In effect in a sort of fixed non floating point way, you are implying low loading conditions when I quies is substantially lower than for class A with cathode resistors, but amp in low loading conditions behaves in fixed bias on both speech and music but max o/p only available on short music/speech passages. That's the nub.
Low loading produces lower thd but higher peak current conditions than for normal AB Iquies setup.
Catch ? Higher o/p tranny Z required. As I see it, amp cannot be cont driven higher than a few watts to measure thd.
The implications of not be able to measure thd properly, is down to the listener to detect it on sound passages. Tricky on non trained ear, but a gated burst tone generator/thd analyser can do it.
I need to read up more on this technique as it has alot going for it, and any lay method regarding tube longevity should be examined.
Perhaps other members can fill in.
richj
Low loading produces lower thd but higher peak current conditions than for normal AB Iquies setup.
Catch ? Higher o/p tranny Z required. As I see it, amp cannot be cont driven higher than a few watts to measure thd.
The implications of not be able to measure thd properly, is down to the listener to detect it on sound passages. Tricky on non trained ear, but a gated burst tone generator/thd analyser can do it.
I need to read up more on this technique as it has alot going for it, and any lay method regarding tube longevity should be examined.
Perhaps other members can fill in.
richj
I have been experimenting with methods to reduce the idle dissipation in a class A amp for about a year. I have tried several methods to vary the voltage or current (or both) in the output tubes in a dynamic fashion. Most of the methods proved useless. If you think about it, a current change through the output tubes looks just like signal to the tube, and to your speakers.
I had to discover this the hard way. I developed a few theories on efficiency improvement that needed to be tested, so I designed and built this fancy DSP powered controller where every parameter on the output tubes and the driver tubes could be modified in real time according to the instantaneous needs of the amp. I have built several different amplifier designs and connected them up to the controller. Several different programs have been tested. Any attempt to vary the bias voltage and thus the tube current in real time is met with the two following obstacles:
In a conventional capacitor coupled amplifier the bias change must overcome the charge held in the coupling capacitor. The time constant formed by the coupling cap and the grid resistor is by design too slow to allow dynamic change. So I ditched the cap and installed an interstage transformer.
Now I could change the output tubes bias at any rate that I choose, what did I discover? If you have a sufficiently large OPT Hammond 1628SE) the speaker cone moves in step with your bias change. Slow changes (compared to the actual audio signal) cause a resultant distortion caused by the operating point shift causing OPT saturation. Fast changes look just like the audio signal. I wrote a DSP program to vary the grid bias exactly in step with the audio signal, and discovered that I could get sound without the driver tube in its socket. It sounded ugly, but I had created a Class A digital vacuum tube amp, not what I had intended.
Is there any hope? Yes I have found a few possibilities.
It may be possible to vary the current, voltage, or both through a Class A push pull amp in a fashion that keeps the tubes operating with just enough headroom for the given signal. If the two tubes are controlled carefully enough so that their currents are exactly equal the dynamic changes will cancel in the OPT.
It IS possible to vary the plate voltage in a cathode follower output stage in step with the drive signal such that the VOLTAGE ACROSS THE OUTPUT TUBE remains constant. It turns out that there are some other neat side effects to this. When the voltage across the tube remains constant all of the distortion products that are created by tube parameters shifting with applied voltage go away. Now you pick a tube that can operate with say 100 volts across it, crank the current up to say 300 mA, and you can get about 30 watts of output power for 30 watts of tube dissipation at less than 1% distortion. The output tube OPERATES ENTIRELY IN CLASS A all of the time.
I have built this amp. It works quite well. The power supply that constantly varies its output voltage over the range of +420 to - 300 volts in step with the audio signal is another matter. It blows up a lot! I have made some serious progress over the past six months, with more to come.
The amplifier and DSP controller design was submitted to a design contest that is now over so the details will be posted on my web site as time permits over the next month or two. I have almost 1 years worth of notes, experiments, and data to go through. Most of it is on paper.
I had to discover this the hard way. I developed a few theories on efficiency improvement that needed to be tested, so I designed and built this fancy DSP powered controller where every parameter on the output tubes and the driver tubes could be modified in real time according to the instantaneous needs of the amp. I have built several different amplifier designs and connected them up to the controller. Several different programs have been tested. Any attempt to vary the bias voltage and thus the tube current in real time is met with the two following obstacles:
In a conventional capacitor coupled amplifier the bias change must overcome the charge held in the coupling capacitor. The time constant formed by the coupling cap and the grid resistor is by design too slow to allow dynamic change. So I ditched the cap and installed an interstage transformer.
Now I could change the output tubes bias at any rate that I choose, what did I discover? If you have a sufficiently large OPT Hammond 1628SE) the speaker cone moves in step with your bias change. Slow changes (compared to the actual audio signal) cause a resultant distortion caused by the operating point shift causing OPT saturation. Fast changes look just like the audio signal. I wrote a DSP program to vary the grid bias exactly in step with the audio signal, and discovered that I could get sound without the driver tube in its socket. It sounded ugly, but I had created a Class A digital vacuum tube amp, not what I had intended.
Is there any hope? Yes I have found a few possibilities.
It may be possible to vary the current, voltage, or both through a Class A push pull amp in a fashion that keeps the tubes operating with just enough headroom for the given signal. If the two tubes are controlled carefully enough so that their currents are exactly equal the dynamic changes will cancel in the OPT.
It IS possible to vary the plate voltage in a cathode follower output stage in step with the drive signal such that the VOLTAGE ACROSS THE OUTPUT TUBE remains constant. It turns out that there are some other neat side effects to this. When the voltage across the tube remains constant all of the distortion products that are created by tube parameters shifting with applied voltage go away. Now you pick a tube that can operate with say 100 volts across it, crank the current up to say 300 mA, and you can get about 30 watts of output power for 30 watts of tube dissipation at less than 1% distortion. The output tube OPERATES ENTIRELY IN CLASS A all of the time.
I have built this amp. It works quite well. The power supply that constantly varies its output voltage over the range of +420 to - 300 volts in step with the audio signal is another matter. It blows up a lot! I have made some serious progress over the past six months, with more to come.
The amplifier and DSP controller design was submitted to a design contest that is now over so the details will be posted on my web site as time permits over the next month or two. I have almost 1 years worth of notes, experiments, and data to go through. Most of it is on paper.
Attachments
Tubelab,
Interesting, great minds think alike. During the early 1990's I rabbled with a Texas DSP with the aim of examining a floating point tube amp biassing concept but in/end discarded it simply that it took away the kudos of what a tube amp should really be. i.e self governing. i.e should be able to run by itself without SS intervention ! In any case now I'm still against the SS getting too involved with tube stuff..
In certain applications DSP (esp. echo cancelling) is a very strong mathematical arguement and where I was in R&D for a large company it was a very effective in communications cleanup. Now the applications list is everywhere.
Back to the tube amp. The main problem, was having got a DSP algorthim to run, the physics and math looked good, the trouble started when a transient appeared causing instantaneous thd ; let alone coping with out of balance p-p. Hanking with variable switch-mode psu is another area worth persuing , but again on reproducing a transient, instant cycle per cycle response is required. Switchmode on a basis can do it, but there are so many other issues of topology, time constants, EMC. ....Herr-Gott ! enough to pull my hair out.
When I get closer to retirement I might try a SE design which inflicts less of a wound on the pocket.
So I'm back to sq'one. For simplicity I prefer the low loading concept and with todays test gear makes easy setup and analysis. The Radiotron handbook 4th ed has no mention of Low Loading.
richj
Interesting, great minds think alike. During the early 1990's I rabbled with a Texas DSP with the aim of examining a floating point tube amp biassing concept but in/end discarded it simply that it took away the kudos of what a tube amp should really be. i.e self governing. i.e should be able to run by itself without SS intervention ! In any case now I'm still against the SS getting too involved with tube stuff..
In certain applications DSP (esp. echo cancelling) is a very strong mathematical arguement and where I was in R&D for a large company it was a very effective in communications cleanup. Now the applications list is everywhere.
Back to the tube amp. The main problem, was having got a DSP algorthim to run, the physics and math looked good, the trouble started when a transient appeared causing instantaneous thd ; let alone coping with out of balance p-p. Hanking with variable switch-mode psu is another area worth persuing , but again on reproducing a transient, instant cycle per cycle response is required. Switchmode on a basis can do it, but there are so many other issues of topology, time constants, EMC. ....Herr-Gott ! enough to pull my hair out.
When I get closer to retirement I might try a SE design which inflicts less of a wound on the pocket.
So I'm back to sq'one. For simplicity I prefer the low loading concept and with todays test gear makes easy setup and analysis. The Radiotron handbook 4th ed has no mention of Low Loading.
richj
Microchip makes a few dsPIC (dsp enabled PIC chip) chips that have SMPS features built into them, 2 MSPS 10 bit A/D's too, so the audio modulated buck converter becomes doable.
I imagine that most tubeheads would have a hard time accepting a DSP based tube amp. I have found no tube based methods for the kind of efficiency enhancements demonstrated by this amp, yet. But I have already tested the idea of using one cathode follower (or source follower) to "modulate" the supply voltage on another cathode follower. It works, and affords the same kinds of distortion improvements as the DSP based system, yet contains no silicon. The efficiency is low, but the distortion is in the 0.1% range at full power. The "distortion signature" can still be adjusted as in the DSP solution.
I never would have discovered any of this if I didn't start playing around with DSP stuff.
tubelab.com said:
That has rekindled my thoughts....a quantum leap when
I started DSP at a time when Motorola emulation debug mainframe was trend 1989....but speeds were very slow compared to todays 32bit standards. However in todays energy fussy market it should be actively encouraged and the Class A tube amp is a shockingly energy wasteful and makes a good candidate.
I am using DSP for intelligent PFC power supplies for renewable sources but annoyingly but anticipated the subject of EMC filtering slows down performance for compliance. This costs more than all the dsp bits summed up !
I'm going to have another look at high speed psu's, but but the end subject is compliance. The authorities here are serious about this.
richj
This problem is not new...
This is the same situation that audio compressors were dealing with in the 1940's and 1950's... Some units were modulating the DC bias for AGC for record cutting as well as FM stereo modulation techniques... The dreaded "thump" was very audible.....monkeying with the time constants reduced the problem but it degarded the performance of the unit...
The way they solved it was to use a P-P balanced stage to cancel the effects of the shifting DC...since the DC thump is a common-mode noise in a differential audio system...a balancing pot is used durring calibration to best null this thumping...
ALso...the "quasi" Class A circuit proved to be efficient....low idle current with a modulated bias scheme....but the sacrifice is LINEARITY...
Yes you can make such as bias scheme that follows the envelope of the applied audio Signal and make it work fine....but you give up the linearity.... Mathamatically if you do the systems analysis for this problem...the solution will always arrive and point to the non-linearity of the gm of the device.... The typical solution to maintaining a constant gm is a using a constant current source, which brings you back to the original dilema of efficiency vs. linearity.....
Chris
This is the same situation that audio compressors were dealing with in the 1940's and 1950's... Some units were modulating the DC bias for AGC for record cutting as well as FM stereo modulation techniques... The dreaded "thump" was very audible.....monkeying with the time constants reduced the problem but it degarded the performance of the unit...
The way they solved it was to use a P-P balanced stage to cancel the effects of the shifting DC...since the DC thump is a common-mode noise in a differential audio system...a balancing pot is used durring calibration to best null this thumping...
ALso...the "quasi" Class A circuit proved to be efficient....low idle current with a modulated bias scheme....but the sacrifice is LINEARITY...
Yes you can make such as bias scheme that follows the envelope of the applied audio Signal and make it work fine....but you give up the linearity.... Mathamatically if you do the systems analysis for this problem...the solution will always arrive and point to the non-linearity of the gm of the device.... The typical solution to maintaining a constant gm is a using a constant current source, which brings you back to the original dilema of efficiency vs. linearity.....
Chris
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.