OK, so here's a question I couldn't find anywhere on the internet.
I am using a modded ST-70, but I doubt the details of the line stage are all too critical.
The 390pF feedback capacitor connected from the screen tap to the cathode of the input stage somewhat puzzles me.
My understanding of the feedback is as follows: the resistor connected from the 16 ohm tap to cathode is (in an ideal world) frequency-independent negative feedback. This is the primary gain controlling mechanism.
The capacitor is used to decrease the overall gain at high frequencies. The cap still gets a complete (i.e. not only the pull, but the full push-pull) ac signal at the screen tap due to autotransformer action. I assume the tap is taken from the input side of the OPT due to the bandwidth limitation of the OPT, so a more accurate high frequency reference is available at this point.
Feel free to chime in at this point and correct my above assumptions.
What I fail to understand is why limit bandwidth here when the OPT will do the same? Further, most P-P designs I see apply a parallel RC feedback network taken off the output of the OPT.
So is a better method to move the C from the UL tap to the output of the OPT, in parallel with the existing R? My input stage is a cascoded triode, not a pentode.
Once the purpose and location is determined, just how critical is the value of the cap? Is this the value determined by subjective listening, calculated stability equations, or square wave injection?
Thanks for clearing this esoteric issue up.
I am using a modded ST-70, but I doubt the details of the line stage are all too critical.
The 390pF feedback capacitor connected from the screen tap to the cathode of the input stage somewhat puzzles me.
My understanding of the feedback is as follows: the resistor connected from the 16 ohm tap to cathode is (in an ideal world) frequency-independent negative feedback. This is the primary gain controlling mechanism.
The capacitor is used to decrease the overall gain at high frequencies. The cap still gets a complete (i.e. not only the pull, but the full push-pull) ac signal at the screen tap due to autotransformer action. I assume the tap is taken from the input side of the OPT due to the bandwidth limitation of the OPT, so a more accurate high frequency reference is available at this point.
Feel free to chime in at this point and correct my above assumptions.
What I fail to understand is why limit bandwidth here when the OPT will do the same? Further, most P-P designs I see apply a parallel RC feedback network taken off the output of the OPT.
So is a better method to move the C from the UL tap to the output of the OPT, in parallel with the existing R? My input stage is a cascoded triode, not a pentode.
Once the purpose and location is determined, just how critical is the value of the cap? Is this the value determined by subjective listening, calculated stability equations, or square wave injection?
Thanks for clearing this esoteric issue up.
While I'm not a degreed engineer of transformer design, I think I can safely say that this capacitor is there to compensate for anomalies that appear during square wave (pulse) testing of the amplifier. In particular, overshoot of the leading edge and perhaps any ringing just past the corner. I'm almost certain it isn't done by listening or calculation.
This small capacitor appears in some of Dynaco's other amplifiers as well. Because of this, I'm leaning toward a characteristic function of their transformers. And since all facets of transformer design are a compromise, this is how they tweeked it.
Curiously enough, Acrosound transformers (same time period) never used this cap. Personally, I've always prefered Acrosound over Dynaco anyway.
Victor
This small capacitor appears in some of Dynaco's other amplifiers as well. Because of this, I'm leaning toward a characteristic function of their transformers. And since all facets of transformer design are a compromise, this is how they tweeked it.
Curiously enough, Acrosound transformers (same time period) never used this cap. Personally, I've always prefered Acrosound over Dynaco anyway.
Victor
So if the idea is to inject a square wave and tweak, what is the final product we are looking for? Minimal overshoot on the rising edge? Various square wave frequencies between 100 and 10k?
Has anyone attempted placing the cap in parallel with the feedback resistor, as most UL designs seem to do, or is this the best way with the A470 trannies?
Thx.
Has anyone attempted placing the cap in parallel with the feedback resistor, as most UL designs seem to do, or is this the best way with the A470 trannies?
Thx.
I would say Hollowstate is on target...
ALthough it is possible to calculate the compensation, it is "quicker" to just do it with the scope and square wave...
With loop compensation, you can have several loops, depending on what and where you need to correct... YOu have to correct with POLES and ZEROS...
The reason for the ST-70 feedback arrangement is for convienience of using least amount of parts, designer was smart.. ALso to put the least amount of DC offset on the secondary of the Output Transformer..
The first stage tube has a divider in the cathode.... The divider is to control the amount of feedback, but also the lower you go on the divider the smaller the DC voltage...This undesirable DC voltage would offset the speaker, since it is on the secondary side... So they provided a ton of AC signal voltage from the screen tap....Then divided it down with the 47/620 ohm resistors to get it where they wanted it, now the DC offset voltage is tiny, but the AC signal for feedback is where they wan it... To calculate the time constant..you would need to know All the resistances of the circuit, such as screen resistance of tube, impedance looking into the screen tap, cathode resistance looking up into 7199..ect..ect.. The cap is strictly for DC blocking..it overlaps the feedback resistor from the secondary side of the OPT....since the 1K is taking over for low band...calculated for the same dB of feedback at the cross-over ... SO technically, the R and C for feedback are in parallel but not normalized..
Keep in mind the bandwidth on the primary is the same as the seconday for the OPT... The resistor on secondary is also used to keep the size of the cap small...
Chris
ALthough it is possible to calculate the compensation, it is "quicker" to just do it with the scope and square wave...
With loop compensation, you can have several loops, depending on what and where you need to correct... YOu have to correct with POLES and ZEROS...
The reason for the ST-70 feedback arrangement is for convienience of using least amount of parts, designer was smart.. ALso to put the least amount of DC offset on the secondary of the Output Transformer..
The first stage tube has a divider in the cathode.... The divider is to control the amount of feedback, but also the lower you go on the divider the smaller the DC voltage...This undesirable DC voltage would offset the speaker, since it is on the secondary side... So they provided a ton of AC signal voltage from the screen tap....Then divided it down with the 47/620 ohm resistors to get it where they wanted it, now the DC offset voltage is tiny, but the AC signal for feedback is where they wan it... To calculate the time constant..you would need to know All the resistances of the circuit, such as screen resistance of tube, impedance looking into the screen tap, cathode resistance looking up into 7199..ect..ect.. The cap is strictly for DC blocking..it overlaps the feedback resistor from the secondary side of the OPT....since the 1K is taking over for low band...calculated for the same dB of feedback at the cross-over ... SO technically, the R and C for feedback are in parallel but not normalized..
Keep in mind the bandwidth on the primary is the same as the seconday for the OPT... The resistor on secondary is also used to keep the size of the cap small...
Chris
Essentially yes. Basic audio square wave testing is generally done at 2KHz. Remember, square waves contain many harmonics, and these higher harmonics will let you extraoplate up to 20KHz. Of course you can go higher to see what happens for your own edification.
Simple, the best looking square wave (at the output) that mimics the input as close as possible with the fewest aberrations.
If you really want to know more and have a decent square wave generator and a good scope, terminate the transformer secondary with a non-inductive load, feed in a signal and scope the output. Then disconnect and reconnect the cap to see what it's doing.
I suppose I should say that you should only do this if you are experienced and feel confidante working with a circuit under power and the high voltages involved. Otherwise don't.
I'd also give the manufacturer a little credit for knowing what they were doing when they placed that capacitor there.
Victor
HollowState said:
Actually, in my world this is low power and low voltage. I work in the power industry, where transformers are sized in MVA and voltages are listed in kV (i.e. 345 kV is high voltage, 15 kV is medium voltage, 480V is low voltage). So tinkering with 400V stuffed in 100uF isn't too scary (does hurt though). The transformers I work on are about 15 feet tall, and most capacitors only used for power factor correction.
That partially explains why feedback is not my specialty. Additionally, I was mostly drooling during Feedback Control Systems in college, as the theory was so beyond practical common sense and application.
I've got the scope and function generator, so I'll give it a try with the square wave. Will be interesting to see the differences, and I'll probably look at what the result is with the cap moved to the low-z side of OPT, just for kicks.
SY said:
And just when I thought I had a handle on things, SY jumps in and adds a head-scratcher.
Asymmetry? as in one half of the primary winding behaves a little differently than the other half? I assume that would necessitate the feedback cap be attached to the correct half?
Ahhhh, now it gets good.
SY knows what he speaks of. Dude, you spend way too much time with tubes !!
I made a copy of what was posted in a prior forum:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=88403
I took resistance readings from the center tap to the two green and two blue leads. My results are as follows (in order of blue, green, blue/white, green/white).
OPT1: 90, 30, 103, 34
OPT2: 160, 30, 102, 34
OPT3: 90, 30, 105, 35
OPT4: 163, 30, 103, 34
So, regarless of which OPT's I use (I will be selecting the 90's, not the 160's) I can see the asymmetry.
The ST70 manual specifies to use the green (30 ohm) tap for feedback, which is apparently the inside winding. Almost appears to serve more as an AC balance than a HF compensation.
Very nice. Thanks EVERYONE for the advice !!
SY knows what he speaks of. Dude, you spend way too much time with tubes !!
I made a copy of what was posted in a prior forum:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=88403
I took resistance readings from the center tap to the two green and two blue leads. My results are as follows (in order of blue, green, blue/white, green/white).
OPT1: 90, 30, 103, 34
OPT2: 160, 30, 102, 34
OPT3: 90, 30, 105, 35
OPT4: 163, 30, 103, 34
So, regarless of which OPT's I use (I will be selecting the 90's, not the 160's) I can see the asymmetry.
The ST70 manual specifies to use the green (30 ohm) tap for feedback, which is apparently the inside winding. Almost appears to serve more as an AC balance than a HF compensation.
Very nice. Thanks EVERYONE for the advice !!
Ok zigzag, now I won't worry about you. I work with HV too, but nothing like you. I keep it under 20KV for repairing power supplies and testing transmitting tubes.
Yes true. But a well designed transformer (not Dynacos) would use progressively larger gauge wire as the coil build increases. But this can only be carried so far as it poses it's own problems. Uneven DC resistance is the least of an output's troubles. Uneven distributed capacitance and leakage inductance comes to mind as a bigger bugaboo.
Victor
Yes true. But a well designed transformer (not Dynacos) would use progressively larger gauge wire as the coil build increases. But this can only be carried so far as it poses it's own problems. Uneven DC resistance is the least of an output's troubles. Uneven distributed capacitance and leakage inductance comes to mind as a bigger bugaboo.
Victor
I used DCR as an indicator of the asymmetry, not as the problem itself. If you think about a long single winding, the outer layer has more leakage inductance and higher DCR because it's farther from the core and the radius of the turns is longer.
To achieve better balance, the winding technique has to be different- sectioning is the key. But that's expensive, so very few "classic" amps used those techniques. Thus, you see tricks like Dynaco's or (more often) odd caps and Zobels tied across half the primary as in the Eico HF-87.
To achieve better balance, the winding technique has to be different- sectioning is the key. But that's expensive, so very few "classic" amps used those techniques. Thus, you see tricks like Dynaco's or (more often) odd caps and Zobels tied across half the primary as in the Eico HF-87.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.