Citation II hack
- Thread starter s-petersen
- Start date
S- You are making real progress now. With no step network installed, it looks like your FB cap is ever so slightly a little high. You might check out the thread on my Fisher SA-100 clone I recently built, where in I show a pic of the 10 kHz sq wave that unit produces. It is a text book example and can well serve as a reference to go by if you have no other references.
It appears that the step network will likely be able to be greatly reduced from the original values. Those values are best determined in conjunction with HF stability tests once the values of the FB network are set.
With pentode input stages, external level controls should have minimal impact on a 10 kHz sq wave presentation. You might check your cabling to see if it is not introducing excessive capacitance that affects the display.
It also appears that there is a significant amount of noise present in your signals. Hopefully that is something you can chase down as well.
Keep up the great work!
Dave
It appears that the step network will likely be able to be greatly reduced from the original values. Those values are best determined in conjunction with HF stability tests once the values of the FB network are set.
With pentode input stages, external level controls should have minimal impact on a 10 kHz sq wave presentation. You might check your cabling to see if it is not introducing excessive capacitance that affects the display.
It also appears that there is a significant amount of noise present in your signals. Hopefully that is something you can chase down as well.
Keep up the great work!
Dave
s-petersen
Scott
I watched that thread,I only wish I could make things as nice and neat as you did,I thought that I was trying for no bump on the leading edge,if I reduce the cap,I get more ringing,if I increase it I get rounding.
The pic in your thread is the square wave I'm looking for?
What does the step network affect besides rounding the leading edge as it did.
My pics looked worse than real life,as far as line thickness,My camera doesn't like low light without flash,it pretty much over exposes when I try to tahe a pic,I'll have to try manual mode to get better pics
The pic in your thread is the square wave I'm looking for?
What does the step network affect besides rounding the leading edge as it did.
My pics looked worse than real life,as far as line thickness,My camera doesn't like low light without flash,it pretty much over exposes when I try to tahe a pic,I'll have to try manual mode to get better pics
Last edited:
s-petersen
Scott
When I adjusted the cap,I can only change pretty much the first spike on the square wave,the sucessive ripples don't change all that much
s-petersen
Scott
I think you are right about the cabling,I have a little bump at the input,I'll have to look into it
s-petersen
Scott
I think the generic,condensed version ,or the last parts of the thread,would be a great Idea
Last edited:
s-petersen
Scott
I think I need to reduce my gain,currently I get full output @.072v gives 22v across the output(16 ohm)gives 100ma at the tube
S -- your gain does seem excessive -- and this with a verified 20 db of NFB? Whoa! Unless you are looking to use this unit with a passive pre, then some sort of input attenuator will be a must to get any reasonable S/N ratio in the final product.
OK. The 10K sq wave I referenced in my clone thread with the small hump followed by a smooth top is in fact the the desired final appearance.
The FB cap by itself will not produce a flat top -- the step network will take care of that. The goal of the FB cap is to create a flat loop response, best adjusted by the method I mentioned earlier. If your indicator response is not flat enough to adjust the cap that way, then using the pic from my clone will suffice as a good reference to adjust for. Again, with regards to a 10 kHz sq wave presentation, the FB cap primarily affects the size of the leading edge hump. By itself, it will not dampen the top of the wave.
The step network affects the amplitude/phase response of the amplification stages, and works in combination with the loop R/C components to produce the most stable transient response. When both networks are properly tailored, the circuit will achieve critical damping and display a flat wave top presentation.
Once you are satisfied you have your loop R/C components properly chosen, then we will move on to the step network and check the level of HF stability achieved.
Understand that developing a stable global NFB network around a basic amplification block is the hardest part of amplifier design. A great feedback system can make a good amplifier wonderful, while a poor feedback system will ruin any amplifier no matter how good the basic amplifier is. This is the reason that many people either shy away from feedback altogether, or only use it sparingly. Too much feedback can in fact be used -- and you saw the results of that first hand. But dispensing with it altogether reduces performance dramatically. Properly applied and in the correct amounts, you can achieve the most of its benefits, while minimizing its drawbacks.
The procedures used to achieve a "proper" feedback system is only exceeded in its vagueness by the definition of what a proper feedback system is.
You can make a feedback system that targets low distortion, or one that favors stability -- or one that favors frequency response -- or -- or.......
And therein lies a good portion of the problem. What does a well balanced feedback system produce in terms of these criteria? In the early days, extended frequency response and low distortion was favored to make the specifications sell the product -- until such designs started burning out tweeters at an alarming rate due to their instability. Oops.
Then they went the other direction favoring stability, except that then the designs sounded dull and unexciting. Even when the poor bloke came in to buy a new tweeter, they couldn't give that stuff away. Unfortunately, about the time they could have really gotten it right, the transistor came along and put the whole issue out of its misery. It simply faded away with the switch from vacuum tube to SS designs.
The point is, that creating a really good feedback system then is all about achieving the best mix of the various criteria mentioned -- and doing that involves as much art as it does science.
I mention all of this because the process I am giving you is of necessity, my own process that I have developed over many years of working with FB amplifiers, and targets the various criteria in a mix that I have found to be most practical and useful. I have found it to produce very satisfying sonic and measured results, and others who have heard the results of this approach seem to agree.
There are other approaches that can be used as well which involve measuring phase angles, circuit modeling, and the like, but at some point, all these methods boil down to trial and error tailoring of the final components chosen for the targeted mix of results.
I just though you might like to know about the can of worms you've opened!
Dave
OK. The 10K sq wave I referenced in my clone thread with the small hump followed by a smooth top is in fact the the desired final appearance.
The FB cap by itself will not produce a flat top -- the step network will take care of that. The goal of the FB cap is to create a flat loop response, best adjusted by the method I mentioned earlier. If your indicator response is not flat enough to adjust the cap that way, then using the pic from my clone will suffice as a good reference to adjust for. Again, with regards to a 10 kHz sq wave presentation, the FB cap primarily affects the size of the leading edge hump. By itself, it will not dampen the top of the wave.
The step network affects the amplitude/phase response of the amplification stages, and works in combination with the loop R/C components to produce the most stable transient response. When both networks are properly tailored, the circuit will achieve critical damping and display a flat wave top presentation.
Once you are satisfied you have your loop R/C components properly chosen, then we will move on to the step network and check the level of HF stability achieved.
Understand that developing a stable global NFB network around a basic amplification block is the hardest part of amplifier design. A great feedback system can make a good amplifier wonderful, while a poor feedback system will ruin any amplifier no matter how good the basic amplifier is. This is the reason that many people either shy away from feedback altogether, or only use it sparingly. Too much feedback can in fact be used -- and you saw the results of that first hand. But dispensing with it altogether reduces performance dramatically. Properly applied and in the correct amounts, you can achieve the most of its benefits, while minimizing its drawbacks.
The procedures used to achieve a "proper" feedback system is only exceeded in its vagueness by the definition of what a proper feedback system is.
You can make a feedback system that targets low distortion, or one that favors stability -- or one that favors frequency response -- or -- or.......
And therein lies a good portion of the problem. What does a well balanced feedback system produce in terms of these criteria? In the early days, extended frequency response and low distortion was favored to make the specifications sell the product -- until such designs started burning out tweeters at an alarming rate due to their instability. Oops.
Then they went the other direction favoring stability, except that then the designs sounded dull and unexciting. Even when the poor bloke came in to buy a new tweeter, they couldn't give that stuff away. Unfortunately, about the time they could have really gotten it right, the transistor came along and put the whole issue out of its misery. It simply faded away with the switch from vacuum tube to SS designs.
The point is, that creating a really good feedback system then is all about achieving the best mix of the various criteria mentioned -- and doing that involves as much art as it does science.
I mention all of this because the process I am giving you is of necessity, my own process that I have developed over many years of working with FB amplifiers, and targets the various criteria in a mix that I have found to be most practical and useful. I have found it to produce very satisfying sonic and measured results, and others who have heard the results of this approach seem to agree.
There are other approaches that can be used as well which involve measuring phase angles, circuit modeling, and the like, but at some point, all these methods boil down to trial and error tailoring of the final components chosen for the targeted mix of results.
I just though you might like to know about the can of worms you've opened!
Dave
s-petersen
Scott
Kind of like trying to open a good combination lock without the combination.
I tried it on my test speaker,and a Dynaco preamp "for the hell of it"and I had to use 2meg in with 100k shunt and it was still a little too sensitive,but on the C20 it was pretty low,I may have to put 100k across the input to give it a more usual input impedance and go back to 1 meg.
It is a little bass heavy,treble shy,but in every other way sounds good
The excessive gain was at the lowest output of my siggen,and might have had something to do with the interference.
I have to take a few days away from this project,I work a 12 hour shift,so when I get home I am usually too tired,and have too little time to get anything accomplished.Back at it Saturday
Can you give some clue as to what shape on the square wave resistance affects,vs cap.
I really appreciate the help you are giving me!,Thanks
I tried it on my test speaker,and a Dynaco preamp "for the hell of it"and I had to use 2meg in with 100k shunt and it was still a little too sensitive,but on the C20 it was pretty low,I may have to put 100k across the input to give it a more usual input impedance and go back to 1 meg.
It is a little bass heavy,treble shy,but in every other way sounds good
The excessive gain was at the lowest output of my siggen,and might have had something to do with the interference.
I have to take a few days away from this project,I work a 12 hour shift,so when I get home I am usually too tired,and have too little time to get anything accomplished.Back at it Saturday
Can you give some clue as to what shape on the square wave resistance affects,vs cap.
I really appreciate the help you are giving me!,Thanks
Last edited:
No problem. We can start up again when you are ready.
Within the FB R/C network, the R component sets the basic level of feedback -- usually adjusted at 1 kHz for the desired FB level sought -- in this case, a ten times reduction for 20 db NFB. The C component helps to correct the HF phase response of the OPT, best adjusted as we've already discussed.
Dave
Within the FB R/C network, the R component sets the basic level of feedback -- usually adjusted at 1 kHz for the desired FB level sought -- in this case, a ten times reduction for 20 db NFB. The C component helps to correct the HF phase response of the OPT, best adjusted as we've already discussed.
Dave
s-petersen
Scott
oops..,I should have clarified,in the step network.
It will be easier to go over that when we get to that process once the basic loop components are settled.
Dave
Dave
s-petersen
Scott
I was just curious,this amp has so much more gain than the original,is it because the output transformer has only about 2/3 (3800/6000)the ohms as the original?
The original circuit for the input had a .05 input cap for low freq roll off,the manual said it served as a rumble filter,and could be bypassed,so I didn't include it.Can dc offset cause any problems?Some older pas preamps have this problem,(mine is a 3x,the problem was corrected)If I use an input cap,
what size would you recommend.
One nice thing about doing feedback this way,the math shy will love it,not my favorite subject.
At work,we had a bonified rocket scientist(Patriot system),he was an electronic tech/engineer,he was the go-to guy to help with setting up the math.He retired,I miss him,as a source,and a Friend
The original circuit for the input had a .05 input cap for low freq roll off,the manual said it served as a rumble filter,and could be bypassed,so I didn't include it.Can dc offset cause any problems?Some older pas preamps have this problem,(mine is a 3x,the problem was corrected)If I use an input cap,
what size would you recommend.
One nice thing about doing feedback this way,the math shy will love it,not my favorite subject.
At work,we had a bonified rocket scientist(Patriot system),he was an electronic tech/engineer,he was the go-to guy to help with setting up the math.He retired,I miss him,as a source,and a Friend
Last edited:
Scott -- The Mullard design (which you are using) has been as roundly praised for its performance capabilities as it has been criticized for its unbearably excessive gain.
To put it into perspective, the same Mullard circuit you are using is the basis for many of the Eico amplifiers -- except that Eico wisely used a 6SN7 in place of the 12AX7 for inverter service that not only provided a lower impedance (i.e. better) drive for the output stage, but also cut a big chunk of the gain out of that circuit. But still -- even with that modification, the Eico amplifiers still require only .5 volt input to generate full power output. This is still too sensitive for my tastes, causing problems with noisy interconnects and amplifying the quiescent noise from the preamp. As a result, Eico still had to provide input level controls to contain the still excessive gain from the design.
In your case, the gain issue is a bit of a big elephant in the middle of the room. Even with the FB issues dealt with, you're still left will all that gain. To that point, in my opinion, using large attenuation networks ahead of the input stage is poor design. It's a huge invitation to reducing HF performance and loss of detail. More directly, why produce all that gain, only to have to compensate for it (i.e attenuate it) somewhere else?
If it is the problem for you I think it is, it should really be dealt with first -- and properly -- before moving forward. If it is to be dealt with through basic circuit design, then just like using the actual OPT when developing the feedback/stability circuits, the actual amplifying block should be used as well. Otherwise, just like before, you've got to do it all over again when the actual circuit is used.
I would strongly consider converting your 12AX7 inverter tube to a 6FQ7/6CG7 (basically a miniature version of the 6SN7). This tube uses the same size socket, and can be converted with simple rewiring. It will require new components associated with that tube, but it will also be a big step in the right direction regarding this problem. Then, if the gain is still excessive for you, we can reconfigure the input stage to operate in triode mode. This -- along with the inverter change -- will deal a serious blow to all that unnecessary gain, and almost certainly eliminate the need for any input attenuation circuits -- which you would not want to use with a triode type input tube anyway.
Regarding the function of the step network components, both components work together in harmony -- in conjunction with the circuit impedance present -- to control not only the point at which the HF begins to roll off, but also the amount of overall attenuation the network provides at those frequencies. Therefore, a change in one component within the network requires a change in the other component to achieve both goals being targeted. For example, if a given network provides the correct roll off frequency, but not enough attenuation, then reducing the R component will provide more attenuation, but also change the "corner" frequency of the network. So, the C component must then be adjusted to compensate, etc., -- and so it goes.
As I said, this is quite an ambitious project, but one if you see through, you will also learn a lot from.
Dave
To put it into perspective, the same Mullard circuit you are using is the basis for many of the Eico amplifiers -- except that Eico wisely used a 6SN7 in place of the 12AX7 for inverter service that not only provided a lower impedance (i.e. better) drive for the output stage, but also cut a big chunk of the gain out of that circuit. But still -- even with that modification, the Eico amplifiers still require only .5 volt input to generate full power output. This is still too sensitive for my tastes, causing problems with noisy interconnects and amplifying the quiescent noise from the preamp. As a result, Eico still had to provide input level controls to contain the still excessive gain from the design.
In your case, the gain issue is a bit of a big elephant in the middle of the room. Even with the FB issues dealt with, you're still left will all that gain. To that point, in my opinion, using large attenuation networks ahead of the input stage is poor design. It's a huge invitation to reducing HF performance and loss of detail. More directly, why produce all that gain, only to have to compensate for it (i.e attenuate it) somewhere else?
If it is the problem for you I think it is, it should really be dealt with first -- and properly -- before moving forward. If it is to be dealt with through basic circuit design, then just like using the actual OPT when developing the feedback/stability circuits, the actual amplifying block should be used as well. Otherwise, just like before, you've got to do it all over again when the actual circuit is used.
I would strongly consider converting your 12AX7 inverter tube to a 6FQ7/6CG7 (basically a miniature version of the 6SN7). This tube uses the same size socket, and can be converted with simple rewiring. It will require new components associated with that tube, but it will also be a big step in the right direction regarding this problem. Then, if the gain is still excessive for you, we can reconfigure the input stage to operate in triode mode. This -- along with the inverter change -- will deal a serious blow to all that unnecessary gain, and almost certainly eliminate the need for any input attenuation circuits -- which you would not want to use with a triode type input tube anyway.
Regarding the function of the step network components, both components work together in harmony -- in conjunction with the circuit impedance present -- to control not only the point at which the HF begins to roll off, but also the amount of overall attenuation the network provides at those frequencies. Therefore, a change in one component within the network requires a change in the other component to achieve both goals being targeted. For example, if a given network provides the correct roll off frequency, but not enough attenuation, then reducing the R component will provide more attenuation, but also change the "corner" frequency of the network. So, the C component must then be adjusted to compensate, etc., -- and so it goes.
As I said, this is quite an ambitious project, but one if you see through, you will also learn a lot from.
Dave
Kegger
R.I.P. 1/12/1966 - 6/1/2017
I'm with Dave on this, and I was hinting to much gain early on here, but my suggested
mod used the 12AT7 to keep gain similar to what you had and or seemed to want here.
But now like Dave(and as you see)..
Before I finished all of your feedback and adjusting there I'd "fix" the circuit, an use as
Dave suggests an 6CG7(or 12AU7/12BH7) instead of the 12AT7 in my posted version or
the 12AX7 that you are using now in yours as wired,
(redoing the splitter section)
To use 6CG7(or 12AU7/12BH7) in the version I posted just use a 10K pot in place of the
20k pot, with the rest of changes I had there and your set. (less gain and better drive)
Than again as Dave suggests if you do still feel you have to much gain the EF86 can be
triode strapped to further reduce the overall circuit gain there.
.
mod used the 12AT7 to keep gain similar to what you had and or seemed to want here.
But now like Dave(and as you see)..
Before I finished all of your feedback and adjusting there I'd "fix" the circuit, an use as
Dave suggests an 6CG7(or 12AU7/12BH7) instead of the 12AT7 in my posted version or
the 12AX7 that you are using now in yours as wired,
(redoing the splitter section)
To use 6CG7(or 12AU7/12BH7) in the version I posted just use a 10K pot in place of the
20k pot, with the rest of changes I had there and your set. (less gain and better drive)
Than again as Dave suggests if you do still feel you have to much gain the EF86 can be
triode strapped to further reduce the overall circuit gain there.
.
Last edited:
s-petersen
Scott
If later on I switched in a Kt88,or Kt120 which driver would be the best choice.I have 12A*7's and 12BH7's.I'm not sure,I have to look for the 6CG7's
I never thought of them as audio,I don't think I've ever seen them in an amp.Possibly the CJ MV45 I had,or a Marantz? I'm not sure.
If I change the driver I may as well go right to the old Tungsol 6550's I have,As there would,'t be much of the original circuit left.
Would dropping the plate voltage on ehe EF86 as was sugested earlier,lower the gain?
I never thought of them as audio,I don't think I've ever seen them in an amp.Possibly the CJ MV45 I had,or a Marantz? I'm not sure.
If I change the driver I may as well go right to the old Tungsol 6550's I have,As there would,'t be much of the original circuit left.
Would dropping the plate voltage on ehe EF86 as was sugested earlier,lower the gain?
Last edited:
The 12BH7 will do an excellent job! Changing the plate voltage on the EF86 will have only a slight effect on the gain of that stage. The gain is determined much more by the values of the associated components used with that stage.
Dave
Dave
s-petersen
Scott
Ok we'll try the BH7,it's a shame, I don't have any in Mullard and such.12a*7's I have up the kazoo.I'd like to keep the EF-86 if I could,I think that is where the sound I like comes from,call it a feeling,Marantz,Quad,Eico,BT-H all used it
Last edited:
