I've noticed on a Power Amp I have which uses a DC-Servo to correct DC-Offset that the level of offset tends to drift around 0v by a few milli-volts. The oscilliscope trace sort of floats around 0v.
The Servo is a TLO71 with 1M and 470nF resistors and caps. I also tried a OPA134 which gave the same result.
I was wondering if anyone else had seen this effect and if it is considered normal.
Could it be leakage through the 470nF cap?
The Servo is a TLO71 with 1M and 470nF resistors and caps. I also tried a OPA134 which gave the same result.
I was wondering if anyone else had seen this effect and if it is considered normal.
Could it be leakage through the 470nF cap?
Dave said:
------------------------------------------------------------
A few mV is fine. What's the problem? You get this even in preamps.
Dave,
That is normal. Most manufacturers that use servos set the time constant real low so as to prevent the the servo from influencing the sound of the amplifier.
You could play around with the servo to get a faster response (adjust the values of the capacitor) but that usually hurts the sound.
On several amplifiers I have removed the servo and inserted a good quality capacitor in the feedback loop with with better results.
An offset voltage of about up to + or - 25 mv should be fine for a poweramplifier
That is normal. Most manufacturers that use servos set the time constant real low so as to prevent the the servo from influencing the sound of the amplifier.
You could play around with the servo to get a faster response (adjust the values of the capacitor) but that usually hurts the sound.
On several amplifiers I have removed the servo and inserted a good quality capacitor in the feedback loop with with better results.
An offset voltage of about up to + or - 25 mv should be fine for a poweramplifier
This uses integrating feedback, which means that the output has to drift around. This is the circuit working normally. You may be able to reduce the amount of the drift by increasing the gain of the feedback loop, but I don't recommend it. As the others say, a few mv is fine.
Dave:
Maybe you must change the op amp for one bipolar...if you look at the data shets of the bifet op amp TL071 you will see that the ofset voltage change a lot with temperature...a precision op amp
as the OP27 is very good as servo.
Many people in this forum claim about sound diferences betwen op amps as servos... but i have find that the more important is to choose one frequence of integration the more low the better...
In servo operation the DC perfomance of the op amp is more important that the AC operation!Then the choice of op amp must be done with this parameter in first place.
Just my opinion! 😉
Regards
Jorge
Maybe you must change the op amp for one bipolar...if you look at the data shets of the bifet op amp TL071 you will see that the ofset voltage change a lot with temperature...a precision op amp
as the OP27 is very good as servo.
Many people in this forum claim about sound diferences betwen op amps as servos... but i have find that the more important is to choose one frequence of integration the more low the better...
In servo operation the DC perfomance of the op amp is more important that the AC operation!Then the choice of op amp must be done with this parameter in first place.
Just my opinion! 😉
Regards
Jorge
Offset drift with servo circuit.
A servo circuits respond to error voltages. Therefore you must have and error before the circuit can respond. Normally this error voltage will be on the order of a few millivolts and will dance positive or negative over a short periods of time. The actual number of millivolts will be determined by servo loop gain and integration time of the RC network used. The shorter the integration time the worse the low frequency response of the amplifier will become due to modulation the main signal by the DC servo loop..
Normally the output-offset voltage of an audio amplifier is measured with the input of the amplifier shorted to ground or terminated by a specific resistance. This is presents a unnatural condition for the amplifier since it is not used in this manor during normal use. The audio signal normally present at the output of a audio amplifier is none-symmetrical in nature. Thus most servo circuits cannot maintain an amplifiers zero output DC offset voltage once a audio signal is present. Measuring the DC offset of a audio amplifier when there is no audio signal present may not accurately reflect what happens once a audio signal is applied. An improperly designed servo circuit may be worse than no servo circuit at all.
One of my tests is to use a pulse generator to generate only positive or negative going pulses. In the best DC coupled amplifiers utilizing a properly designed DC servo circuit these pulses will be accurately reproduced with no DC offset on the output of the amplifier. While a amplifier with a normal or poorly designed DC servo circuit will have the pulses reproduced along with a offset voltage. The offset voltage being produced is a product of the servo circuit as it tries to compensate for the none-symmetrical input signal that the pulse generator provides. Many amplifiers can sound better without the marginal DC servo circuits.
In one of my past notes I have described a servo circuit that can be adapted to most amplifiers and will give superior results since it is designed to ignore the actual audio signal to be amplified and cares only about the difference between the input and output DC signal when properly adjusted. Thus the circuit will compensate only for amplifier DC drift and is unaffected by the audio present as long as the amplifier is flat from input to output and is not driven into saturation.
I believe that the notes, however brief they may be on this design are still available on my web site, www.audioamps.com ,which be the way my wife wants me to shut down. I have also developed out a suitable circuit board layout for this circuit if anyone is interested in improving their DC servo circuit. All component values must be determined based on the actual application the circuit is the be used in and thus must be determined by the user.
Remember in audio very simple circuits are not always the best. Make them as complicated as needed to get the job done properly.
Best regards,
John Fassotte
A servo circuits respond to error voltages. Therefore you must have and error before the circuit can respond. Normally this error voltage will be on the order of a few millivolts and will dance positive or negative over a short periods of time. The actual number of millivolts will be determined by servo loop gain and integration time of the RC network used. The shorter the integration time the worse the low frequency response of the amplifier will become due to modulation the main signal by the DC servo loop..
Normally the output-offset voltage of an audio amplifier is measured with the input of the amplifier shorted to ground or terminated by a specific resistance. This is presents a unnatural condition for the amplifier since it is not used in this manor during normal use. The audio signal normally present at the output of a audio amplifier is none-symmetrical in nature. Thus most servo circuits cannot maintain an amplifiers zero output DC offset voltage once a audio signal is present. Measuring the DC offset of a audio amplifier when there is no audio signal present may not accurately reflect what happens once a audio signal is applied. An improperly designed servo circuit may be worse than no servo circuit at all.
One of my tests is to use a pulse generator to generate only positive or negative going pulses. In the best DC coupled amplifiers utilizing a properly designed DC servo circuit these pulses will be accurately reproduced with no DC offset on the output of the amplifier. While a amplifier with a normal or poorly designed DC servo circuit will have the pulses reproduced along with a offset voltage. The offset voltage being produced is a product of the servo circuit as it tries to compensate for the none-symmetrical input signal that the pulse generator provides. Many amplifiers can sound better without the marginal DC servo circuits.
In one of my past notes I have described a servo circuit that can be adapted to most amplifiers and will give superior results since it is designed to ignore the actual audio signal to be amplified and cares only about the difference between the input and output DC signal when properly adjusted. Thus the circuit will compensate only for amplifier DC drift and is unaffected by the audio present as long as the amplifier is flat from input to output and is not driven into saturation.
I believe that the notes, however brief they may be on this design are still available on my web site, www.audioamps.com ,which be the way my wife wants me to shut down. I have also developed out a suitable circuit board layout for this circuit if anyone is interested in improving their DC servo circuit. All component values must be determined based on the actual application the circuit is the be used in and thus must be determined by the user.
Remember in audio very simple circuits are not always the best. Make them as complicated as needed to get the job done properly.
Best regards,
John Fassotte
Thanks for the replies they were most helpful. I would have thought the servo would be able to hold the DC-Offset at a constant value of maybe a couple of mV but it appears this is not the case. I had never see this mentioned and was rather surprised to see it.
Tube-Dude, I think a bi-polar op-amp is not such a great idea as the input current it draws will create a voltage drop across the 1 Meg resistor. Also the drifting is too fast to indicate that it is thermally related.
The difference in sound quality between op-amps is probably due to the way the Servo is connected into the amplifier. With the Servo's output driving essentially the - input of the amp to some degree non-linearities in the op-amp output stage may cause distortion at the output of the amp itself.
I think a better way to do it is to get the servo to control the current through a constant current source which would in turn control the DC-Offset.
AlaskanAudio - That's a very clever idea.
Here is the link for those interested -
http://www.audioamps.com/diyprojects/other1/servo01/servo01.1.htm
Its a great site by the way - It would be a pity to shut it down.
Tube-Dude, I think a bi-polar op-amp is not such a great idea as the input current it draws will create a voltage drop across the 1 Meg resistor. Also the drifting is too fast to indicate that it is thermally related.
The difference in sound quality between op-amps is probably due to the way the Servo is connected into the amplifier. With the Servo's output driving essentially the - input of the amp to some degree non-linearities in the op-amp output stage may cause distortion at the output of the amp itself.
I think a better way to do it is to get the servo to control the current through a constant current source which would in turn control the DC-Offset.
AlaskanAudio - That's a very clever idea.
Here is the link for those interested -
http://www.audioamps.com/diyprojects/other1/servo01/servo01.1.htm
Its a great site by the way - It would be a pity to shut it down.
You have got a bunch of well written answers but I will add one thing:
Don't make the servo too slow, then you will get fluctuations from temperature. -3dB aorund 0.1-1 Hz is suitable.
Don't make the servo too slow, then you will get fluctuations from temperature. -3dB aorund 0.1-1 Hz is suitable.
More good info
I have read the comment by peranders and would like to make an additional comment.
I feel that the majority of DC output drift caused temperature variations is to be controlled by temperature compensation circuits and not the DC servo circuit. A well-designed amplifier may very well have various temperature compensation circuits and a DC servo circuit. When one compares temperature and DC offset compensation circuits you might consider temperature compensation to be more like mechanical feedback while DC servo feedback being more electrical in nature.
The mechanical temperature control loop/loops will control the majority of temperature-induced output DC offset voltage and maintain proper output stage bias over a wide temperature range.
The DC servo loop is usually a lot faster than mechanical thermal feedback loops and should need to take care of only minor DC offset voltages that the fairly long time constant temperature control loops cannot compensate for.
Temperature control loops/circuits by nature are slower since we are usually talking about transferring heat in metal parts, sensing the change in temperature and then correcting whatever needs correcting to maintain the proper DC operating points.
When the DC servo loop is disabled in a fairly well designed amplifier the output DC offset should stay very close to zero over the entire temperature range of the amplifier of say 70 the 190 degrees Fahrenheit when the DC servo loop is disabled. Perhaps the happiest point being around 145 degrees Fahrenheit. Close to zero being perhaps less than plus or minus 200 millivolts. If an amplifier is naturally unbalanced or little attention was paid to thermal stability the output DC offset voltage of the amplifier may drift close to one of the supply rails once the DC servo circuit is disabled. Thus the only thing that makes such an amplifier useable is the DC servo circuit, which has to work overtime to correct for design problems. This is not a good situation to be in.
With circuits that use little or no negative feedback, or have high gain the DC stability will be a lot worse so a DC servo circuit may be an absolute necessity in order to regain control over DC balance. Good temperature compensation circuits will help prevent this to a large degree.
The speed of the servo circuit can be adjusted to suit particular requirements and it is not unusual to see a fast loop around a slow loop to help in speeding up settling time. Once the fast loop gets the DC offset close to were it should be that loop is disabled and the slow loop takes control. It should be remembered however that a DC servo circuit, even though they operate at very low frequencies add or subtract arbitrarily from the signal that is being amplified. Thus when using a DC servo circuit to stabilize an amplifier it is wise to have the circuit do as little as possible correction by using good design techniques through out the design. Don’t expect the DC servo circuit to be a cure all.
John Fassotte
Alaskan Audio
I have read the comment by peranders and would like to make an additional comment.
I feel that the majority of DC output drift caused temperature variations is to be controlled by temperature compensation circuits and not the DC servo circuit. A well-designed amplifier may very well have various temperature compensation circuits and a DC servo circuit. When one compares temperature and DC offset compensation circuits you might consider temperature compensation to be more like mechanical feedback while DC servo feedback being more electrical in nature.
The mechanical temperature control loop/loops will control the majority of temperature-induced output DC offset voltage and maintain proper output stage bias over a wide temperature range.
The DC servo loop is usually a lot faster than mechanical thermal feedback loops and should need to take care of only minor DC offset voltages that the fairly long time constant temperature control loops cannot compensate for.
Temperature control loops/circuits by nature are slower since we are usually talking about transferring heat in metal parts, sensing the change in temperature and then correcting whatever needs correcting to maintain the proper DC operating points.
When the DC servo loop is disabled in a fairly well designed amplifier the output DC offset should stay very close to zero over the entire temperature range of the amplifier of say 70 the 190 degrees Fahrenheit when the DC servo loop is disabled. Perhaps the happiest point being around 145 degrees Fahrenheit. Close to zero being perhaps less than plus or minus 200 millivolts. If an amplifier is naturally unbalanced or little attention was paid to thermal stability the output DC offset voltage of the amplifier may drift close to one of the supply rails once the DC servo circuit is disabled. Thus the only thing that makes such an amplifier useable is the DC servo circuit, which has to work overtime to correct for design problems. This is not a good situation to be in.
With circuits that use little or no negative feedback, or have high gain the DC stability will be a lot worse so a DC servo circuit may be an absolute necessity in order to regain control over DC balance. Good temperature compensation circuits will help prevent this to a large degree.
The speed of the servo circuit can be adjusted to suit particular requirements and it is not unusual to see a fast loop around a slow loop to help in speeding up settling time. Once the fast loop gets the DC offset close to were it should be that loop is disabled and the slow loop takes control. It should be remembered however that a DC servo circuit, even though they operate at very low frequencies add or subtract arbitrarily from the signal that is being amplified. Thus when using a DC servo circuit to stabilize an amplifier it is wise to have the circuit do as little as possible correction by using good design techniques through out the design. Don’t expect the DC servo circuit to be a cure all.
John Fassotte
Alaskan Audio
Hello !
Thank you about your comments of the DC-servo ! Can you help me to install it to NAD 310 amplifier ? If you need, I can send the schematic of the NAD 310 to you.
Thank you !
Thank you about your comments of the DC-servo ! Can you help me to install it to NAD 310 amplifier ? If you need, I can send the schematic of the NAD 310 to you.
Thank you !
Similar issue I observe by a multichannel (7-channel) power amp at five channels, which uses a dual-jFET (2SJ109) at the input - both after switch on and after toggle between RCA/XLR-Cannon (unbal/bal) input mode. The serial resistor value between DC-offset output and LTP-input is 10 M-Ohm.
At the both other channels, which use a BjT (2x2SA970) at the input, there is no drift to observe - also both after switch on and after toggle between RCA/XLR-Cannon (unbal/bal) input mode. The serial resistor value between DC-offset output and LTP-input is here only 1 M-Ohm.
Could be this resistor value therefore the reason?
Thank you for advices.
BTW - this thread exist additional in this case:
http://www.diyaudio.com/forums/solid-state/1166-servo-correcting-dc-offset-drift.html#post3579790
Last edited:
The slow movement of a few mv or so is infrasonic noise that results from slow variations in the mains/voltage rails, and gets through to the amplifier output due to finite PSRR. You will get this same infrasonic noise if the amplifier is truly DC coupled and DC offset trimmed.
This infrasonic noise is not a result of the DC servo. If one used the conventional electrolytic capacitor in series with the feedback shunt resistor instead of the servo, and its LF corner was the same as that of the servo, the same infrasonic noise would be observed.
Often the use of the DC servo results in a lower LF corner than what people achieve with the capacitor, but this need not be the case. You can set the DC servo's LF corner to whatever you want.
A good low-noise audio-quality JFET op amp should always be used for the DC servo integrator. A quality film capacitor between 0.1uF and 1.0uF should be used for the integrator. See the DC servo design discussion in my book in chapter 8.
There should be no static temperature-dependent offset if the DC servo is properly designed with a decent op amp. The input offset voltage of the op amp controls the amplifier DC offset, and it should not vary much with temperature. If you use a BJT op amp for the servo integrator (not a good idea), its input bias current may cause some drift under some conditions.
Cheers,
Bob
- Status
- Not open for further replies.