PA100 use three parallel 0.5Ohm resistor to protect output DC voltage
problem. the function of these resistors is a "buffer". let two of LM3886's offset voltage not short directly.
But i think DC Servo is a better solution.
This circuit needs two DC-Servo, one DC-Servo for one LM3886, and
control each LM3886's output offset voltage to zero.
btw, the performance of DC-Servo is the key point! You needs a high impedance , low input offset and enough slew-rate OP if you want to made it with OP. OPA604, OPA134, or LM318 is a good choice.
problem. the function of these resistors is a "buffer". let two of LM3886's offset voltage not short directly.
But i think DC Servo is a better solution.
This circuit needs two DC-Servo, one DC-Servo for one LM3886, and
control each LM3886's output offset voltage to zero.
btw, the performance of DC-Servo is the key point! You needs a high impedance , low input offset and enough slew-rate OP if you want to made it with OP. OPA604, OPA134, or LM318 is a good choice.
If you matched the gain-setting resistors very closely, you "could" leave out the DC Servo. But then you'd need DC blocking capacitors, which will probably not sound as good.
If you use the DC Servo, which is a good idea in my opinion, be aware that the design and implementation of the DC Servo are very important. DC Servos have gotten somewhat of a bad reputation, maybe because of the many poorly-designed ones that have appeared in commercial equipment, and maybe also because it's relatively easy to do them wrong, especially in DIY situations.
You want the DC Servo's integrator's F-3dB to be below 1 Hz. The servo's opamp needs to have high performance, but with very low offsets. The integrating capacitors should be high quality film types, with low Dielectric Absorption, such as those with dielectrics made of teflon, polystyrene, or polypropylene (in order by preference [and probably by cost, too]).
The "better" DC Servo designs have additional low-pass filtering, most-often after the integrator (but sometimes also before the integrator). Usually, the post-filter's F-3dB is set to be about 10x higher in frequency than the F-3dB of the integrator, to ensure stability. You could, for example, split (into two resistors) the resistor shown after the integrator output (in AN-1192) and put a capacitor to ground between the two resistors, to form a passive low-pass filter, there.
However, if you use a dual opamp for the integrator, you'll have a spare opamp. So why not use an active filter?
Here is one example of a DC Servo circuit that I have used:
http://www.fullnet.com/~tomg/DC_SERVO.jpg
For that circuit (which uses a simple 3rd-order active Sallen-Key "DC-accurate" low-pass post-filter): From the integrator input to the amplifier input, the response is 98dB down, from DC up to about .01 Hz. Then it rolls off to about 147dB down at 10 Hz. Then the roll-off steepens to bring it to about 276dB down at 1 kHz, and about 327dB down at 6 kHz. (NOTE that those figures assume that the servo's output resistor's value is 20X the value of the resistor at the amp input that forms a voltage divider with it.)
NOTE that if the amplifier output voltage can swing farther than the servo can handle (which might often be the case), then a voltage divider might need to be used on the servo's input.
Disclaimer: "Other parameters may need to be changed. Your mileage may vary. Use at your own risk."
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
If you use the DC Servo, which is a good idea in my opinion, be aware that the design and implementation of the DC Servo are very important. DC Servos have gotten somewhat of a bad reputation, maybe because of the many poorly-designed ones that have appeared in commercial equipment, and maybe also because it's relatively easy to do them wrong, especially in DIY situations.
You want the DC Servo's integrator's F-3dB to be below 1 Hz. The servo's opamp needs to have high performance, but with very low offsets. The integrating capacitors should be high quality film types, with low Dielectric Absorption, such as those with dielectrics made of teflon, polystyrene, or polypropylene (in order by preference [and probably by cost, too]).
The "better" DC Servo designs have additional low-pass filtering, most-often after the integrator (but sometimes also before the integrator). Usually, the post-filter's F-3dB is set to be about 10x higher in frequency than the F-3dB of the integrator, to ensure stability. You could, for example, split (into two resistors) the resistor shown after the integrator output (in AN-1192) and put a capacitor to ground between the two resistors, to form a passive low-pass filter, there.
However, if you use a dual opamp for the integrator, you'll have a spare opamp. So why not use an active filter?
Here is one example of a DC Servo circuit that I have used:
http://www.fullnet.com/~tomg/DC_SERVO.jpg
For that circuit (which uses a simple 3rd-order active Sallen-Key "DC-accurate" low-pass post-filter): From the integrator input to the amplifier input, the response is 98dB down, from DC up to about .01 Hz. Then it rolls off to about 147dB down at 10 Hz. Then the roll-off steepens to bring it to about 276dB down at 1 kHz, and about 327dB down at 6 kHz. (NOTE that those figures assume that the servo's output resistor's value is 20X the value of the resistor at the amp input that forms a voltage divider with it.)
NOTE that if the amplifier output voltage can swing farther than the servo can handle (which might often be the case), then a voltage divider might need to be used on the servo's input.
Disclaimer: "Other parameters may need to be changed. Your mileage may vary. Use at your own risk."
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Ponz said:
Ponz,
Thanks, very much, for your positive opinion!
And I agree that the capacitors' specs, as well as the opamps' specs, should be very good and pretty-carefully considered, in order to get the very best performance.
I originally developed that circuit for use in my Curve Tracer product's power amplifier, to remove any remaining DC offset from the sweep signal that is applied to the device-under-test. (And by the way, that amplifier design has been powered by an LM1875T since around 1998 (or 1999; I forget), way before I even thought much about using them for audio.) For that application, I used the indicated polypropylene and polyester capacitors, and an Analog Devices OP275 dual opamp, with "good enough" results, since it wasn't really too critical (being, already, actually, "overkill", for that particular amplifier, which was already very precise and very tightly controlled in several other ways).
I just ran some LT-Spice simulations of the DC Servo circuit, again, with an amp model attached. Testing it with no signal, except for a DC offset, it lowers a 1-volt initial DC offset to 50 mV in about 2.25 seconds, and gets it down to 1 mV within about 4.72 seconds. With an initial 1-volt DC offset, the integrator output starts at only +2.5 volts and eventually goes to only -2.5 volts by the time the offset has been removed. So there may be a bit of an "overkill" component for this case and the speed and dynamic range for audio applications might be able to be optimized better (i.e. maybe slower speed and less range could be better, for this application).
With a 22V p-p 60 Hz sine signal included, I can see that the integrator's output (which then still has the same average levels and timing as the "no signal" case, of course) has about 391 mV p-p of the sine riding on it, and the post-filter's output has about 424 uV p-p of the sine, which would mean about 20 uV of the sine would be present at the amp's input pin, after the R/20R divider, whereas the total signal at that input measured 1 V p-p.
So the residual sine component after the filter, at the amp's input, is about 94 dB below the signal at the amp's input, at 60 Hz.
With a 1 kHz sine signal (1 V p-p @ amp input, 22V p-p @ amp output), the integrator's output has a sine component that is about 23mV p-p, while the post-filter's output has a 1 kHz sine component of about 610 nV p-p (or about 29 nV p-p at the amp's input, which is 150 dB below the signal level). And that just gets better, as the frequency is increased.
For a kind-of "worst case", I also ran it with a 20 Hz 22v p-p sine output. For that case, the integrator's output has about 1.2 v p-p of the sine and the post-filter's output has about 43 mV p-p of the 20 Hz sine, which would mean about 2 mV of the sine is at the amp's input pin, versus the 1 V p-p signal, which puts the sine fed back through the servo to the amp's input 54 dB below the signal, at 20 Hz.
In the Curve Tracer product, I only used this DC servo circuit for signals down to 60 Hz. And I also had a response-speed-related design-goal component, for that application. So maybe there's room for improvement, for use with audio and down to 20 Hz.
The post-filter's cutoff frequency can be varied by changing only the capacitor values, more-or-less in proportion to each other, i.e. using standard capacitor values. So perhaps the 0.47uF/0.47uF/2.2uF should be changed to 1uF/1uF/4.7uF, at least for the purists. But then the integrator's caps' or resistors' values would probably also need to be approximately doubled. Or, if that's too slow, maybe 0.68uF/0.68uF/3.3uF could be used for the filter and either the caps' or resistors' values for the integrator could be increased by about 1.5X.
There may also be a small residual DC offset that varies slightly with the configuration, that someone might want to trim out, although it's probably already <= 1 mV.
Sorry to "blather-on", for so long, about all of that. I love this kind of stuff.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
gootee said:
CORRECTION:
The post-filter is NOT a "Sallen-Key" topology.
I stole that filter design directly from Figure 13 of the datasheet for the LTC1968, from http://www.linear.com , several years ago. They don't name the filter topology, there, except to call it "DC Accurate".
AndrewT pointed out in an email that it didn't look like the Sallen-Key topology, which prompted me to search for the datasheet I'd gotten it from. It turns out that they were only _comparing_ it to a Sallen-Key filter, and I somehow got confused about the names.
So, I still don't know the name of the filter's topology. Obviously, the first pole is just a passive RC lowpass section. It's the scond-order part, with the opamp, two capacitors, and two resistors that I'm referring to. To me, it seems to most-closely resemble the classic FDNR filter topology (Frequency-Dependent Negative Resistor). But's it's different from that, too. If anyone knows of a name for that type of filter, please let us know.
At any rate, I do recommend reading that section of the LTC1968 datasheet, if you're interested. You will see exactly why that particular filter topology is perfect for a DC Servo's post-filter.
I apologize for the error.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
DA is hardly a problem in a DC-servo. I have noticed it in a dual-slope ADC, there you had an offset which you could notice just by reading the display.gootee said:
A DC servo doesn't have to be so complicated, a good opamp connected as in integrator.
Example here:
http://sjostromaudio.com/_unsql/hifi/qrp02/downloads_qrp02.html
Enough slew rate is < 15 V /us, you don't need a fast opamp like LM318. OPA134, OPA227, NE5534(!) or slower is a good choice. The opamp should have decent audio properties up to 1 kHz which very many have. In my QRP02 Gainclone I have one half of AD8620, quite an overkill but it does a really good job though.Ponz said:
Hi, peranders.
It's really always just a question of how MUCH of a problem it is. The original application for my servo circuit needed well-under 1 mV of offset. And the WIMA 0.22uF polypropylenes are relatively cheap, anyway. So it's probably reasonable-enough to use them, even if it's just for audio.
I agree that it doesn't HAVE to be so complicated. It just makes it better. When I first made the circuit and used it in the original application, there was a measurable increase in output THD under some conditions, without the additional filtering.
peranders said:
It's really always just a question of how MUCH of a problem it is. The original application for my servo circuit needed well-under 1 mV of offset. And the WIMA 0.22uF polypropylenes are relatively cheap, anyway. So it's probably reasonable-enough to use them, even if it's just for audio.
I agree that it doesn't HAVE to be so complicated. It just makes it better. When I first made the circuit and used it in the original application, there was a measurable increase in output THD under some conditions, without the additional filtering.
peranders said:
Yes, 15V/us is probably more than enough, for most chipamps at least. The slew-rate just needs to be greater than the slew rate of the main amplifier, by some factor (possibly as low as >= 1). However, faster might be better, too, up to some point.
And another very important specification for a servo's integrator opamp is low offset drift.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
peranders said:
Hi, peranders.
In your qrp02r0schema_p2.pdf, what is "LM3386TF"?
(I assume you meant to have "LM3886TF".)
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.