I fine tune stuff like hum, parasitics and NFB. Hum is often caused by PSU ripple, heater-cathode induction and the ever-mystical, ever-simple, ground loop. Parasitics such as HF (squeal if audible) or LF (motorboating) resonances must be tackled before anything and may be caused by PSU wiring, wiring layout in general, or an unstable (such as a Williamson) or undamped circuit (grid stoppers, zobel networks). NFB in particular magnifies any phase shifts and really gives your troubleshooting a run for your money. I enjoy a challenge. 
Tim
Tim
This is a sadly neglected topic in most tube amp books and probably the thing which causes most grief to newbies.
Rule 1: Get the amp as linear and as hum free etc. before trying to add the feedback loop (assuming it has one).
For amps using global feedback its best to try and suppress any high frequency resonances in the output transformer at this stage - before closing the feedback loop. This is done by adding zobel networks to output transformer connections - usually on the primary side.
The best explanation/method fro doing that can be found here:
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html.
When it comes time to close the feedback loop it may be necessary to adjust the frequency and/or phase response in order for the amp to not oscillate.
The best explanation/methods for this is probably:
TUBE AUDIO DESIGN - The Beginners Guide by Bruce Rozenblit
I hope this is of some help.
Cheers,
Ian
Rule 1: Get the amp as linear and as hum free etc. before trying to add the feedback loop (assuming it has one).
For amps using global feedback its best to try and suppress any high frequency resonances in the output transformer at this stage - before closing the feedback loop. This is done by adding zobel networks to output transformer connections - usually on the primary side.
The best explanation/method fro doing that can be found here:
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html.
When it comes time to close the feedback loop it may be necessary to adjust the frequency and/or phase response in order for the amp to not oscillate.
The best explanation/methods for this is probably:
TUBE AUDIO DESIGN - The Beginners Guide by Bruce Rozenblit
I hope this is of some help.
Cheers,
Ian
Well, except for like I said. Example: you can have an amplifier perfectly flat from 20Hz to 20kHz, but if there's any phase error at the ends, you won't see it until NFB kicks it in the middle anatomy and it gets red and swollen. Williamson topology, and OPTs being used well beyond their useful range, are wonderful things to watch LF rise at 3Hz on, or resonance at 50kHz. Seriously, I had to drop the coupling caps to like 0.022uF, and there was still a 3dB rise at 13Hz! On that one I eventually rewired it to something better...
Tim
Tim
As you rightly point out, misapplied feedback can actually make things worse in an amp that was behaving well prior to its introduction. This is certainly true of the Williamson and any other design that has too many coupling caps in the loop and/or has an OP tranny with deficiencies such as excessive leakage inductance.
Incidentally, I wonder why Mr. Williamson designed a global FB loop in his amp, when he was using triode OP tubes? It seems to be inviting problems unnecessarily, considering the already low distortion and low OP impedance of his triode-strapped KT66s. A bit of local FB would have done the job with no risk of instability. He designed his amp to be built by hobbyists and must have known that many people wouldn't have access to a sufficiently good OPT to make it work properly. However, one can easily provide a simple and effective fix to make global feedback stable in the Williamson, by converting the second coupling cap (before the OP tube) in each half into a step network and rearranging the biasing accordingly.
People always say they will 'fine tune' their amps after building it. Actually how to do so? Any steps or procedure?
Don't know; don't care.
Solid state or hollow state; RF or audio: makes no difference since I always test as I build. It's SSSSSOOOOO much easier to get one stage working right, to apply optimizations, and do all the tweaking before moving onto the next one. At least, you will know where the durn thing is misbehaving! Saves much aggravation as opposed to wiring everything up, then having to go back to track down a misbehaving stage. Also makes it much easier to do whatever final tweaking may be necessary to get full functionality.
Even if it's a proven design, I still test while building. I've caught lots of potential problems that way. While repricating a solid state design of mine, I caught a dead transistor right away when the problem was easy to find, and, more importantly, before it could potentially cause other transistors to fail in later stages.
Don't know; don't care.
Solid state or hollow state; RF or audio: makes no difference since I always test as I build. It's SSSSSOOOOO much easier to get one stage working right, to apply optimizations, and do all the tweaking before moving onto the next one. At least, you will know where the durn thing is misbehaving! Saves much aggravation as opposed to wiring everything up, then having to go back to track down a misbehaving stage. Also makes it much easier to do whatever final tweaking may be necessary to get full functionality.
Even if it's a proven design, I still test while building. I've caught lots of potential problems that way. While repricating a solid state design of mine, I caught a dead transistor right away when the problem was easy to find, and, more importantly, before it could potentially cause other transistors to fail in later stages.
Thanks for the input from you guys.
Agree with Miles that always test when building the amp. What I can do is to make sure all the calcualtion and values are correct when building (supply voltage, value of the cathode resistor.... etc), and that is what I can only learn from the tube amp books or websites. You guys mentioned the problem of phase shift. Can the problem be eliminate when you're in the designing stage? Or is it an inevitable problem?
Agree with Miles that always test when building the amp. What I can do is to make sure all the calcualtion and values are correct when building (supply voltage, value of the cathode resistor.... etc), and that is what I can only learn from the tube amp books or websites. You guys mentioned the problem of phase shift. Can the problem be eliminate when you're in the designing stage? Or is it an inevitable problem?
You guys mentioned the problem of phase shift. Can the problem be eliminate when you're in the designing stage? Or is it an inevitable problem?
Problems such as these can only be treated empirically. You won't actually know if it will be a problem until you have the amp built and running, as it depends on so many factors that you can't know ahead of time. A phase shift problem may or may not show up, as it depends on the characteristics of the audio iron you use, the circuit stray capacitance, the amount of global feedback you apply to solve linearity problems. Yet another reason to test while still constructing the whole amp. That way, you'll know what characteristics your subsystems have, and can best determine where something like grid stoppers, or phase compensation can best be applied to nail any instabilities that do show up.
Problems such as these can only be treated empirically. You won't actually know if it will be a problem until you have the amp built and running, as it depends on so many factors that you can't know ahead of time. A phase shift problem may or may not show up, as it depends on the characteristics of the audio iron you use, the circuit stray capacitance, the amount of global feedback you apply to solve linearity problems. Yet another reason to test while still constructing the whole amp. That way, you'll know what characteristics your subsystems have, and can best determine where something like grid stoppers, or phase compensation can best be applied to nail any instabilities that do show up.
Miles Prower said:
You can estimate phase shifts, rolloffs and estimate NFB all you want but it's all theory until you add the OPT. That hunk of iron is one big cluster-[um, oops sorry adult word belongs here], but we have to live with it. Wiring can bring interesting suprises too, well if you don't draw up and analyze a diagram before hand. I like to do wiring improv.
Tim
Phase shifts can be and should be addressed during the design stage. They form one of your design requirements. In any engineering design if you don't have design requirements you never know if you've succeded or failed. Or worse, if its a software design you never know when its finished (creeping features or as we eng. types call them "feeping creatures").
Phase shift is 99% of the time due to a resistive element and a reactive element (either capacitance or inductance ) in series or in parallel. Most commonly in series. The resitive element is often the input iimpedance of a circuit stage or the output impedance of a circuit stage. If you specify what phase shift you are prepared to tolerate at what frequency you can then specify the approriate input or output impedance goal. If it is a feedback system / amplifier, you can also recognize where the most difficult phase shift problem exists in a circuit and use this to your advantage by making the pole causing that phase shift be your dominant pole.
People seem to keep wanting to slug the frequency response of the input stage to make it the dominant pole - why they don't use the pole associated with the output tube grid capacitance (usually the most difficult to handle) I just can't understand. Well, there are difficulties in a push pull amp making sure the two sides are driven by the same impedance BUT these are easily managed.
A bit of a rave but hopefully there is some useful stuff in there.
Cheers,
Ian
Phase shift is 99% of the time due to a resistive element and a reactive element (either capacitance or inductance ) in series or in parallel. Most commonly in series. The resitive element is often the input iimpedance of a circuit stage or the output impedance of a circuit stage. If you specify what phase shift you are prepared to tolerate at what frequency you can then specify the approriate input or output impedance goal. If it is a feedback system / amplifier, you can also recognize where the most difficult phase shift problem exists in a circuit and use this to your advantage by making the pole causing that phase shift be your dominant pole.
People seem to keep wanting to slug the frequency response of the input stage to make it the dominant pole - why they don't use the pole associated with the output tube grid capacitance (usually the most difficult to handle) I just can't understand. Well, there are difficulties in a push pull amp making sure the two sides are driven by the same impedance BUT these are easily managed.
A bit of a rave but hopefully there is some useful stuff in there.
Cheers,
Ian
Agreed even though it can be difficult to know the exact characteristics of the output transformer at this stage.
Yes, it seems that the dominant pole method is used more frequently in modern tube amps than before, could be influence from SS design. A more effective approach is to use phase corrective networks ala Bode, this make it possible to always keep better open loop bandwidth than with dominant pole wíth the same amount of feedback, also the fedback get more effective at the frequency extremes.
Williamson used phase correction for higher frequencies, (the series connected capacitor and resistor in parallell with the first stage anode resistor) but he didn't add a phase correction for the lower frequency poles which is not so difficult to do but I assume he had trouble to find the relatively high value of capacitors that are needed for this.
Look here for original Williamson without low frequency phase correction http://www.tubetvr.com/Williamson.pdf
And here with low frequency phase correction added http://www.tubetvr.com/Williamson_comp2.pdf
Regards Hans
I don't use feedback so that's not what I tune. What I am interested in tuning is the bias for the output valves - in my case two pairs of PP 6S4A triodes. I have fixed bias for the output tubes, and the supply can be tuned for a start, as to how you filter it. I have CRC at the moment - I'd add a bit more filtering but I don't have a high enough voltage to drop.
But the interesting part is tuning the cathode resistors. I've tried 82 ohms unbypassed on each cathode and the sound is forward and dynamic. With 34 ohms shared between all four cathodes and 13 ohms on each individual cathode the sound is quite different - smoother, more detailed, but less leading edge definition and a rather recessed sound. Haven't yet tried 47 ohms shared between all four cathodes. All these options affect the sound quality quite audibly. I tried 100 ohm with a cathode bypass cap and found that inferior to an unbypassed cathode resistor. Tuning this all by ear could take a while! Andy
But the interesting part is tuning the cathode resistors. I've tried 82 ohms unbypassed on each cathode and the sound is forward and dynamic. With 34 ohms shared between all four cathodes and 13 ohms on each individual cathode the sound is quite different - smoother, more detailed, but less leading edge definition and a rather recessed sound. Haven't yet tried 47 ohms shared between all four cathodes. All these options affect the sound quality quite audibly. I tried 100 ohm with a cathode bypass cap and found that inferior to an unbypassed cathode resistor. Tuning this all by ear could take a while! Andy
Easy.........
Tune with changing to different manufacture tubes called tube rolling. A search can find upper half of noted best quality tubes worthy of auditioning.
Changing low grade coupling caps from say a nasty Sprague orange drop to an AuriCap or more expensive capacitors brings good results.
Tune with changing to different manufacture tubes called tube rolling. A search can find upper half of noted best quality tubes worthy of auditioning.
Changing low grade coupling caps from say a nasty Sprague orange drop to an AuriCap or more expensive capacitors brings good results.
Re: Easy.........
If you believe in spending money for better percieved quality, yes.
Also assumes all the advertised products work as well in your particular amplifier as claimed, which come to think of it is suprisingly lacking, in contrast to the usual attitude of "your amp may vary".
Tim
amperex said:
If you believe in spending money for better percieved quality, yes.
Also assumes all the advertised products work as well in your particular amplifier as claimed, which come to think of it is suprisingly lacking, in contrast to the usual attitude of "your amp may vary".
Tim
Given that we're dealing with one very inexact science here, there are lots of scam artists out there. The $500.00 Volume Control Knob
So, yeah, I am quite suspicious about such claims. Indeed, I believe that the folks at AuriCap are making a big mistake in relying solely on testimonials. I happened to find another web site (independent lab) where they actually put various capacitors to the test. Those "orange drop" types (which I was originally planning on using) did test rather poorly in terms of E/D characteristics and linearity. I recently changed out a 1.0uF electrolytic coupling capacitor in a couple of solid state speaker box amps for 1.0uF polypropylene capacitors (not Auricaps) and this really did improve the sonics. Most noticeable, even though the speakers aren't the best, was the improvement in bass.
Sometimes, I wonder if these speaker amps aren't too good now. I actually had one back on the bench as I thought that there was something wrong with it. It tested out OK, and it turned out that it wasn't the amp, but rather imperfections in the mp3's (and some early CDs) I was listening to.
The Auricap folks are definitely onto something. They're a bit more expensive than the "orange drops" I'd've used otherwise, but certainly not too onerous. I have a good enough reason to believe that Auricaps are worth it, and they are going into the VT amp I'm working on.However, I'm not going to be conned into paying $20.00+ (lots more sometimes) for something like these paper 'n' oil capacitors in teak wood tubes. I'm sure they'd look real pretty framed and hanging on the wall, but who's going to notice when they're mounted under a chassis? Or that outfit that even had the chutzpah to advertise that their beeswax capacitors were made "with all natural ingrediants"
Is this health food or electronic components they're selling ? 
can anyone please elaborate why a Williamson is unstable?
This is a four stage toplogy which uses RC coupling between every stage. This excessive use of RC coupling introduces phase shifts that, when combined with inverse feedback, promotes instability since a 360 deg phase shift around the loop is highly likely to occur at a frequency where the closed loop gain is still greater than unity. That gives you an oscillator.
Best to use less open loop gain, and try for using less global feedback.
Here is a preliminary design I'm working on right now. This differs from a Williamson by using just one capacitor coupled stage, as the finals are direct coupled. It eliminates two stages: the voltage amp, and the cathodyne. The 6SL7 LTP is the only gain block inside the feedback loop in that design. Since both the 6SL7 and 6SN7 are quite linear, what I'm shooting for is requiring just enough feedback to correct for the audio iron. In any case, I'm not planning on using more than 20db(v).
Any additional front end gain can come from a preamp module that would also provide the sonic controls.
This is a four stage toplogy which uses RC coupling between every stage. This excessive use of RC coupling introduces phase shifts that, when combined with inverse feedback, promotes instability since a 360 deg phase shift around the loop is highly likely to occur at a frequency where the closed loop gain is still greater than unity. That gives you an oscillator.
Best to use less open loop gain, and try for using less global feedback.
Here is a preliminary design I'm working on right now. This differs from a Williamson by using just one capacitor coupled stage, as the finals are direct coupled. It eliminates two stages: the voltage amp, and the cathodyne. The 6SL7 LTP is the only gain block inside the feedback loop in that design. Since both the 6SL7 and 6SN7 are quite linear, what I'm shooting for is requiring just enough feedback to correct for the audio iron. In any case, I'm not planning on using more than 20db(v).
Any additional front end gain can come from a preamp module that would also provide the sonic controls.
The real issue with the Williamson is not so much the number of rolloffs, but their ratios. To stabilize the circuit, the dominant rolloffs have to be spaced more widely from the remainder. If you can get a copy of Crowhurst's "Understanding Hifi Circuits," there's a very handy set of charts showing what ratios between rolloffs are necessary to achieve stability for varying numbers of rollofffs and varying feedback factors.
The circuit you show does not have enough gain to get any meaningful feedback, and certainly not 20dB. You'll find that it either needs to be run open-loop or that you'll have to insert another gain stage. Direct coupling to the output tubes is very useful, not so much for overall stability but rather to improve overload recovery.
The circuit you show does not have enough gain to get any meaningful feedback, and certainly not 20dB. You'll find that it either needs to be run open-loop or that you'll have to insert another gain stage. Direct coupling to the output tubes is very useful, not so much for overall stability but rather to improve overload recovery.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.