Jeff R said:
In my experience, one can certainly use either method in the design of a sonically transparent box of gain. But the servo gives the designer one more tool, with certain obvious advantages- like being able to take the loop around a noninverting amplifier block or a buffer. It's a tradeoff of cost and complexity, and neither are particularly problematic. One has to judge matters design by design as to what's appropriate.
I'll post a schematic once I finish wrestling with the schematic symbols in .bmp that Frank sent me.
In the meantime, I can give you an outline: the servo goes around a hybrid complementry source/cathode follower. This is configured with a topology very similar (OK, exactly the same!) as in the Berning TF-10, but using half of a 6KN8 and a 2N5114 per channel as the complementry pair, with the output taken from the junction of the bottom of the source resistor and the bottom of the cathode resistor. (The gain portion is not enclosed by the servo loop, and is also a 6KN8/2N5114 hybrid, grounded grid, FET configured as source follower directly coupled to tube cathode.)
The servo is an opamp inverting integrator (plain vanilla LF353) with a turnover time constant of 1 second (10 megohm, 0.1 uF). The input to the integrator is the output of the source/cathode follower. The output of the integrator feeds the bottom of the input grid/source leak resistor (also 10 meg, fed by a 0.01 uF input cap) through an RC low pass with an f3 of about 50 Hz.
Power rails are +200 and -15V, the HV fed by a simple regulator with CG electrolytic and polypropylene bypasses coming after a decoupling resistor, the low voltage fed by an opamp-based (LF353 and LH0002) regulator, again through a decoupling resistor and e-lytic and polypropylene bypass caps.
For the scorn of the fashionable, the e-lytics are Mallory CGs, the polypropylenes are Wima MKP.
Clearly, 100% DC feedback is not an option in this application!
In the meantime, I can give you an outline: the servo goes around a hybrid complementry source/cathode follower. This is configured with a topology very similar (OK, exactly the same!) as in the Berning TF-10, but using half of a 6KN8 and a 2N5114 per channel as the complementry pair, with the output taken from the junction of the bottom of the source resistor and the bottom of the cathode resistor. (The gain portion is not enclosed by the servo loop, and is also a 6KN8/2N5114 hybrid, grounded grid, FET configured as source follower directly coupled to tube cathode.)
The servo is an opamp inverting integrator (plain vanilla LF353) with a turnover time constant of 1 second (10 megohm, 0.1 uF). The input to the integrator is the output of the source/cathode follower. The output of the integrator feeds the bottom of the input grid/source leak resistor (also 10 meg, fed by a 0.01 uF input cap) through an RC low pass with an f3 of about 50 Hz.
Power rails are +200 and -15V, the HV fed by a simple regulator with CG electrolytic and polypropylene bypasses coming after a decoupling resistor, the low voltage fed by an opamp-based (LF353 and LH0002) regulator, again through a decoupling resistor and e-lytic and polypropylene bypass caps.
For the scorn of the fashionable, the e-lytics are Mallory CGs, the polypropylenes are Wima MKP.
Clearly, 100% DC feedback is not an option in this application!
Re: James Bongiorno's Response
Maybe, maybe not. Let' say we have a really low noise non-inverting amp with 5 ohms as the "lower" resistor. <10 Hz => 3200 µF, 1 ohms => 16000 µF
Elso Kwak said:Peranders you placed one zero too much. I used a 1000µF electrolytic in the phono amp. The DC-servo was made as in the book by Walt Jung.
Maybe, maybe not. Let' say we have a really low noise non-inverting amp with 5 ohms as the "lower" resistor. <10 Hz => 3200 µF, 1 ohms => 16000 µF
SY said:
The stability comes from the amp itself. If you have much air movements you can get much sub frequencies but the frequency of the servo must be ajusted for the circumstances. You can't designs a DC-servo without using the head and then hope for the best. All servos, no matter what kind must be tuned for best performance.
Re: Re: James Bongiorno's Response
If you quote me, please quote me corectly.
Just as John Curl I used 1000µF in my phono amp <B>NOT</B> 10000µF. Thank you. 
http://marklev.com/marklev/JC2/jc2schematics.jpg
Peranders ,peranders said:
If you quote me, please quote me corectly.
Just as John Curl I used 1000µF in my phono amp <B>NOT</B> 10000µF. Thank you.
http://marklev.com/marklev/JC2/jc2schematics.jpg
Re: Re: Re: James Bongiorno's Response
Yes, but I ment a more extreme amp, with VERY low resistor values. Then the caps tend to get big, which I'm not very comfortable with.
Elso Kwak said:
Peranders ,
If you quote me, please quote me corectly.
Just as John Curl I used 1000µF in my phono amp <B>NOT</B> 10000µF. Thank you.
http://marklev.com/marklev/JC2/jc2schematics.jpg
Yes, but I ment a more extreme amp, with VERY low resistor values. Then the caps tend to get big, which I'm not very comfortable with.
Matching
Even with the original DC-servo in the circuit sound is much better with all four inputtransistors matched for equal gain.
But it still works, I mean you get sound out of your loudspeakers, with non-matched transistors if you retain the servo. That's why I suspect this whole Servo hype made by the GAS company is a production issue. You need a lot of 2N5401 and 2N5551 to find four equal! They made it a commercial issue like "Worlds first Servo Power Amplifier". Enough ranting, the company is long
.
With the servo removed I doubt the amplifier will work with non-matched transistors.
Ampzilla schematic:
http://home.kimo.com.tw/skychutw/ampzilla/schematics/ampzilla2_sch.png
Hi Jeff,Jeff R said:
Even with the original DC-servo in the circuit sound is much better with all four inputtransistors matched for equal gain.
But it still works, I mean you get sound out of your loudspeakers, with non-matched transistors if you retain the servo. That's why I suspect this whole Servo hype made by the GAS company is a production issue. You need a lot of 2N5401 and 2N5551 to find four equal! They made it a commercial issue like "Worlds first Servo Power Amplifier". Enough ranting, the company is long
.With the servo removed I doubt the amplifier will work with non-matched transistors.
Ampzilla schematic:
http://home.kimo.com.tw/skychutw/ampzilla/schematics/ampzilla2_sch.png
Seeking "The Truth".....
"Eric, why did you start this thread in the first place? Plan to make a DC-servo? Or maybe remove one?"
Mainly because of curiosity.
For a long time I have read derisory comments in HiFi mags and on the net regarding DC servo operation - the usual lemmings following each other effect I suspect.
It is refreshing to read that DC servo is not neccesarily sonically bad.
I have always been dubious of the sonic effects of the usual capacitor blocking feedback path to ground (to cause unity DC gain).
In previous experimentation I have found this capacitor to be sonically intefering, and at least needing bypassing with lower values.
You guys have given me some good information to go on regarding future designs - thankyou.
Eric.
"Eric, why did you start this thread in the first place? Plan to make a DC-servo? Or maybe remove one?"
Mainly because of curiosity.
For a long time I have read derisory comments in HiFi mags and on the net regarding DC servo operation - the usual lemmings following each other effect I suspect.
It is refreshing to read that DC servo is not neccesarily sonically bad.
I have always been dubious of the sonic effects of the usual capacitor blocking feedback path to ground (to cause unity DC gain).
In previous experimentation I have found this capacitor to be sonically intefering, and at least needing bypassing with lower values.
You guys have given me some good information to go on regarding future designs - thankyou.
Eric.
Re: Seeking "The Truth".....
If you are worried you can as I and others have pointed out, add one pole extra to reduce the DC-servo's influence even more. If the main pole of the servo is at 1-2 Hz, the other can be at 7-15 Hz.
One important thing is to apply max input voltage at the lowest possible (from signal source) frequency and then measure (or calculate) how much the servo have to work. It's really important that the servo isn't saturated!
BTW: Where is halojoy these days, from 17 posts a day (nearly 1000 posts in one month) to a couple a week?
mrfeedback said:
If you are worried you can as I and others have pointed out, add one pole extra to reduce the DC-servo's influence even more. If the main pole of the servo is at 1-2 Hz, the other can be at 7-15 Hz.
One important thing is to apply max input voltage at the lowest possible (from signal source) frequency and then measure (or calculate) how much the servo have to work. It's really important that the servo isn't saturated!
BTW: Where is halojoy these days, from 17 posts a day (nearly 1000 posts in one month) to a couple a week?
"If you are worried you can as I and others have pointed out, add one pole extra to reduce the DC-servo's influence even more. If the main pole of the servo is at 1-2 Hz, the other can be at 7-15 Hz."
Thankyou, that sounds like good information - even longer time constant ought to be fine I expect, and correct injection point ought to make it sonically irrelevant.
"BTW: Where is halojoy these days, from 17 posts a day (nearly 1000 posts in one month) to a couple a week?"
I think he spat the dummy when the forum powers that be removed the 'Enlightened' status monikor when he got to 997 posts - they stole his thunder you might say.
I think such status monikors are useful in most cases, but certainly not in his case !.
Despite my protest at the time, I laughed my head off when I realised that this was done to him, especially at the last stroke, and I do not miss his BS in the slightest.
Eric.
Thankyou, that sounds like good information - even longer time constant ought to be fine I expect, and correct injection point ought to make it sonically irrelevant.
"BTW: Where is halojoy these days, from 17 posts a day (nearly 1000 posts in one month) to a couple a week?"
I think he spat the dummy when the forum powers that be removed the 'Enlightened' status monikor when he got to 997 posts - they stole his thunder you might say.
I think such status monikors are useful in most cases, but certainly not in his case !.
Despite my protest at the time, I laughed my head off when I realised that this was done to him, especially at the last stroke, and I do not miss his BS in the slightest.
Eric.
Halojoy
<B>He will be back!</B>
Whasn't that Jean-Paul's credo...? No It was "I will be back". I hope
BTW Halojoy has reached his 1000th posts. It is this one:
http://www.diyaudio.com/forums/showthread.php?s=&postid=108358#post1083580
(Memberlist indicates this as his last and the counter is on 1000.)
<B>He will be back!</B>
Whasn't that Jean-Paul's credo...? No It was "I will be back". I hope
BTW Halojoy has reached his 1000th posts. It is this one:
http://www.diyaudio.com/forums/showthread.php?s=&postid=108358#post1083580
(Memberlist indicates this as his last and the counter is on 1000.)
Hi,
A lot has been said here one way or the other. I will abstain from long, lengthy discussions of the why it makes sonic differneces, but rather give my experiences. These where derived with a Phonostage (classic 1 IC MM - namely my the one publiched ages ago as "Analog Addicts Phono Pre").
Op-Amp Phonostages are a classic problem child with regards to offset and offset drift as the DC gain without DC Blocker in the feedback loop is in the region of 60db or more and at the same time the impedances in the feedback loop are often quite low, which makes any DC Blocker neccesarily very large in value. In my case with a 475 Ohm feedback Resistor I'd have had to use around 100uF, hard to come by in Film Capacitors.
In my first round I compared having a classic output coupling Capacitor (Jung/March combo pre "Audio" article with bypass capacitors and biasing). I found that having the same combo as output coupling device sounded subjectively more open, detailed dynamic and natural than having the same combo in the ground leg of the feedback resistor.
I then tried a Servo using the same chip (actually a OPA2604 the Servo Piggy Backed on top of the main chip) as I had used for the Phonostage and with about the same -3db point as the output coupling combo (appx. 165uF into a 10k Pot = 0.096Hz) and worked a lot to optimise the servo. The FET Nature of the OPA2604/604 allows a very high value resistor (10MegaOhm) in the Servo and thus a moderatly small value capacitor (220nF metalised polypropylene), allowing a Film Capacitor to be used in the circuit.
I also optimised the "pull" range of the servo so that extreme conditions (heat/cold - use hairdryer/Ice Spray) just allowed the Phonostage output to remain at 0V with the Servo Output near rail voltage. This way the impact of the servo chip's noise and distortion is minimised, as well as the amplified capacitor distortion.
The disappointing result after all that effort was that subjectively the output coupling combo still had the edge. It sounded less mechanical and more detailed. The first servo I started with (1M/1uF MKT, not optimised pull range) was sonically miles of the simple output coupling combo, though better than the feedback cap.
So, in application where it is possible to have an output coupling combo and to accept some offset (preamps and the like) I would recommend capacitively coupling the output using the full Jung/March Combo or sing good film capacitors if the value suits film capacitors.
I appreciate that in power amplifiers this may not be an option. Hence I would recommend there to first of all optimise the Offset and DC stability of the actual circuit. If bipolar Transistors form the input this means to use equal DC resistances on both inputs. Then make a Servo using a "good sounding" FET Op-Amp with decent DC specs and optimise the "pull" range so the maximum of possible Output Voltage of the servo Amp (usually around +/-10V DC for Op-Amp's on 15V Rails) will just be able to kill the worst offset in the stabilised (burned in and warmed up) circuit. This will minimise the sonic impact of the servo.
Anyway, my few shilling here.
Sayonara
A lot has been said here one way or the other. I will abstain from long, lengthy discussions of the why it makes sonic differneces, but rather give my experiences. These where derived with a Phonostage (classic 1 IC MM - namely my the one publiched ages ago as "Analog Addicts Phono Pre").
Op-Amp Phonostages are a classic problem child with regards to offset and offset drift as the DC gain without DC Blocker in the feedback loop is in the region of 60db or more and at the same time the impedances in the feedback loop are often quite low, which makes any DC Blocker neccesarily very large in value. In my case with a 475 Ohm feedback Resistor I'd have had to use around 100uF, hard to come by in Film Capacitors.
In my first round I compared having a classic output coupling Capacitor (Jung/March combo pre "Audio" article with bypass capacitors and biasing). I found that having the same combo as output coupling device sounded subjectively more open, detailed dynamic and natural than having the same combo in the ground leg of the feedback resistor.
I then tried a Servo using the same chip (actually a OPA2604 the Servo Piggy Backed on top of the main chip) as I had used for the Phonostage and with about the same -3db point as the output coupling combo (appx. 165uF into a 10k Pot = 0.096Hz) and worked a lot to optimise the servo. The FET Nature of the OPA2604/604 allows a very high value resistor (10MegaOhm) in the Servo and thus a moderatly small value capacitor (220nF metalised polypropylene), allowing a Film Capacitor to be used in the circuit.
I also optimised the "pull" range of the servo so that extreme conditions (heat/cold - use hairdryer/Ice Spray) just allowed the Phonostage output to remain at 0V with the Servo Output near rail voltage. This way the impact of the servo chip's noise and distortion is minimised, as well as the amplified capacitor distortion.
The disappointing result after all that effort was that subjectively the output coupling combo still had the edge. It sounded less mechanical and more detailed. The first servo I started with (1M/1uF MKT, not optimised pull range) was sonically miles of the simple output coupling combo, though better than the feedback cap.
So, in application where it is possible to have an output coupling combo and to accept some offset (preamps and the like) I would recommend capacitively coupling the output using the full Jung/March Combo or sing good film capacitors if the value suits film capacitors.
I appreciate that in power amplifiers this may not be an option. Hence I would recommend there to first of all optimise the Offset and DC stability of the actual circuit. If bipolar Transistors form the input this means to use equal DC resistances on both inputs. Then make a Servo using a "good sounding" FET Op-Amp with decent DC specs and optimise the "pull" range so the maximum of possible Output Voltage of the servo Amp (usually around +/-10V DC for Op-Amp's on 15V Rails) will just be able to kill the worst offset in the stabilised (burned in and warmed up) circuit. This will minimise the sonic impact of the servo.
Anyway, my few shilling here.
Sayonara
Re: servo for lm3875...?
Hi,
If building the inverted schematic the question does not arise. If building non-inverted I would think that using a220kOhm Feedback resistor and a 10k Resistor to ground with a 2.2uF...4.7uF Film Capacitor to block DC will sound better than most if not all servos.
It all depends yupon how you are doing things, no "hard and fast" rules exist. Try all options and decide for yourself.
L8er T
Hi,
If building the inverted schematic the question does not arise. If building non-inverted I would think that using a220kOhm Feedback resistor and a 10k Resistor to ground with a 2.2uF...4.7uF Film Capacitor to block DC will sound better than most if not all servos.
It all depends yupon how you are doing things, no "hard and fast" rules exist. Try all options and decide for yourself.
L8er T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.