Here you may find the project, photos, circuit, measurement, curves, schematic, full documentation of my headphone audio amplifier based on AD826 opamp
Free for non-commercial purpose only. Please read the disclaimer
http://ygg-it.tripod.com/
Please leave here any your comment that will be very appreciated!
Free for non-commercial purpose only. Please read the disclaimer
http://ygg-it.tripod.com/
Please leave here any your comment that will be very appreciated!
Congrats... I am sure it sounds perfect.
But some questions;
- The input buffer is labelled "Class A". How can you provide that? I couldnt see any bias system for that opamp....
- At the second and output buffer sections, the DC offset is determined 0mv... Also how can you provide that? There is no DC servo and/or DC decoupling on NFB line..
- I cannot see any frequency compansation circuit in the amp.. In this case you can listen a lot of Russian (or Arabic) radio station...
????????
But some questions;
- The input buffer is labelled "Class A". How can you provide that? I couldnt see any bias system for that opamp....
- At the second and output buffer sections, the DC offset is determined 0mv... Also how can you provide that? There is no DC servo and/or DC decoupling on NFB line..
- I cannot see any frequency compansation circuit in the amp.. In this case you can listen a lot of Russian (or Arabic) radio station...
????????
answer..
- The input buffer is labelled "Class A". How can you provide that? I couldnt see any bias system for that opamp....
Already answer in FAQ section
- At the second and output buffer sections, the DC offset is determined 0mv... Also how can you provide that? There is no DC servo and/or DC decoupling on NFB line..
RE: for ideal op-amp in order to minimize DC (goal is zero but 10-20mV at final output is even tolerated) the input (+) and input (-) should see the same DC impedance.
To do this you need that Rin(+)=Rf||Rin(-). And this was done with 1% precision resistor for both the stages. This usually is enough to get a very low DC offset. In real world you have also DC voltage output due to the AD826 input offset voltage, input bias current and input offset current.
These values can range from microvolt to millivolt, can be either polarity and vary among devices along a Gaussian curve: with an accurate selection of Op Amp devices and the right cascade connection you can furtherly minimise the Dc offset. I don't like too much the Dc servo solution since it would feed the opamp with power supply noise.
- I cannot see any frequency compensation circuit in the amp.. In this case you can listen a lot of Russian (or Arabic) radio station...
RE: you are completely right, but fortunately this is not my case... It was very risky and in fact my first project had a small 100pF capacitor in parallel to the input. If you are lucky you can take it away since any additional capacitor is additional distortion...But you should have a very good layout, star grounding circuit and short shielded wiring inside.
In my circuit you may also find three ferrite beads that wrap wires and stop RF from going in (one onto the input cable, the second onto the pot cable and the third onto the main earth wire.
Also remember that the big EPCOS 33uF capacitor has a self resonance of about 320KHz (see Metallized Polyester Film Capacitors (MKT) EPCOS datasheet- B32524 - pag.20 ) due to its intrinsic inductive property. So from this frequency it starts to act like an inductor and this may help too.
- The input buffer is labelled "Class A". How can you provide that? I couldnt see any bias system for that opamp....
Already answer in FAQ section
- At the second and output buffer sections, the DC offset is determined 0mv... Also how can you provide that? There is no DC servo and/or DC decoupling on NFB line..
RE: for ideal op-amp in order to minimize DC (goal is zero but 10-20mV at final output is even tolerated) the input (+) and input (-) should see the same DC impedance.
To do this you need that Rin(+)=Rf||Rin(-). And this was done with 1% precision resistor for both the stages. This usually is enough to get a very low DC offset. In real world you have also DC voltage output due to the AD826 input offset voltage, input bias current and input offset current.
These values can range from microvolt to millivolt, can be either polarity and vary among devices along a Gaussian curve: with an accurate selection of Op Amp devices and the right cascade connection you can furtherly minimise the Dc offset. I don't like too much the Dc servo solution since it would feed the opamp with power supply noise.
- I cannot see any frequency compensation circuit in the amp.. In this case you can listen a lot of Russian (or Arabic) radio station...
RE: you are completely right, but fortunately this is not my case... It was very risky and in fact my first project had a small 100pF capacitor in parallel to the input. If you are lucky you can take it away since any additional capacitor is additional distortion...But you should have a very good layout, star grounding circuit and short shielded wiring inside.
In my circuit you may also find three ferrite beads that wrap wires and stop RF from going in (one onto the input cable, the second onto the pot cable and the third onto the main earth wire.
Also remember that the big EPCOS 33uF capacitor has a self resonance of about 320KHz (see Metallized Polyester Film Capacitors (MKT) EPCOS datasheet- B32524 - pag.20 ) due to its intrinsic inductive property. So from this frequency it starts to act like an inductor and this may help too.
Is it only on vero board? You haven't made any pcb? The circuit looks OK but I have no idea how good the 230 VAC arrangment is.
Why have you excluding the cap in the offset trimming circuit? Isn't it good to block incomming supply noise directly into the audio signal?
You write 21 kHz bandwidth? How do you manage this without any filters and using 50 MHz opamps?
Why have you excluding the cap in the offset trimming circuit? Isn't it good to block incomming supply noise directly into the audio signal?
You write 21 kHz bandwidth? How do you manage this without any filters and using 50 MHz opamps?
You are free to call it whatever you'll like. My stuff is "high performance", "ultra high performance" or even "extremely high performance" but then I can't figure out the next level 
Hi Definition Dual Power Supply
"Hi Definition Dual Power Supply" : check this site to really understand what I mean. It seems that a 100nF cap is even too much!!
lm317 article
"Hi definition" name is used since it is the best compromise to get the best performance for image precision and the less AC noise. There is no marketing intention on it... I tried to go down with less than 100nF but the amplifier started to oscillate. By the way the 100nF cap is the only one which bypass the power supply after the voltage regulator.
"Hi Definition Dual Power Supply" : check this site to really understand what I mean. It seems that a 100nF cap is even too much!!
lm317 article
"Hi definition" name is used since it is the best compromise to get the best performance for image precision and the less AC noise. There is no marketing intention on it... I tried to go down with less than 100nF but the amplifier started to oscillate. By the way the 100nF cap is the only one which bypass the power supply after the voltage regulator.
Softness,
Tone realism,
Dynamic,
Definition,
Bass level,
Bass depth,
Space clarity,
Voice position
Can I read...
If you read carefully the datasheet it says something about power supply bypassing and personally I would be very carefull when I have a 350 V/us with minimal decoupling and a not too fast voltage regulator. The opamp is much faster than the regulator.
If a LM317 is high definition what might a super regulator be which is 20-50 times faster?
Tone realism,
Dynamic,
Definition,
Bass level,
Bass depth,
Space clarity,
Voice position
Can I read...
If you read carefully the datasheet it says something about power supply bypassing and personally I would be very carefull when I have a 350 V/us with minimal decoupling and a not too fast voltage regulator. The opamp is much faster than the regulator.
If a LM317 is high definition what might a super regulator be which is 20-50 times faster?
Lovely articles but I respect his opinions.
http://tech.juaneda.com/en/articles/electrolyticcapacitors.html
My opinion: Don't believe everything you read on the net.
http://tech.juaneda.com/en/projects/preamp3.html
33 kohms into a common base transistor stage?
.... and once more.... surely some advantage but someone must enlighten me.
http://tech.juaneda.com/en/articles/electrolyticcapacitors.html
My opinion: Don't believe everything you read on the net.
http://tech.juaneda.com/en/projects/preamp3.html
33 kohms into a common base transistor stage?
.... and once more.... surely some advantage but someone must enlighten me.
re
First at all thank you very much for all your comments.
1) no idea how good the 230 VAC arrangment is
VAC connection are short, with isolated wire and I try to take them as far as possible from the low voltage circuit and the sides of the box. Transformers are both protected by low value fuses and the cabinet is grounded
Do you mean that veroboard (matrixboard in my case) can be more dangerous
or noiser than a PCB ?
2) Why have you excluding the cap in the offset trimming circuit?
I escluded for three reasons: first I didn't hear any additional hum in my headphone; second the measured SN was the same than the standard circuit; third I did a try with a 100nF, 68nf and 10nF (less in not useful for this purpose): the sondstage become a little narrower. I'm in a point now that I can hear differences such these...
3) You write 21 kHz bandwidth? How do you manage this without any filters and using 50 MHz opamps?
You are right. The frequency response @ -0.1 dB is just the minimum possible I can measure with a PC soundborad which has a sample rate of 44kHz. Surely it is much wider, and unless the circuit starts to oscillate
I found there is no reason to limit it from a sound performance point of view. Than I'm sure that the parasitic element in my circuit will limit the bandwidth: I'm not especting to get the same slew rate as it was indicated in the datasheet.
First at all thank you very much for all your comments.
1) no idea how good the 230 VAC arrangment is
VAC connection are short, with isolated wire and I try to take them as far as possible from the low voltage circuit and the sides of the box. Transformers are both protected by low value fuses and the cabinet is grounded
Do you mean that veroboard (matrixboard in my case) can be more dangerous
or noiser than a PCB ? 2) Why have you excluding the cap in the offset trimming circuit?
I escluded for three reasons: first I didn't hear any additional hum in my headphone; second the measured SN was the same than the standard circuit; third I did a try with a 100nF, 68nf and 10nF (less in not useful for this purpose): the sondstage become a little narrower. I'm in a point now that I can hear differences such these...
3) You write 21 kHz bandwidth? How do you manage this without any filters and using 50 MHz opamps?
You are right. The frequency response @ -0.1 dB is just the minimum possible I can measure with a PC soundborad which has a sample rate of 44kHz. Surely it is much wider, and unless the circuit starts to oscillate
I found there is no reason to limit it from a sound performance point of view. Than I'm sure that the parasitic element in my circuit will limit the bandwidth: I'm not especting to get the same slew rate as it was indicated in the datasheet.
Re: re
Hi Ygg
Let me try to explain my scepticism. I don't dispute such audible differences exist, i've heard similar effects myself. The problem is that once you spend too long fine-tuning a simple circuit, like you've done, your ears become hypersensitive to its sound and greatly exaggerate the differences. I also suspect that you use a very narrow selection of music, as auditioning becomes too time consuming with a wide selection.
In the end, the results are often valid only for a particular track and your particular system/room. Just my opinion.
ygg-it said:the sondstage become a little narrower. I'm in a point now that I can hear differences such these...
Hi Ygg
Let me try to explain my scepticism. I don't dispute such audible differences exist, i've heard similar effects myself. The problem is that once you spend too long fine-tuning a simple circuit, like you've done, your ears become hypersensitive to its sound and greatly exaggerate the differences. I also suspect that you use a very narrow selection of music, as auditioning becomes too time consuming with a wide selection.
In the end, the results are often valid only for a particular track and your particular system/room. Just my opinion.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.