Re lundahl push-pull tranny....I ahven't used these......only to add that one pays for swedish iron and C core......and O.M.V that wound with 4 sections with single o/p Z isn't enought for todays varied LS impedances. Lundahl stipulates 30Hz LF cutoff freq. With alot of digital reply material these days.....this lower response isn't low enough.....20Hz min. Rule of thumb....p-p o/p trans half-size per double freq.
There's an article (single ended amps..et alia) from Sound Practices spring 1994 vol 2 which digresses more into differences between triode & pentodes in output stages. (It's rather long and I have no copying possibility up here).
UL transformer design is generally more design critical... that is sectional balance parameters tighter when compared to o/p tranny for standard tetrode config.....A good quality UL transformer may have 7-14 sections..(reflected in price) and the interwinding capacitances and resonant leakage inductances have more influence on the % UL tap .....and high gm tubes usually require g2-a snubber networks to avoid ringing and oscillation. Not to put off....read on.
Many early designs using Mullard EL84's and EL34's 20W amp (and others) both in UL mode have no record of output stage instability ..the quality of the o/p tranny is the single most important item....it's too easy to say all's good when in reality alot comes flying back into ones face.....with some transformers these snubber networks won't be needed as the UL taps become a greater part of the primary i.e towards triode........result lower leakage inductance and a lesser chance of output stage instability problems. But look at p-p triode drive requirements for a meagre o/p....Disadvantages with true tetrode is high intermodulation 15% at moderate Pouts compared with 4% with same Pout with UL at 43% taps........BUT apply nfb to UL and tetrode p-p and the intermodulation ratio drops around 1% i.e roughly 5:1 with 14dB feedback.. Another minor drawback with nfb req's is more input gain req....Whether true tetrode or UL.....with nfb feedback the advantages and drawdacks from each o/p stage can eclipse each other and in many instances the type of sound determines the choice of stage. In my designs I chose UL fixed bias because I blast trumpet for power and also use amp as HiFi with low thd. Fixed bias with no nfb creates a lovely sound with wind instruments......but with nfb is better on replay. UL is a compromise between true tetrode and triode.......
I recently designed a p-p 150W amp using 4 x 6550 with 43% taps....and to avoid any hint of ringing in the UL output stage without curtailing upper freq bandwidth of 50Khz wasn't easy.
Things go wrong in an amplifier which cannot drive any load...debugging an inherent flaw from a crappy design...the worst being an amp trying to drive an LS with a capacitive load.
tough subject ...up to the individual.# more anon
richj
There's an article (single ended amps..et alia) from Sound Practices spring 1994 vol 2 which digresses more into differences between triode & pentodes in output stages. (It's rather long and I have no copying possibility up here).
UL transformer design is generally more design critical... that is sectional balance parameters tighter when compared to o/p tranny for standard tetrode config.....A good quality UL transformer may have 7-14 sections..(reflected in price) and the interwinding capacitances and resonant leakage inductances have more influence on the % UL tap .....and high gm tubes usually require g2-a snubber networks to avoid ringing and oscillation. Not to put off....read on.
Many early designs using Mullard EL84's and EL34's 20W amp (and others) both in UL mode have no record of output stage instability ..the quality of the o/p tranny is the single most important item....it's too easy to say all's good when in reality alot comes flying back into ones face.....with some transformers these snubber networks won't be needed as the UL taps become a greater part of the primary i.e towards triode........result lower leakage inductance and a lesser chance of output stage instability problems. But look at p-p triode drive requirements for a meagre o/p....Disadvantages with true tetrode is high intermodulation 15% at moderate Pouts compared with 4% with same Pout with UL at 43% taps........BUT apply nfb to UL and tetrode p-p and the intermodulation ratio drops around 1% i.e roughly 5:1 with 14dB feedback.. Another minor drawback with nfb req's is more input gain req....Whether true tetrode or UL.....with nfb feedback the advantages and drawdacks from each o/p stage can eclipse each other and in many instances the type of sound determines the choice of stage. In my designs I chose UL fixed bias because I blast trumpet for power and also use amp as HiFi with low thd. Fixed bias with no nfb creates a lovely sound with wind instruments......but with nfb is better on replay. UL is a compromise between true tetrode and triode.......
I recently designed a p-p 150W amp using 4 x 6550 with 43% taps....and to avoid any hint of ringing in the UL output stage without curtailing upper freq bandwidth of 50Khz wasn't easy.
Things go wrong in an amplifier which cannot drive any load...debugging an inherent flaw from a crappy design...the worst being an amp trying to drive an LS with a capacitive load.
tough subject ...up to the individual.# more anon
richj
Ok Thats it, phn. I am going to say it...
I have noticed on MANY occassions in mine and other member's thread and in other forums that your sole intention of posting is disrupting, skewing and hijacking a current discussion and you are demonstrating that once again here as I write. Please stop this. Now!
I repeat - if you do not have anything constructive to contribute to the subject please stop posting.
I have noticed on MANY occassions in mine and other member's thread and in other forums that your sole intention of posting is disrupting, skewing and hijacking a current discussion and you are demonstrating that once again here as I write. Please stop this. Now!
I repeat - if you do not have anything constructive to contribute to the subject please stop posting.
rdf said:
Well, that settles it then. We should shift over to the Chipamp Forum, doing otherwise is irrational.
Nothing wrong with being a little irrational. That's why some people (I'm one) would rather drive a '57 T-bird than a BMW. It's not that the T-bird is objectively faster or handles better or gets better mileage- it's because it's really cool in a retro sort of way.
Zobels - Anode to Screen Tap for ultralinear
Adding Zobel Networks across:
a) The Full Primary (Anode to Anode)
b) Half Promary (Anode to CT)
OR
c) Anode to Screen Tap on each side
This is a topic which causes a lot of grief - When designing a recent amplifier (moderate cost) 4 x KT88 in Ultralinear with a Hammond 1650T Output tranny I asked a lot of my network of valve gurus how best to do this. The answers were basically that there was no correct way and that it was largely and exersize in "trial and error".
My response (as an Electronic Design Eng) was "****, there must be a correct, relatively simple way" of determining the position (ie a,b OR c above) and the component values. When people say that trial and error is the only way to do something its a FLAG that they don't understand what they are doing.
There has been a LOT of work done on suppressing resonances (ringing) in transformers in recent years - MOSTLY to do with DC to DC Converter design. At least one person has interpreted this information and applied it to valve interstage and output transformers.
Read THIS
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html
I found this method worked well - with the Hammond 1650T I found that the optimal position of the zobels was anode to screen tap on each side (that is c above) AND that for optimum results the required zobels were slightly different on the push side than the pull side - from memory the zobels were 1K and 4n7 on the BROWN wires side and 1K2 and 6n9 (4n7 parallel 2n2) on the BLUE wires side - If anyone is using the 1650T and wants the exact values email me and I'll go back thru' my work book to check that these were in fact the values I used.
REMEMBER - do this to the amplifier open loop i.e. BEFORE applying any feedback.
Cheers,
Ian
Adding Zobel Networks across:
a) The Full Primary (Anode to Anode)
b) Half Promary (Anode to CT)
OR
c) Anode to Screen Tap on each side
This is a topic which causes a lot of grief - When designing a recent amplifier (moderate cost) 4 x KT88 in Ultralinear with a Hammond 1650T Output tranny I asked a lot of my network of valve gurus how best to do this. The answers were basically that there was no correct way and that it was largely and exersize in "trial and error".
My response (as an Electronic Design Eng) was "****, there must be a correct, relatively simple way" of determining the position (ie a,b OR c above) and the component values. When people say that trial and error is the only way to do something its a FLAG that they don't understand what they are doing.
There has been a LOT of work done on suppressing resonances (ringing) in transformers in recent years - MOSTLY to do with DC to DC Converter design. At least one person has interpreted this information and applied it to valve interstage and output transformers.
Read THIS
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html
I found this method worked well - with the Hammond 1650T I found that the optimal position of the zobels was anode to screen tap on each side (that is c above) AND that for optimum results the required zobels were slightly different on the push side than the pull side - from memory the zobels were 1K and 4n7 on the BROWN wires side and 1K2 and 6n9 (4n7 parallel 2n2) on the BLUE wires side - If anyone is using the 1650T and wants the exact values email me and I'll go back thru' my work book to check that these were in fact the values I used.
REMEMBER - do this to the amplifier open loop i.e. BEFORE applying any feedback.
Cheers,
Ian
Presumably, people want to drive actual loudspeakers rather than test resistors. With that in mind, I urge anyone contemplating output stages to look at the Crowhurst reference I gave earlier, and particularly to Figures 203, 208, and 211.
Unfortunately, copyright inhibits me from posting those figures, but in short, they graph variation in distortion and output power for varying load impedances for pentode (tetrode), triode, and UL. These are, of course, open loop curves, so the designer needs to think through the effects of feedback, if that path is chosen.
Unfortunately, copyright inhibits me from posting those figures, but in short, they graph variation in distortion and output power for varying load impedances for pentode (tetrode), triode, and UL. These are, of course, open loop curves, so the designer needs to think through the effects of feedback, if that path is chosen.
It's interesting that you would bring this up. I personally don't believe what Crowhurst wrote is true and it has never been borne out in any other context. What I mean is, I have read dozens of other texts, engineering and popular, and have found no other reference to this idea of his. As smart as the man was, I wonder if he really knew what he was talking about in this instance. He is the only person I've read who talks about an inductive load in this manner, and I believe he is incorrect. All other theories, experiments and tests have shown that in an Ultralinear or triode transformer-coupled tube output stage, distortion always decreases as impedence increases.
John
Actually, the curves shown in the Mullard "Circuits for Audio Amplifiers" look qualitatively similar (they are for a different tube, so it's not surprising that they're not identical). You see a minimum, then the distortion starts rising again.
What makes interpretation difficult is that the distortion curves in Crowhurst (for a 5881) are taken at full power, which itself is varying as the load impedance varies. What would be nice would be to see the curves at a single reference power level, say, 5W. The Mullard reference shows curves for the EL34 at 20W for UL, and unspecified (but presumably at full power) for the triode connection.
Here's a case where I wish I had my own data. I'd love to see if someone else has replicated (or not) Crowhurst's curves.
What makes interpretation difficult is that the distortion curves in Crowhurst (for a 5881) are taken at full power, which itself is varying as the load impedance varies. What would be nice would be to see the curves at a single reference power level, say, 5W. The Mullard reference shows curves for the EL34 at 20W for UL, and unspecified (but presumably at full power) for the triode connection.
Here's a case where I wish I had my own data. I'd love to see if someone else has replicated (or not) Crowhurst's curves.
Well, yes, of course for a given circuit at full power it may well be that way, but not because the speaker voice coil presents an inductance of a few tenths of a millihenry. It's more likely because the output tube itself (as well as the other tubes in the circuit) must be driven harder in order to make the same power with a higher primary to secondary turns ratio.
John
SY said:
In moderation that's unconventional, not irrational, with which I empathize completely. In my case there's no cycling better than on a fixed gear but, as you say, I'ld never claim it's the fastest or most efficient over long distances and varied terrain. It's the experience I enjoy.
Tubes, however, I drifted into long before they were considered cool because to my ear they sound closer to the real thing. This was at a time I was engineering a weekly classical music competion for live broadcast with finals at Orchestra Hall in Detroit, so it wasn't for lack of exposure. The youngest brother inherited my Bryston 2B not long after I was given a MC-240 and I never regretted it. Now I'm being mezmerised by a low power SET on Jordan JX92Ss. At work I spend all day listening to solid state gear and every manner of speaker, including digital-input-biamped systems. Even my office speakers are Tannoy 800's. Within its limitation none of those approach the wealth of natural musical detail of the peanut whistle system at home. If I was building only for the cool retro factor and thought an LM3886 was in every way a more accurate sound reproduction device because of its measured performance, now that would be irrational.
However I'm not one to discount the meaningful correlation between some numbers and audible performance. The SET's a Mullard 3-3 and THD rarely exceeds .1% at normal listening with a decent damping factor. That's what's so interesting about the current topic, the possibility of getting good numbers (dist, Zout, etc.) with a topology I'm wagering is responsible for the highly musical quailities of my system without resorting to big dollops of global NFB with its attendent difficulties and effect on the distortion spectrum. I'm far from powercntrl's Category 2 (though he might not agree) and looking forward to seeing what this thread kicks up.
phn said:
Sorry phn, I thought the conversation was drifting towards 'all amps sound the same' and my Usenet muscle twitched. I should stop posting from work (and reading Usenet!)
Sorry my comments were not what you were expecting. However, you might like to read what Max Robinson has to say at his "Fun With Tubes" web site. He discusses this subject on his Ultralinear or Pentode? page.
Yes, this is probably true. The great advantage of Ultralinear has always been lower IM distortion and I doubt Robinson measured that. You should always remember too that ham radio guys are hardly ever audiophiles (Joe Roberts is a notable exception). They basically are interested in a powerful amplifier that measures well and call that good. My friends and I at the Dallas Audio Club build and listen, and tweak from there. I've never taken anything but the simplest measurements on my amps.
John
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.