I have pairs of each of Plitron VDV2100 and VDV2100-CFB/H output transformers sitting on the shelf waiting for me to get around to them (plus the matching Power Trannies).
The VDV2100 is 2000:5 Ohm with 40% UL taps
The VDV2100 CFB/H is the same BUT it has 2 separate 10% cathode feedback windings and the UL taps are at 30%. Thats because the 10% cathode feedback effectively makes the UL 40% (i.e. 10 + 30 = 40).
The VDV2100 is destined have triode strapped output pairs BUT I am going to try Mennos "Super Triode" (Ultralinear plus cathode feedback) with the VDV2100 CFB/H.
A hint for those ordering stuff from Plitron. Their quantity price break is 5 units. I needed 4 Power trannies to go with those output trannies BUT was able to order 6 for $15 more than 4 would have cost (excluding shipping charges).
Cheers,
Ian
P.S. I've just finished a new fixed biased 6V6 "Baby Huey" style Ultralinear Amp with Tube Rectifier and all Polypropylene HV caps. Hammond 1650E Output Trannies. It matches the performance of my 845 SET. Will post some pics and schematic (as soon as I've drawn it) to a new thread.
The VDV2100 is 2000:5 Ohm with 40% UL taps
The VDV2100 CFB/H is the same BUT it has 2 separate 10% cathode feedback windings and the UL taps are at 30%. Thats because the 10% cathode feedback effectively makes the UL 40% (i.e. 10 + 30 = 40).
The VDV2100 is destined have triode strapped output pairs BUT I am going to try Mennos "Super Triode" (Ultralinear plus cathode feedback) with the VDV2100 CFB/H.
A hint for those ordering stuff from Plitron. Their quantity price break is 5 units. I needed 4 Power trannies to go with those output trannies BUT was able to order 6 for $15 more than 4 would have cost (excluding shipping charges).
Cheers,
Ian
P.S. I've just finished a new fixed biased 6V6 "Baby Huey" style Ultralinear Amp with Tube Rectifier and all Polypropylene HV caps. Hammond 1650E Output Trannies. It matches the performance of my 845 SET. Will post some pics and schematic (as soon as I've drawn it) to a new thread.
Hi Douglas. I understand that part, which is the UL-like element of CF, but what about the difference between the two in regards to effect on the G1-k voltage? By acting directly on the screen UL has none, CF modulates Vg1-k directly. That seems too big a difference to equate CF with UL. Then again, it was a long day at work.
rdf, you have it right. It is the g2-k voltage modulation that gets it the 'like' badge. The g1-k is a part which cannot be taken by itself unless you operate as McIntosh did/does with their topology.
cheers,
Douglas
cheers,
Douglas
Bandersnatch said:
Checkin on me, are ye mate?
Which is just as well, I did make an error for which my serious apology. 😱 (Not calculation only; measurements; the thing is working.)
I was inadvertently thinking of the high peak on the driver anode (scope reading), which is about 380V. But the anode does of course not go to 0!
The driver anode swings between 64V and 380V (rounded), i.e. Vpp fed is 316V. This is reached by taking 25% of the 6L6 Va.p of 450V, being 113Vp, which is the cathode (winding) signal. Add 45V for input signal (the 6L6 cathode bias is about 45V), giving 158Vp required G1 signal, or 316Vpp. As said, I am considering making the "tap" at 20% in the future.
Not to bore with analysis, but the next question is probably why cathode bias? It is in fact fixed bias on the cathodes, by way of a regulated current sink (constant Vc). This because I am doing other regulating on G1s, which requires that they be common-referenced.
Thanks for pointing this out.
Johan Potgieter said:
Why 20%?
100% and you've set (in the Solid State forum one lady described her amp with an OPT squeezed between couple of source followers)
rdf said:
Hi Rdf,
Already answered, but again apologies for being a little vague. I did in fact mean the output circuit only, not full equivalency. That is, if you look at output between cathode and anode, the screen is at a tap - which is now where you have inserted the power supply instead of at the cathode end of the OPT (equivalent diagram). The tube then "sees" the same output conditions as per ordinary UL. The cathode-G1 feedback is an extra bonus, separate from the above.
Bandersnatch,
My tertiary winding is not a single layer, but two. Primary arrangement is then 2-S-4-S-4-S-4-S-2 (S=secondary), to give the topology for lowest leakage and a 25% "tap". For 6L6 there is not a defined optimum at 43%; it is similar to the KT88 data previously shown on this thread (different values). I sacrifice some increase in rp and perhaps D, seen from a UL point.
Johan Potgieter said:
Thx Johan, I didn't fully catch the drift and 'dun learnt sumptin' today. Pentode with cathode feedback sound now like an oddly appealling form of 'super UL'.
Anyone using UL p-p 6550C's or KT88's tried increasing the screen grid resistors from published values typ from 100 to 330 ohms and noticing what happens to the thd ?
Keep the exisiting a-g2 Zobel (if fitted) but I find increasing screen grid res to 470 ohms per tube signifigantly reduces thd. This may fall into line with remarks applicable to EL34's (Mullard 20W sch using 1K) but by doing this I'm now getting good performance using 6550's. It seems less current through the screen je better the results. With beam tetrodes can anyone explain why this is so ?
richj
Keep the exisiting a-g2 Zobel (if fitted) but I find increasing screen grid res to 470 ohms per tube signifigantly reduces thd. This may fall into line with remarks applicable to EL34's (Mullard 20W sch using 1K) but by doing this I'm now getting good performance using 6550's. It seems less current through the screen je better the results. With beam tetrodes can anyone explain why this is so ?
richj
Hi Richj,
What I can explain is that it is not simply G2 current (as in dc). It was mentioned before that one must keep in mind that G2 is not simply a grid associated with high imput impedance as one may have become used to classicaly. As it draws current it behaves like a second anode, thus having a (dynamic) internal resistance like rp. This can vary substantially with beam tubes - I did once long ago plot this, but notes be no more. (It could be interesting doing it again - I am tempted.)
That would represent a different picture signal-wise than just thd vs. I(G2). Did you measure over the full output range? You would have noticed that distortion for the EL34 did not increase linearly with output with a screen resistor (although reduced compared to fixed V.G2).
I cannot at present measure D
, but hope to find time to again plot "r.g2" vs voltage, g2-series resistor, etc. Being able to measure D, you would be in a better position to do all. The "r.g2" would be a function of tube current, screen tap, series resistance (R.g2) and even signal amplitude - quite a handful. Presuming one is interested mainly in D, it would be sufficient to plot only R.g2/D for convenience.
Regards
What I can explain is that it is not simply G2 current (as in dc). It was mentioned before that one must keep in mind that G2 is not simply a grid associated with high imput impedance as one may have become used to classicaly. As it draws current it behaves like a second anode, thus having a (dynamic) internal resistance like rp. This can vary substantially with beam tubes - I did once long ago plot this, but notes be no more. (It could be interesting doing it again - I am tempted.)
That would represent a different picture signal-wise than just thd vs. I(G2). Did you measure over the full output range? You would have noticed that distortion for the EL34 did not increase linearly with output with a screen resistor (although reduced compared to fixed V.G2).
I cannot at present measure D
, but hope to find time to again plot "r.g2" vs voltage, g2-series resistor, etc. Being able to measure D, you would be in a better position to do all. The "r.g2" would be a function of tube current, screen tap, series resistance (R.g2) and even signal amplitude - quite a handful. Presuming one is interested mainly in D, it would be sufficient to plot only R.g2/D for convenience.Regards
Many thanks, Rdf!
Memory! (on my part, or poor filing).
As said before, may factors here. With large resistors one would expect max. output to fall. Your analysis does not indicate a spectacular difference, though I was worried by the increase in high order harmonic distortion. It also might be that there is an optimum somewhere, since some things would be increasing while others would go the opposite way. The EL34 data does show a meaningful change, but for one specific conditon, though one might hope that was the optimum. And as said, I expect a noticable difference between a pentode and beam tube.
Now I really long for the use of a spectrum analyser!
Regards
Memory! (on my part, or poor filing).
As said before, may factors here. With large resistors one would expect max. output to fall. Your analysis does not indicate a spectacular difference, though I was worried by the increase in high order harmonic distortion. It also might be that there is an optimum somewhere, since some things would be increasing while others would go the opposite way. The EL34 data does show a meaningful change, but for one specific conditon, though one might hope that was the optimum. And as said, I expect a noticable difference between a pentode and beam tube.
Now I really long for the use of a spectrum analyser!
Regards
Reading through this thread most of the comments have not really been about why people dislike UL but about how to improve UL, Looking at many of the O/P transformers that I have seen on offer most have the UL tap.
Probably be well shot down but how about this-
Stabilised g2 supply and resistive feed to g2, g2 is then coupled to the UL tap with a capacitor. this has the advantage that a separate UL winding is not required and the g2 voltage can be set as desired.
Boiss
Probably be well shot down but how about this-
Stabilised g2 supply and resistive feed to g2, g2 is then coupled to the UL tap with a capacitor. this has the advantage that a separate UL winding is not required and the g2 voltage can be set as desired.
Boiss
Boiis,
No fine - but stabilized, and then to g2 through a resistor big enough so that a capacitor of practical value could be used (the impedance of all also big enough not to significantly load the OPT) might be a problem. One intuitively feels that with the lowest value of resistance thus required, g2 voltage will no longer be stabilised anyway!
Other schemes have been suggested, some very laudible ones using mosfets. Separate screen supplies become mandatory at high output power only. It would appear that up to 100W output this is still not necessary.
Regards
___________________________________________
I like your preference for good food! Count me in.
No fine - but stabilized, and then to g2 through a resistor big enough so that a capacitor of practical value could be used (the impedance of all also big enough not to significantly load the OPT) might be a problem. One intuitively feels that with the lowest value of resistance thus required, g2 voltage will no longer be stabilised anyway!
Other schemes have been suggested, some very laudible ones using mosfets. Separate screen supplies become mandatory at high output power only. It would appear that up to 100W output this is still not necessary.
Regards
___________________________________________
I like your preference for good food! Count me in.
Some checks on two AB tube amps rated at 100W each with different o/p trannies both 43% taps and using ident tubes. 6550C were also tried with no thd improvement. Using KT90 represents a thd disaster as these are NOT optimised for audio work so thd was x10 higher.
The circuits were sim.
The value of the g2 resistor selected for lowest thd varied more with o/p transformer makes, but selecting diferent tubes specifically designed for UL operation showed hardly any variation.
Measuring the leakage inductance/capacitance between two o/p transformer vendors, the one with a 14 section winding had a higher leakage capacitance and didn't show any thd variation when the g2 resistor was varied from 220R - 1K and gave a consistent 0.25% thd at 1Khz 75W whatever res value.This o/p transformer manufacturer claims -3dB at 60Khz.
The other was a high performance 18 section configuration which measured 50% lower leakage inductance & capacitance values wound to Williamson spec and thd was same as above with nominal values but lower thd when g2 = 1K instead of 220R. This o/p transformer manufacturer claims -3dB at 40Khz. This 18 section design with 1K g2 res dropped thd to 0.05% at 75W. This is an excellent result but this shows how confusing it is without a thd analyser. Infact impossible for most.
....The 18 sect looked better on paper but required higher 1st stage zobel values to avoid a self resonance ring at 80Khz. The 14 sect gave a better square wave performance and seemed better behaved through to 50Khz and resp faded therafter. Both transformers are modern good quality winding designs with balanced windings, but this shows the variation between makes
that gave identical Pout.
This does imply that a designer can "overkill" a sectionalised design...where fewer sections can create a better performance. I take my hat off to the designer of the 18 sect design as he used a slide rule and the 14 sect design was computer optimised. Both makers know exactly what is involved..
The difficult bit is understanding the relationship with the tap leakage parasitics at the screen grid in UL and if anyone knows any literature on this I would gladly know.
richj
The circuits were sim.
The value of the g2 resistor selected for lowest thd varied more with o/p transformer makes, but selecting diferent tubes specifically designed for UL operation showed hardly any variation.
Measuring the leakage inductance/capacitance between two o/p transformer vendors, the one with a 14 section winding had a higher leakage capacitance and didn't show any thd variation when the g2 resistor was varied from 220R - 1K and gave a consistent 0.25% thd at 1Khz 75W whatever res value.This o/p transformer manufacturer claims -3dB at 60Khz.
The other was a high performance 18 section configuration which measured 50% lower leakage inductance & capacitance values wound to Williamson spec and thd was same as above with nominal values but lower thd when g2 = 1K instead of 220R. This o/p transformer manufacturer claims -3dB at 40Khz. This 18 section design with 1K g2 res dropped thd to 0.05% at 75W. This is an excellent result but this shows how confusing it is without a thd analyser. Infact impossible for most.
....The 18 sect looked better on paper but required higher 1st stage zobel values to avoid a self resonance ring at 80Khz. The 14 sect gave a better square wave performance and seemed better behaved through to 50Khz and resp faded therafter. Both transformers are modern good quality winding designs with balanced windings, but this shows the variation between makes
that gave identical Pout.
This does imply that a designer can "overkill" a sectionalised design...where fewer sections can create a better performance. I take my hat off to the designer of the 18 sect design as he used a slide rule and the 14 sect design was computer optimised. Both makers know exactly what is involved..
The difficult bit is understanding the relationship with the tap leakage parasitics at the screen grid in UL and if anyone knows any literature on this I would gladly know.
richj
Maybe there were other factors affecting performance, besides the number of sections - such as the core materials and dimensions?
Thanks very much, Richj.
Yes, this is the thing. I must say I also pretty much work from experience and a few basic measurements - must really explore avenues to get a decent spec. analyser .... oh, well
(no more rich aunts left, never had any).
I did remark earlier that one must be careful with all those sections. One other thing is the thickness of insulation between sections. It is perhaps known that the thicker, the lower the internal capacitance, but higher the leakage L. The thing is also however that with thin insolation the tolerance increases, as the percentage dimension variation between transformers becomes very dependant on wire tension, proximity of adjacent section layers and ingress of lacquer, etc. This is where layer-winding is superior despite the extra labour cost. The dimensions also have an influence.
[Ah! There is something to be said yet for the experience of a home-made transformer, never mind how long it took! Those were the days.]
No, Richj, I also find that the tap leakage parasitics are mentioned only and a C-R between anode and screen is advised. Perhaps fortunately thus far, the problems I encountered where I needed correction were all in the region of 50 - 100KHz, so it did not bother me greatly to have to add some C-R there. I know that some folks encourage sectionalising of these (windings) as well (was that perhaps the reason for all those 18 sections?). But in the end it could run up the transformer complexity (and cost) to an unacceptable point. Thus as said I do not mind doing a slight C-R phase correction rather.
Simulating would be nice, but how do you simulate such an output transformer sufficiently accurately? Not by any simplified equivalent circuit, I feel. I measure the product itself, draw phase diagrams etc. and get some satisfaction from that. But it would be nice to have better instruments, as I had at the lab where I worked (HP, Marconi, Rhode & Schwartz....).
Regards.
Yes, this is the thing. I must say I also pretty much work from experience and a few basic measurements - must really explore avenues to get a decent spec. analyser .... oh, well
(no more rich aunts left, never had any).I did remark earlier that one must be careful with all those sections. One other thing is the thickness of insulation between sections. It is perhaps known that the thicker, the lower the internal capacitance, but higher the leakage L. The thing is also however that with thin insolation the tolerance increases, as the percentage dimension variation between transformers becomes very dependant on wire tension, proximity of adjacent section layers and ingress of lacquer, etc. This is where layer-winding is superior despite the extra labour cost. The dimensions also have an influence.
[Ah! There is something to be said yet for the experience of a home-made transformer, never mind how long it took! Those were the days.]
No, Richj, I also find that the tap leakage parasitics are mentioned only and a C-R between anode and screen is advised. Perhaps fortunately thus far, the problems I encountered where I needed correction were all in the region of 50 - 100KHz, so it did not bother me greatly to have to add some C-R there. I know that some folks encourage sectionalising of these (windings) as well (was that perhaps the reason for all those 18 sections?). But in the end it could run up the transformer complexity (and cost) to an unacceptable point. Thus as said I do not mind doing a slight C-R phase correction rather.
Simulating would be nice, but how do you simulate such an output transformer sufficiently accurately? Not by any simplified equivalent circuit, I feel. I measure the product itself, draw phase diagrams etc. and get some satisfaction from that. But it would be nice to have better instruments, as I had at the lab where I worked (HP, Marconi, Rhode & Schwartz....).
Regards.
In general I find the quoted upper b/w of an o/p tranny claimed by a maker i.e -3/-6dB at 60Khz is fine but when one gets down in the chassis, the usable often ends up more like 30Khz. That is one should be able to get a good looking 10Khz square wave peformance without skewing or overshoot at full power. This can be a tough challenge.
Strictly using Fourier analysis implies a 15Khz square wave requires a b/w of 45Khz i.e droop not greater than -6dB at this frequency. Fortunately there is no music content at this frequency.
As I see it the iron permeability and silicon hasn't changed from the good old days.... The core dimensions have a massive influence on performance.....greater copper CSA and longer windings add to leakage capacitance, even though this can be "ironed out". To be fair I get a better performance from a set of o/p trannies designed with cut off at 20Hz than 15Hz with a kg extra iron. With smaller cores the reduction of leakage parasitics can be beneficial although this conflicts with the optimum core requirements for a specific power throughput (in Radiotron H/book).
Johan# you rightly mentioned it's all down to cost and the law of diminishing returns. One can pay a massive sum for a quality lump and someone can say "get same with tranny half the size".
The way forward is with computer simulation, if one can get "ideal performance" with a lower section than going to the trouble of a high sectioned design then so the better. Unfortunately the real world often ends up with larger o/p trannies being a law to themselves and often smaller is better.
That 18 sectioned design I mentioned previously was interleaved with paper (ideal dielectric). Another point is with larger o/p trannies I always have to fit an R/C snubber between the anodes.
Another point that many find is different square wave quality on varying combination taps of o/p transformers i.e 4/8/15 Ohm secs. That can be dealt in another mail.
richj
Strictly using Fourier analysis implies a 15Khz square wave requires a b/w of 45Khz i.e droop not greater than -6dB at this frequency. Fortunately there is no music content at this frequency.
As I see it the iron permeability and silicon hasn't changed from the good old days.... The core dimensions have a massive influence on performance.....greater copper CSA and longer windings add to leakage capacitance, even though this can be "ironed out". To be fair I get a better performance from a set of o/p trannies designed with cut off at 20Hz than 15Hz with a kg extra iron. With smaller cores the reduction of leakage parasitics can be beneficial although this conflicts with the optimum core requirements for a specific power throughput (in Radiotron H/book).
Johan# you rightly mentioned it's all down to cost and the law of diminishing returns. One can pay a massive sum for a quality lump and someone can say "get same with tranny half the size".
The way forward is with computer simulation, if one can get "ideal performance" with a lower section than going to the trouble of a high sectioned design then so the better. Unfortunately the real world often ends up with larger o/p trannies being a law to themselves and often smaller is better.
That 18 sectioned design I mentioned previously was interleaved with paper (ideal dielectric). Another point is with larger o/p trannies I always have to fit an R/C snubber between the anodes.
Another point that many find is different square wave quality on varying combination taps of o/p transformers i.e 4/8/15 Ohm secs. That can be dealt in another mail.
richj
richwalters said:
Exactly so!
Output impedance taps are often just glibly that, thus excluding part of the secondary and upsetting the "leakage" cart.
- Status
- Not open for further replies.