly symmetrical, so thought triangle wave could do the trick.
Anyway, I think I got accurate readings (not using cheap voltmeter)
Maybe youre right and my ears are used to the other amps. Anyway, I miss more bass in some recordings.
Do you think a preamp could help here?
Im learning a lot with that project but it is such a big challenge to get the "right sounding amp"
Andrew, Tom, thanks again for your help.
To be honest, my oscillators needs some calibration (and Im a bit lazy).
The sine is not perfectly symmetrical, so thought triangle wave could do the trick.
Anyway, I think I got accurate readings (not using cheap voltmeter)
Maybe youre right and my ears are used to the other amps. Anyway, I miss more bass in some recordings.
Do you think a preamp could help here?
Im learning a lot with that project but it is such a big challenge to get the "right sounding amp"
Andrew, Tom, thanks again for your help.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.