Hi there Ray-Moth....The distortion in that UL 6L6 circuit is puzzling and goes against all the others.....esp with nfb. Thd in an UL design should be magnitudes lower.........If I had such a design I would use different tubes.............why does an EL34 per Mullard 20 W design specify a thd figure of 0.05% ?
If the EL34 was used as a true pentode p-p then thd would be around 3.8% at 30W with 375V B+
and in UL p-p specifies 1% at 30W with 43% taps at 400V B+. Both these figures are without nfb......so....
This would imply that 6L6's are not ideal for UL ?
richj
If the EL34 was used as a true pentode p-p then thd would be around 3.8% at 30W with 375V B+
and in UL p-p specifies 1% at 30W with 43% taps at 400V B+. Both these figures are without nfb......so....
This would imply that 6L6's are not ideal for UL ?
richj
Consider what is going on with an "ultralinear" circuit.
It's referred to as "Partial Triode Operation Of Pentodes" in the Radiotron handbook, btw.
Triode operation is obtained when the screen is connected to the primary B+ (suitable current limited, of course). But with UL operation it is connected DC to the B+, plus a % of the AC of the ouput signal.
(you could make a circut that applied this signal via a separate stage, and not the ouput tranny... hmmm...)
Imho, this is the critical point. The screen is in effect being modulated by a percentage of the plate signal. Compare with fixed screen operation...
A good exercise is to set up a tube and see how the screen works when it is modulated, not the grid. Try it.
(There's even a circuit in the back of many of the RCA tube books)
Not linear.
Some tubes happen to be "more better" than others when screen modulated. That may account for all sorts of variation in results from UL operation...
And, the 43% tap is a compromise in all respects, and probably ought to be optimized for a given tube or tube type, it's not a universal "magic" point for "UL operation."
Another approach worth of some degree of investigation is "Extended Class A" operation, where a Triode is run Class A, and in parallel with a Pentode biased way up in Class B... there was a recent (90s vintage) company that did this with some sucess... it works out to be sort of like UL in some respects...
The main advantage of UL is "free power" and you don't need a screen supply... but I've always found that triode operation sounds better in the end, at the expense of power output. Ymmv.
Back "in the day" a few extra watts of output power was very attractive to consumers - so UL was widely used. A 60 watt amp was considered to be "BIG." So, getting from 18 watts up to 35 watts by going UL was a big deal to the advertsing people and the consumers. There were no 100 watt amps (sure there were, but no one you knew had one...) except for rich people and in theaters and ball parks... and they were *expensive* if you saw one.
These days, I'd rather go for ultimate triode linearity/ sonic quality over raw power myself.
... just rambling on...
_-_-bear
It's referred to as "Partial Triode Operation Of Pentodes" in the Radiotron handbook, btw.
Triode operation is obtained when the screen is connected to the primary B+ (suitable current limited, of course). But with UL operation it is connected DC to the B+, plus a % of the AC of the ouput signal.
(you could make a circut that applied this signal via a separate stage, and not the ouput tranny... hmmm...)
Imho, this is the critical point. The screen is in effect being modulated by a percentage of the plate signal. Compare with fixed screen operation...
A good exercise is to set up a tube and see how the screen works when it is modulated, not the grid. Try it.
(There's even a circuit in the back of many of the RCA tube books)
Not linear.
Some tubes happen to be "more better" than others when screen modulated. That may account for all sorts of variation in results from UL operation...
And, the 43% tap is a compromise in all respects, and probably ought to be optimized for a given tube or tube type, it's not a universal "magic" point for "UL operation."
Another approach worth of some degree of investigation is "Extended Class A" operation, where a Triode is run Class A, and in parallel with a Pentode biased way up in Class B... there was a recent (90s vintage) company that did this with some sucess... it works out to be sort of like UL in some respects...
The main advantage of UL is "free power" and you don't need a screen supply... but I've always found that triode operation sounds better in the end, at the expense of power output. Ymmv.
Back "in the day" a few extra watts of output power was very attractive to consumers - so UL was widely used. A 60 watt amp was considered to be "BIG." So, getting from 18 watts up to 35 watts by going UL was a big deal to the advertsing people and the consumers. There were no 100 watt amps (sure there were, but no one you knew had one...) except for rich people and in theaters and ball parks... and they were *expensive* if you saw one.
These days, I'd rather go for ultimate triode linearity/ sonic quality over raw power myself.
... just rambling on...
_-_-bear
ok I got the idea about having the screen grid DC as well as AC same and at 43%(or whatever optimum point for the tube) of the anode voltage. Isn't that what Hafler & Kereo and also suggest in their original document ? http://www.aikenamps.com/UL.pdf
but I have no idea why you need to break the bulb and measure the distance between the elements 
to figure out the optimum point ? Can't it be done simply by measurement ?!
but I have no idea why you need to break the bulb
and measure the distance between the elements to figure out the optimum point ? Can't it be done simply by measurement ?!
The gist of the article appears to be that on an AC basis the UL turns ratio should mirror the actual inter-electrode spacing of the tube and the latter can be determined by examining the tube manufacturer's specified voltage limits on a unit-per-unit basis. If the maximum screen voltage is specified as 50% of the plate's the UL tap should be 50% of the primary. Shame he didn't follow it up with test numbers.
Last year I did a test using KT90 & KT88 in a 43% ULp-p circuit at high power (150W) and both tubes behaved different with thd...the KT90 had roughly double the thd compared to the KT88. This proves my point that using a 6L6 in an UL circuit with high thd isn't really at home with it. This may explain why the KT88 is the lion tube in HiFi....although many users mention it's sound is a trifle soft. That is really more to do with the interaction with feedback cricuits and not always blame it on the tube. The EL34, KT66/77/ EL84, etc are all tubes specifically for use in UL hifi designs.....I still rate the 6550A version a supreme tube at home in hifi and MI although a pity no-one makes this version any more.
On averages I find the current cloned 6550 okay in 43% UL solong the B+ is kept down...... and it's thd behaves sim to KT88 in class A.
To be fair........With the KT90 I observed a raft and tangle of harmonics .......odds + evens........compared to the clean harmonics of the 88 and 6550 series. Perhaps I should re visit the thd analyser and tabulate graphically the real differences.
I am still puzzled to vendors of KT90 claiming this tube is okay for HiFi........my opinion is totally other.
richj
On averages I find the current cloned 6550 okay in 43% UL solong the B+ is kept down...... and it's thd behaves sim to KT88 in class A.
To be fair........With the KT90 I observed a raft and tangle of harmonics .......odds + evens........compared to the clean harmonics of the 88 and 6550 series. Perhaps I should re visit the thd analyser and tabulate graphically the real differences.
I am still puzzled to vendors of KT90 claiming this tube is okay for HiFi........my opinion is totally other.
richj
Andrewbee's link was an interesting read, but I see a couple of problems with the approach. First of all, the potential around a cylindrical object varies as 1/radius not linearly with radius. To get it really right would require a Finite Elements Electric field calculation program with a good physical model of the tube in question.
Secondly, if you do successfully arrange for the screen grid to always have the potential that would exist naturally between cathode and plate, then it effectively isn't there, so you have just ended up with a true triode in effect.
An interesting experiment would be to just take a variable resistive divider from the plate to cathode to drive a high volt. MosFet source follower controlling the screen grid. One could separate AC and DC ratios by using a cap and high value resistor to a screen bias supply. Then look at the distortion on a dist., or better yet, spectrum analyzer (M Audio 24 bit sound card, $150) while varying the ratios.
An economy test could be done by just using some variable power supplies for grid 1, screen and plate and a DVM. Set the tube up for a good bias point using a typical plate load resistor. Then adjust the G1 bias thru a range in small increments and see what variation of the screen supply is necessary to get linear plate voltage changes (with respect to G1 changes). Then plot the results. A bit labor intensive though.
One might even go further (out on a limb perhaps!) and arrange for an op. amp. and HV MosFet (drain connected to screen grid with a pullup resistor to B+) to control the screen grid voltage actively for minimum distortion. The op. amp would use a resistive voltage divider between the plate/cathode, and the G1 voltage as its inputs to compare (use inverting Op. Amp. configuration since G1 volts will be neg. and ratio'd plate voltage will be pos.), and its output would control the MosFet gate (make sure to get phases right or will oscullate, maybe will need a 50 KHz RC lowpass for stability).
If this works well enough, one might just use two of these circuits in a P-P amplifier and forget about the whole issue. (But you didn't hear ME say THAT!)
Don
Secondly, if you do successfully arrange for the screen grid to always have the potential that would exist naturally between cathode and plate, then it effectively isn't there, so you have just ended up with a true triode in effect.
An interesting experiment would be to just take a variable resistive divider from the plate to cathode to drive a high volt. MosFet source follower controlling the screen grid. One could separate AC and DC ratios by using a cap and high value resistor to a screen bias supply. Then look at the distortion on a dist., or better yet, spectrum analyzer (M Audio 24 bit sound card, $150) while varying the ratios.
An economy test could be done by just using some variable power supplies for grid 1, screen and plate and a DVM. Set the tube up for a good bias point using a typical plate load resistor. Then adjust the G1 bias thru a range in small increments and see what variation of the screen supply is necessary to get linear plate voltage changes (with respect to G1 changes). Then plot the results. A bit labor intensive though.
One might even go further (out on a limb perhaps!) and arrange for an op. amp. and HV MosFet (drain connected to screen grid with a pullup resistor to B+) to control the screen grid voltage actively for minimum distortion. The op. amp would use a resistive voltage divider between the plate/cathode, and the G1 voltage as its inputs to compare (use inverting Op. Amp. configuration since G1 volts will be neg. and ratio'd plate voltage will be pos.), and its output would control the MosFet gate (make sure to get phases right or will oscullate, maybe will need a 50 KHz RC lowpass for stability).
If this works well enough, one might just use two of these circuits in a P-P amplifier and forget about the whole issue. (But you didn't hear ME say THAT!)
Don
Perhaps the main function of the screen grid is to act like an accelerator - it functions like a phony plate as far as the electrons are concerned, being porous, they aim for it, but slam into the plate instead...
Too many electrons hitting the screen and overload. <flame>
Anyhow, I can't recall now, but is there any appreciable Vdc drop between the 43% tap and the Plate connection?
I thought not. And that the screen was at B+ for DC... no?
_-_-bear
Too many electrons hitting the screen and overload. <flame>
Anyhow, I can't recall now, but is there any appreciable Vdc drop between the 43% tap and the Plate connection?
I thought not. And that the screen was at B+ for DC... no?
_-_-bear
Yeah, normally the ultralinear screen DC component gets operated at the plate B+ voltage. Which is too much for a lot of tubes, particularly horizontal output types. The link in Andrewbee's post to http://www.webace.com.au/~electron/tubes/ulo.html
talks about using a tertiary winding on the transformer for ultralinear taps to allow a lower screen DC voltage. Mentions a few designs that have used this, Bruce de Palma, Acrosound TO-350, Dynaco A-441, McIntosh Mc 3500 and MI-350.
The "Optimised Ultralinear" idea in the article then goes further. Trying to operate the screen at the same DC & AC potential as would exist at the screen's physical location, due to the cathode to plate field, as if the screen weren't there. (by using a calculated % plate winding tap from tube geometry factors) But if the screen has the same potential always as that position, then it has no accelerating influence on the electrons. It's as if it weren't there, transparent except for intercepted electrons. So whats left is a triode.
Its pretty clear from published tube curves for standard ultralinear (usually 43% tap) that they are not normally operated in that manner, since the plate curves don't look anything like triode curves.
Don
talks about using a tertiary winding on the transformer for ultralinear taps to allow a lower screen DC voltage. Mentions a few designs that have used this, Bruce de Palma, Acrosound TO-350, Dynaco A-441, McIntosh Mc 3500 and MI-350.
The "Optimised Ultralinear" idea in the article then goes further. Trying to operate the screen at the same DC & AC potential as would exist at the screen's physical location, due to the cathode to plate field, as if the screen weren't there. (by using a calculated % plate winding tap from tube geometry factors) But if the screen has the same potential always as that position, then it has no accelerating influence on the electrons. It's as if it weren't there, transparent except for intercepted electrons. So whats left is a triode.
Its pretty clear from published tube curves for standard ultralinear (usually 43% tap) that they are not normally operated in that manner, since the plate curves don't look anything like triode curves.
Don
One of the rare examples demonstrating the effect of screen voltage and tap ratio on UL curves:
http://www.mclink.it/com/audiomatica/tubes/kt88.htm
http://www.mclink.it/com/audiomatica/tubes/kt88.htm
Hmmm, guess I have to eat my words, it DOES look like a triode's curves except near Vp=0. Maybe just varying the effective Mu versus % tap point, since the plate resistance is changing. Let's see: Mu = gm * Rp
So, how about using an Op. Amp. and HV MosFet to control the screen voltages for minimum plate voltage non-linearity, instead of xfmr taps. Could call it Hybrid UltraLinear. Well, I guess we Could use an LTP diff stage to control the screen grid voltages, for tube Purists.
Don
So, how about using an Op. Amp. and HV MosFet to control the screen voltages for minimum plate voltage non-linearity, instead of xfmr taps. Could call it Hybrid UltraLinear. Well, I guess we Could use an LTP diff stage to control the screen grid voltages, for tube Purists.
Don
I should have put in some labels on the resistors. But, anyway, here's the intended mode of operation.
The plate (to LTP grid)attached resistor would be N times the value of the grid (to LTP grid) attached resistor, for an anticipated gain of N for plate swing versus input drive swing on the main grids.
If the pentode/tetrode is performing up to snuff at gain N, then the LTP grids remain stationary due to the currents thru the resistors cancelling, and the screen voltages stay the same.
If the plate is underperforming the gain of N, then the LTP grids get some signal and correct the screen voltages until the plate swings are nearly correct again. etc. (typical feedback loop)
Don
The plate (to LTP grid)attached resistor would be N times the value of the grid (to LTP grid) attached resistor, for an anticipated gain of N for plate swing versus input drive swing on the main grids.
If the pentode/tetrode is performing up to snuff at gain N, then the LTP grids remain stationary due to the currents thru the resistors cancelling, and the screen voltages stay the same.
If the plate is underperforming the gain of N, then the LTP grids get some signal and correct the screen voltages until the plate swings are nearly correct again. etc. (typical feedback loop)
Don
Hi,
Messers D. Williamson (MO Valve Co) and Peter Walker (Quad) co-authored a paper "Amplifiers and Superlatives" which appeared in the Sept 1952 issue of Wireless World. A fascinating paper, as they examine the emergence of the Ultra-liner circuit by Hafler and Keroes and compare the same to tetrode and triode stages. More importantly, this paper was the first disclosure by Walker of his circuit involving the cathode in the output transformer. (QuadII)
My original "magazine-rip" copy disappeared, but after some time I located a copy on the Web. Unfortunately, the copy does not disclose a source, so a Google-exercise seems mandatory......
The authors did not like the ultra-linear connection - they provided 'scope photos of responses involving moving the screen from triode to tetrode operation. Yes, the QuadII circuit has the screen returned to B+, but through an LC filter..........
Graeme
Messers D. Williamson (MO Valve Co) and Peter Walker (Quad) co-authored a paper "Amplifiers and Superlatives" which appeared in the Sept 1952 issue of Wireless World. A fascinating paper, as they examine the emergence of the Ultra-liner circuit by Hafler and Keroes and compare the same to tetrode and triode stages. More importantly, this paper was the first disclosure by Walker of his circuit involving the cathode in the output transformer. (QuadII)
My original "magazine-rip" copy disappeared, but after some time I located a copy on the Web. Unfortunately, the copy does not disclose a source, so a Google-exercise seems mandatory......
The authors did not like the ultra-linear connection - they provided 'scope photos of responses involving moving the screen from triode to tetrode operation. Yes, the QuadII circuit has the screen returned to B+, but through an LC filter..........
Graeme
I just thought of another neat little improvement. If the transformer has UL taps, then the plate resistors of the LTP plates could be taken up to them instead of B+. That way the LTP doesn't hardly have to do much to correct the screens to get the gain of N rule enforced, so should be more accurate. (with an effectively higher feedback gain in the LTP)
Well, I guess one would need to check the actual operation to see if the screen grids are being driven by the LTP close to the UL tap swings or not. If the LTP plate load resistors have near identical voltage swings on both ends, then they look like current source loads and the LTP would have higher gain available for feedback. Pentodes, rather than triodes, in the LTP might be better for gain too.
Don
Well, I guess one would need to check the actual operation to see if the screen grids are being driven by the LTP close to the UL tap swings or not. If the LTP plate load resistors have near identical voltage swings on both ends, then they look like current source loads and the LTP would have higher gain available for feedback. Pentodes, rather than triodes, in the LTP might be better for gain too.
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.