|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
68.103.9.72
In Reply to: Local Loop feeback posted by maxhifi on April 3, 2007 at 12:55:12:
Probably more to do with the # of stages involved than the OPT per se.
Follow Ups:
The frequency response limitations and hence the consequent phase shifts are vastly larger in any OPT than in a direct tube to tube local loop, or even a couple of tubes in a not so local loop.The other big importance of taking the NFB off the anodes of the OP tubes rather than the seconadary of the OPT is that it reduces the driving impedance as seen by the OPT - hence reducing the transformer's own inherent distortions.
The RDH states: A transformer's distortion is zero if the driving impedance is also zero...
A surprising result. But probably true. With a zero driving impedance the driving voltage is maintained no matter what the secondary load is. And when the driving impedance is zero, you can remove the core entirely and the performance will be the same.The primary inductance doesn't have to be much at all to prevent a complete short at low frequencies and so the core is not doing anything useful to a zero driving impedance. All the core ever does is support low frequencies in transferring the energy to the secondary by raising the inductance for the low frequencies for the case when the driving impedance is higher than zero.
Since the core has no effect, it is acting like an air core transformer all the way. And there is no distortion in air core mutual inductance.
As well as the tube amp hobby I design servo amps for stabilsed platforms in aircraft and laser scanners in my day job.In any feedback system there is a very usefull "Rule of Thumb" for best performance.
For stability (without special compensating techniques) the loop gain can only be as high as the ratio of the two most dominant poles. In a tube amp with global feedback one (well 2 actually) poles are set by the Output Transformer leakage inductance and interwinding capacitances. You then want to introduce a dominant pole = to that pole divided by the loop gain (say 20 for a typical example). That means for common garden variety Output Transformers you would be wanting to introduce a pole at 1.5 to 2 KHz. Imagine what that would sound like !!!!.So unless using a Toroidal Output Tranny with bandwidth to 250kHz, whenever you apply global feedback, you Will always be fiddling with special compensating techniques. These techniques (from a first principals point of view) seek to modify (reduce)the high frequency gain (amplitude) whilst attempting to leave the phase response intact. This basic point is largely missed or not understood and 98% of "stabilising circuits" I've seen are done poorly. It is also true that to a certain degree (not completely) reducing high frequency gain whilst leaving phase response intact are mutually exclusive requirements.
Do the math - that zobel network across the anode load resistor in the first stage you see as a typical stabilising technique should have a resistor value of 1/10th the anode load - how many times have I seen a value of 1/2 or 1/3 used - a sure sign that someone doesn't know what they are doing and are proceeding by trial and error - mostly the latter.
Aside: HF response will be nearly 6dB down when Zobel Xc + R = the load resistor BUT the 45 degrees phase shift (usually associated with the -3dB point in a simple low pass) will be when Zobel Xc = the Zobel R
Thats why you want the Zobel R = to 1/10th the load resistor (certainly never more than 1/5th)- to maximise the difference in frequency between the amplitude -3dB point and the 45 degrees phase shift point.Having said all the (sorry - I pushed my own button)I'm stongly with Allen on this - any way to avoid global feedback and hence avoid inherently compromised stabilsing techniques should be embraced. As he says reducing the effective rp of the output stage itself helps at both ends of the frequency spectrum - it lowers the low frequency roll off (6dB/octave) associated with rp and Lp (OT Primary Inductance) and raises the High Frequency roll offs associated with rp and leakage inductance and interwinding capacitances (12dB/octave total) Hence I use triode mode or ultralinear with balanced shunt feedback from the output tube anodes. If you still have to introduce global feedback to control Zout then the rp reduction will have made this simpler by extending the frequencies at which problem phase shifts occur in the output tranny.
I know he is not an Ultralinear fan. From the above you can see why that might be. Some OT phase shifts are involved in the feedback to the screens. The same arguement would apply to cathode feedback schemes.
From the above, in a global feedback system, there are a couple of other "facts" that fall out.
In the forward path all other poles MUST be above the OT introduced pole by at least a factor of 3, preferably 5 to 10. MANY driver stages fall down at this point. The theory says that it doesn't matter where in the forward path the dominant pole is introduced and many an amp has been designed an built with "accidental" stability due to inadequate driver stage. Always check your amp open loop to find out where the most dominany pole is located and if required slug that stage harder whilst make sure ever other stage is significantly faster - if that turns out to be the driver stage, that is perfectly acceptable but do it by design rather than by accident.The feedback path MUST be faster than the forward path and here large values of feedback resistor interacting with Miller capacitance at tube grid can be a problem so most feedback is to a low impedance point - usually a cathode so that a low value of feedback resitor can be used and the capacitance at the feedback point will not cause a low pass filter at the frequencies of interest.
You can see why avoiding global feedback simplifies things enormously and reduces the possible number of problems.
Remember Kiss.
Simple circuits have simple problems
Simple problems have simple solutionsAs one of my software lecturers used to point out:
Its 3 times a difficult to debug your software as it is to write it.
If you are as clever as you possibly can be when you write it - when do you reckon you'll have it running by - answer NEVER.
> The RDH states: A transformer's distortion is zero if the driving impedance is also zero...I assume the context of this statememt discounts non-linearities of the core material (?)...
...but it was pretty much an ultimate statement.
This post is made possible by the generous support of people like you and our sponsors: