Well, let's look at some of the work that Pass did with conjugate networks.
For a Lowther DX55 in a KleinHorn, the LF compensation network has a 15 ohm resistor to knock down the resonant peak. Put that in parallel with the speaker's 90 ohm peak and the FirstWatt amp's 47 ohm output impedance and you've got about 10 ohms, which is pretty close to the impedance minimum.
For a Fostex FE166E in the Abbey box, the compensation network has a 10 ohm resistor. With the same amp source Z in parallel with the 85 ohm maximum, the total impedance comes to 7.5 ohms at resonance, a bit below the impedance minimum.
Likewise, if you equalize rather than use speaker-level networks, you'll have to knock down the drive power at resonance by the same proportion.
For a Lowther DX55 in a KleinHorn, the LF compensation network has a 15 ohm resistor to knock down the resonant peak. Put that in parallel with the speaker's 90 ohm peak and the FirstWatt amp's 47 ohm output impedance and you've got about 10 ohms, which is pretty close to the impedance minimum.
For a Fostex FE166E in the Abbey box, the compensation network has a 10 ohm resistor. With the same amp source Z in parallel with the 85 ohm maximum, the total impedance comes to 7.5 ohms at resonance, a bit below the impedance minimum.
Likewise, if you equalize rather than use speaker-level networks, you'll have to knock down the drive power at resonance by the same proportion.
And one more thing- if you go back to the Mills/Hawksford paper, the driver they used (one with a more conventional impedance curve than the Lowthers) gave a peak of 8 dB at resonance. That means that to equalize, you'd reduce power in that region by a factor of 6.3.
SY said:
Huh? Why are you calculating the load impedance seen by the amplifier by including the amplifier's 47 ohm output impedance in parallel with the 15 ohm resistor and the speaker's 90 ohms?
Not that it makes a huge differnece in this case (13 ohms versus 10 ohms), I just don't understand why you're including it.
se
I include it because it's in parallel with the Norton source. For the case of an amp with a very high source Z, that resistance is a real physical resistor, i.e., actually part of the load.
SY said:
If it's part of the load, then it's not the source is it?
The Norton source is fine if you're wanting to know what it looks like looking into the output of the amp. But if you're just lumping everything together into a single shunt resistance, how do know how much power is being delivered to the actual load, i.e. the loudspeaker?
That was the whole point here, wasn't it? When you said that Fs is the point where you least need power.
se
A Norton source that has the compliance to deliver its current sinks more power into a high impedance load (I squared R). I don't see an elegant way around the mechanical resonance with current drive other than an active crossover and voltage/Thiele-Small drive to the bass driver, and current drive to the remainder above their resonance.
I'm not at all happy about Zobel networks plus current drive. The whole thrust of the Hawksford paper was that current drive reduced distortion, so why perform equalisation that increases it? Put the equalisation before the amplifier.
I'm not at all happy about Zobel networks plus current drive. The whole thrust of the Hawksford paper was that current drive reduced distortion, so why perform equalisation that increases it? Put the equalisation before the amplifier.
This is for pedagogical purpose, to try to make myself understood. Clearly, the load absorbs power in inverse proportion to its impedance relative to any shunt impedances. And that brings me back to my original point- at resonance, the power needed for a given SPL is at a minimum.
SY said:
Well you've got me so dizzy on this one I think I'm just gonna go pop a couple Dramamine and go lay down.
se
SY said:
Nope. I distincly remember seeing your backside the entire trip. 🙂
I'll give you a call tomorrow. I think only a realtime discussion can straighten this out.
se
Hello SY,
I don't know very much about electronics, and less about tube circuits, but your conversion of a ST70 to high output impedance is very interesting to me since I have a pair of Dyna Mark 4's that need to be freshened up (they're basically the same as a ST70), and I like the idea of converting them to high-output-impedance.
Do I understand correctly that you simply disconnected the feedback loop? Does the amp function correctly after this mod without further twiddling? Also, what was the nature of the deep bass ugliness? Does this mean they are unusable below 100Hz?
(Newb alert! Please phrase answers simply and use small words! 🙂 )
I don't know very much about electronics, and less about tube circuits, but your conversion of a ST70 to high output impedance is very interesting to me since I have a pair of Dyna Mark 4's that need to be freshened up (they're basically the same as a ST70), and I like the idea of converting them to high-output-impedance.
Do I understand correctly that you simply disconnected the feedback loop? Does the amp function correctly after this mod without further twiddling? Also, what was the nature of the deep bass ugliness? Does this mean they are unusable below 100Hz?
(Newb alert! Please phrase answers simply and use small words! 🙂 )
OK, I'll try to be gentle!
The amp I used does NOT use the stock Dynaco input stage, but an input stage that is lower in gain and more linear. If you simply disconnect the feedback loop on the stock Dyna, the gain will be rather unmanageable and you'll likely have oscillation. Oh, yes, and you'll have a decidedly non-flat frequency response because of the compensation bits thrown in here and there.
Long story short: you need to replace the driver board with something lower gain and more linear. If you're not up to scratch-building something, the Curcio cascode diff amp is available in kit/PCB form and is quite nice. At that point, you can lose the feedback network.
But you still have one more major task: the output stage needs to be converted to pentode. As it sits, it's configured as an ultralinear topology, where the screen grids are being fed by taps on the output transformer primary. The screens need to be disconnected from those taps (the tap leads are unused and can be terminated with some heat shrink tubing to prevent accidental sparks and bangs) and connected to a regulated power supply.
Now that's not a trivial issue, but not insurmountable. Though purist will scream (and they have, they have), a nice simple solid state regulator works superbly in that position. I recommend the floating 317 circuit published by National Semiconductor. If you have a copy of Morgan Jones's "Valve Amplifiers," he goes through that circuit in some detail. I have issues with some of the design details (which, in all fairness, may be my own non-understanding), but what he shows as a circuit is not terribly complicated- nearly part for part what NS recommends- and is inexpensive and reliable. And performs well. For some $$$, you can go the Curcio regulator route with all the bells and whistles, but that's probably overkill.
As far as bass is concerned, output transformers on their own are quite ugly. Driving them from a high source impedance makes them uglier yet. OTOH, I'm firmly of the opinion that tube amps ought to be restricted to frequencies above 100Hz anyway... and indeed, at 60 Hz, the distortion was quite tolerable up to 1.5 amps or so. Additionally, to keep bass response reasonably flat, you're going to need some heavy-duty EQ or use motional feedback. Better yet, kill both birds with a single pebble and biamp, relieving the poor little tube amp of the onerous duty of scraping the bottom.
The amp I used does NOT use the stock Dynaco input stage, but an input stage that is lower in gain and more linear. If you simply disconnect the feedback loop on the stock Dyna, the gain will be rather unmanageable and you'll likely have oscillation. Oh, yes, and you'll have a decidedly non-flat frequency response because of the compensation bits thrown in here and there.
Long story short: you need to replace the driver board with something lower gain and more linear. If you're not up to scratch-building something, the Curcio cascode diff amp is available in kit/PCB form and is quite nice. At that point, you can lose the feedback network.
But you still have one more major task: the output stage needs to be converted to pentode. As it sits, it's configured as an ultralinear topology, where the screen grids are being fed by taps on the output transformer primary. The screens need to be disconnected from those taps (the tap leads are unused and can be terminated with some heat shrink tubing to prevent accidental sparks and bangs) and connected to a regulated power supply.
Now that's not a trivial issue, but not insurmountable. Though purist will scream (and they have, they have), a nice simple solid state regulator works superbly in that position. I recommend the floating 317 circuit published by National Semiconductor. If you have a copy of Morgan Jones's "Valve Amplifiers," he goes through that circuit in some detail. I have issues with some of the design details (which, in all fairness, may be my own non-understanding), but what he shows as a circuit is not terribly complicated- nearly part for part what NS recommends- and is inexpensive and reliable. And performs well. For some $$$, you can go the Curcio regulator route with all the bells and whistles, but that's probably overkill.
As far as bass is concerned, output transformers on their own are quite ugly. Driving them from a high source impedance makes them uglier yet. OTOH, I'm firmly of the opinion that tube amps ought to be restricted to frequencies above 100Hz anyway... and indeed, at 60 Hz, the distortion was quite tolerable up to 1.5 amps or so. Additionally, to keep bass response reasonably flat, you're going to need some heavy-duty EQ or use motional feedback. Better yet, kill both birds with a single pebble and biamp, relieving the poor little tube amp of the onerous duty of scraping the bottom.
Interesting. Thanks for the info, SY.
I was looking at the Curcio input boards already—I guess this is another reason to consider them.
I'd like to end up with adjustable output impedance to tailor the response of wideband drivers. What's the best way to do that? Resistors across the output screw-down terminals?
And that shunt resistor would help keep the bass response in line, right?
Also, do you think it would be possible, with one or a few high-voltage switches, to make my Mark IVs externally switchable between ultralinear and pentode mode?
I was looking at the Curcio input boards already—I guess this is another reason to consider them.
I'd like to end up with adjustable output impedance to tailor the response of wideband drivers. What's the best way to do that? Resistors across the output screw-down terminals?
And that shunt resistor would help keep the bass response in line, right?
Also, do you think it would be possible, with one or a few high-voltage switches, to make my Mark IVs externally switchable between ultralinear and pentode mode?
People have done that sort of HV switching in commercial amps, but it scares me. When I've wanted to go back and forth, I've set it up to move a wire from one spot to another- with power down and caps discharged. Personal prejudice, no doubt.
A parallel resistor will do what you want with regard to adjustable output Z. It's not an efficient way to do it, but hey, tubes aren't efficient.
If a speaker needs low source Z in the bass, a parallel resistor will not get you there. Either low-level eq (ie, before the power amp input) or biamping would be more sensible solutions.
A parallel resistor will do what you want with regard to adjustable output Z. It's not an efficient way to do it, but hey, tubes aren't efficient.
If a speaker needs low source Z in the bass, a parallel resistor will not get you there. Either low-level eq (ie, before the power amp input) or biamping would be more sensible solutions.
Moving a wire sounds simple enough. Beats poking more holes in the collectible chassis, too. 🙂
What I'd like to end up with is an output impedance that's variable between about 10 and 125 ohms. So, worst-case scenario at 40w total output into an 8-ohm load paralleled with 11 ohms across the outputs, The shunt would dissipate maybe 15w, right? Is there a better way?
Since part of my goal for this amp is to experiment with raising the Qes of low-Q drivers, I don't foresee ever needing any less than 10 ohms of output impedance.
I'm still a little unclear about your bass comments a couple posts back--are you saying there will be frequency response anomalies due to OPT behavior unrelated to driver resonance issues?
Thanks,
Dart
What I'd like to end up with is an output impedance that's variable between about 10 and 125 ohms. So, worst-case scenario at 40w total output into an 8-ohm load paralleled with 11 ohms across the outputs, The shunt would dissipate maybe 15w, right? Is there a better way?
Since part of my goal for this amp is to experiment with raising the Qes of low-Q drivers, I don't foresee ever needing any less than 10 ohms of output impedance.
I'm still a little unclear about your bass comments a couple posts back--are you saying there will be frequency response anomalies due to OPT behavior unrelated to driver resonance issues?
Thanks,
Dart
It's not just a matter of frequency response, transformer/output tube combinations distort with real world loads and power demands when asked to handle bass. One can design a transformer that does better in the bass only by giving something up in the treble. That's OK, it's just not my design choice, and the ST-70 transformer fits my criteria.
That 15 watts to the resistor is the majority of power that your amp can put out. If you want to drop to 10 ohms source Z, why not just reconfigure the output stage to triode?
That 15 watts to the resistor is the majority of power that your amp can put out. If you want to drop to 10 ohms source Z, why not just reconfigure the output stage to triode?
I think I'd only rarely need as little as 10 ohms of output Z. For most of my experiments, I imagine I'll need somewhere in the range of 15-40 ohms.
Triode--interesting idea! I hear it's easy to do with these Dynas, but I don't know any specifics. Any idea what kind of output Z I'd see with a Curcio driver board, triode mode, and no feedback?
Triode--interesting idea! I hear it's easy to do with these Dynas, but I don't know any specifics. Any idea what kind of output Z I'd see with a Curcio driver board, triode mode, and no feedback?
I don't have the EL34 triode curves in front of me at the moment, but will look at them later. But as a rule of thumb, source Z open loop for triodes in P-P AB1 is roughly 1-1.4 times the nominal tap impedance, i.e., the 16 ohm tap will give you a source Z of something like 16-20 ohms.
- Status
- Not open for further replies.