I've been fooling with matchng transistors (which don't stay matched) and adding nulling circuits to eliminate offset, and the servo approach seems to kick-butt!
I'd rather not like to get into a discussion on the pro's and con's of servo's, rather I'd like some input on how one might add a servo such as the one below to an existing amplifier which utilizes a differential pair input.
Yes I searched....all over the web, and here in the forums going back nearly two years. There was some good discussion a while back (1 1/2+ years ago), but the pics don't load and all the links are long dead.
OK...here is how one amp company does it... It works very very well. Here they use a NPN pair...
I'd rather not like to get into a discussion on the pro's and con's of servo's, rather I'd like some input on how one might add a servo such as the one below to an existing amplifier which utilizes a differential pair input.
Yes I searched....all over the web, and here in the forums going back nearly two years. There was some good discussion a while back (1 1/2+ years ago), but the pics don't load and all the links are long dead.
OK...here is how one amp company does it... It works very very well. Here they use a NPN pair...
Attachments
...here is the input stage of an amp I built a while back (PNP input). On this amp I have +/- 15V supplies handy, so would power the servo from there. All I have done here is to redraw the input and add the servo. What else should be changed?
The 56K bias resistor ought to be made larger I assume, and could probably toss the grounding cap on the feedback transistor, but this seems too simple...what did I miss...
Thanks ya'all...
The 56K bias resistor ought to be made larger I assume, and could probably toss the grounding cap on the feedback transistor, but this seems too simple...what did I miss...
Thanks ya'all...
Attachments
In similar amps, I've used non-inverting integrators brought back to the "normal" feedback, with good success. That way, you can get rid of the input cap and be truly direct coupled. Yes, you can get rid of the big elytic in the feedback path, one of the primary benefits of the servo.
I agree that servos are the way to go.
I agree that servos are the way to go.
Should work.
Often you see the servo output attached to the inverting input of the diff pair (with the inputs of the opamp swapped of course). Like SY said, it's an integrator parallel to the normal (negative) feedback path. In your drawing, it's an inverting integrator connected as a positive feedback. Does the same actually and you can't go much simpler anyway.
However, in your circuit the integrator seems to be faster than it should be - this could cause quite audible phase shifts in the bass.
Often you see the servo output attached to the inverting input of the diff pair (with the inputs of the opamp swapped of course). Like SY said, it's an integrator parallel to the normal (negative) feedback path. In your drawing, it's an inverting integrator connected as a positive feedback. Does the same actually and you can't go much simpler anyway.
However, in your circuit the integrator seems to be faster than it should be - this could cause quite audible phase shifts in the bass.
Too fast? The 4.7M amd the .1µf cap make it about 2Hz. Anyway, the example I posted in the first pic is about all the firsthand experiance I have had with servo's.
Maybe if it is too fast I could substitute a .18µf cap for the .1µf (I have a bunch of .18µf caps, but no .1µf).
Lastly, can someone explain to the clueless (me..) why the servo allows you to dump the cap on the feedback path? The reasons do not just jump out at me...
Maybe if it is too fast I could substitute a .18µf cap for the .1µf (I have a bunch of .18µf caps, but no .1µf).
Lastly, can someone explain to the clueless (me..) why the servo allows you to dump the cap on the feedback path? The reasons do not just jump out at me...
Too fast for a DC servo IMHO. You can really hear them. It acts like a simple low pass with fg=2Hz which means it will feed the amp with much higher frequencies than 2Hz, only at a lower level. Disconnect its output from the rest of the circuit and watch it with an oscilloscope while sweeping the amp and you will see what I mean.
Sure, SY was talking about the input cap, not the feedback cap. You won't need it as your DC servo will correct a small DC offset even at the input of the amp (which otherwise would get amplified and could damage your speakers which is why you have the cap)
Another hint: I recommend two anti-parallel or anti-series zeners of say 10V each at the input of the servo opamp to protect it in case your output stage fails.
Well, that cap is superfluous- the servo acts to drop the amp gain to near-zero at DC, so is even better than the cap in that respect. Replace that nasty cap with a piece of your favorite wire. If you decide to use a noninverting integrator, you can get rid of the input cap, too (assuming you absolutely KNOW that nothing feeding the amp can POSSIBLY have output offset or could fail to one of its DC rails). For an inverting integrator, you need that input cap to prevent the servo from trying to feed DC back to the device driving the input.
Two Hz is well low enough. I mean, how low do you expect your speakers to go? If you're worried about contamination of the audio signal from the output of the servo, an RC low pass network with a pole at about 50 Hz can be put between the servo output and the amp node that it's feeding. Since its more than a decade above the servo time constant, it's unlikely to give the amp a stability problem.
Two Hz is well low enough. I mean, how low do you expect your speakers to go? If you're worried about contamination of the audio signal from the output of the servo, an RC low pass network with a pole at about 50 Hz can be put between the servo output and the amp node that it's feeding. Since its more than a decade above the servo time constant, it's unlikely to give the amp a stability problem.
Neither saw the big el'lytic nor your comment about it. It's 5'o clock in the morning here 
You will need the big caps in the servo then. My approach is to take the biggest film caps that I can fit (2.2..10µF, low voltage) and resistors a bit lower than yours (noise).
You can tweak your time constant by listening, too. Select one of those recordings they use to show off their subwoofers and increase your time constant until the bass sounds as clean and tight as without the servo.
Some discussion on this subject:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=9525
You will need the big caps in the servo then. My approach is to take the biggest film caps that I can fit (2.2..10µF, low voltage) and resistors a bit lower than yours (noise).
You can tweak your time constant by listening, too. Select one of those recordings they use to show off their subwoofers and increase your time constant until the bass sounds as clean and tight as without the servo.
Some discussion on this subject:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=9525
Echowars, two things:
1 The base current will create an offset. Which this you can only eliminate a part of the error voltage.
2 At 20 Hz and full power the servo circuit must not saturate (=dist)! This will give you the maximum speed (which is rather slow....)
If you want to correct the offset to 100% you must insert the error voltage at the inverting input or some other place internally, and also have a input bias current servo.
If you want to get inspired you can look here.
Your 4.7 M at the inverting input of the servo is totally useless if you compare the rest.
Note also that these servos creates large initial offsets until they have started to correct.
1 The base current will create an offset. Which this you can only eliminate a part of the error voltage.
2 At 20 Hz and full power the servo circuit must not saturate (=dist)! This will give you the maximum speed (which is rather slow....)
If you want to correct the offset to 100% you must insert the error voltage at the inverting input or some other place internally, and also have a input bias current servo.
If you want to get inspired you can look here.
Your 4.7 M at the inverting input of the servo is totally useless if you compare the rest.
Note also that these servos creates large initial offsets until they have started to correct.
OK...that makes sense.
I kinda like having a cap at the input. Put too much money into my speakers and transistors to roll the dice here. But if I try this (and I more than likely will), will remove the 47µf cap and jumper it.
BTW AMT-Freak...here's one of my speaks... Great Heil on top (about '78 vintage with the veiled diaphrams), and two Dynaudio 20W75's parallelled on the bottom. Biamped with a Kenwood 700M driving the Dynaudio's, and a modified Kenwood KA-3500 driving the Heils. These are my favorite's...
I kinda like having a cap at the input. Put too much money into my speakers and transistors to roll the dice here. But if I try this (and I more than likely will), will remove the 47µf cap and jumper it.
Wow!! Definately time for a nap, my friend...
BTW AMT-Freak...here's one of my speaks... Great Heil on top (about '78 vintage with the veiled diaphrams), and two Dynaudio 20W75's parallelled on the bottom. Biamped with a Kenwood 700M driving the Dynaudio's, and a modified Kenwood KA-3500 driving the Heils. These are my favorite's...
Attachments
Peranders...your site is a wealth of info...much of it beyond me. Regarding the desired DC correction, 100% perfection is not necessary. The first schematic posted is from a working commercial amp, and does a pretty decent job. With the AD820, initial offset at power up is about 40mV and drops quickly (3 seconds or so) to less than 1mV. With high level 20Hz signals into this amp, the servo barely flinches on the scope. It did such a decent job it is what got me interested in the servos to begin with. Also, the couple of amps here that I wanted to try one of these in both have output relays with a turn on delay of about 4-6 seconds, so initial offset is no real big deal.
Was looking to keep this as simple as possible and parts count low. But I may indeed experiment with time constants. As far as I can tell, there is a bit of leeway here, and the few (very few) designs I have seen use values that are all over the map. So I'll just do some parts swapping and see what I can come up with.
Was looking to keep this as simple as possible and parts count low. But I may indeed experiment with time constants. As far as I can tell, there is a bit of leeway here, and the few (very few) designs I have seen use values that are all over the map. So I'll just do some parts swapping and see what I can come up with.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.