Wow.
OK, let's get back to reality. The DAC is putting out 2V. That's close to enough to clip any reasonable power amp. What you really want, with your gain structure, is a unity gain buffer with the volume control ahead of it. This will give you far lower distortion and output impedance than that... ummm.... "unusual" suggestion. At most, you might want a gain of 2-4.
Now a different approach is to have all the voltage gain on the preamp end, then have the power amp be closer to unity gain. Nelson Pass has pioneered that approach. This means a wholesale rebuilding of the system, but there could be some advantages if there's a long cable run between pre and power amps.
Back to the topic, the gain of an LTP with one input grounded is something less than 1/2 mu at each plate. The formulas are in VA, along with their derivations.
Yeah, I'm a "hobbyist." What can I say? 😱
So I have the basic idea for how voltage gain works from one stage into the next?
I suppose that means a cathode follower. In that case, I'd better buy me some DN2540's... As for the tube to be used, the higher the gm, the better?
I've read that part several times, but it seems it hasn't sunk in yet. I remember thinking that the mu of the tube used in the LTP is applied only to its half of the push-pull signal, but was unsure if I understood correctly... Makes sense then that the tube's mu would be approximately halved if you ground the grid of one side.
--
I was thinking about what I've picked up from this conversation...
I'm still struggling with the math.
So I'm driving a 2A3 with a 6N30P:
The bias on the 2A3 is -42V. I need to reach 42Vpk input signal on the 2A3 grid to drive it to clipping.
6N30P mu is about 14.
42/14 = 3
Does that mean 3Vpk input to the 6N30P should drive the 2A3 to full power/clipping? If yes, then that equals 2.12Vrms input to 6N30P.
Now let's say I want 6dB of extra gain to account for weaker sources than the CD player that might need a boost...
Add an extra 2Vrms of input, for a total of 4Vrms, or 5.6Vpk required at the 6N30P input.
Now I need a gain of 2 from somewhere. That would be the line stage. Of course, this is exactly the problem that MJ describes in his discussion of line stages in "Valve Amplifiers" where he explains why line preamps are undesirable. What triode gives you a mu < 5 with low input C, low rp and low distortion? None that I know of.
But a buffer (cathode follower) has gain of ~1. If I use a CF, I won't have any extra boost should I need it.
What would you think of a 5687 LTP with the output taken from one side only (mu of about 7)? Would the input C be too high? Distortion would be low, would it not? That might be Plan A.
How's this for a Plan B...
I'm going to put together a 5-position input switch and Noble 100k stereo pot in a box ("passive preamp") and use that as a module for experimenting. Then I'll build a cathode follower and plug the passive pre into that. If it floats my boat, then that's the solution.
Sound like a reasonable plan?
--
I'm still struggling with the math.
So I'm driving a 2A3 with a 6N30P:
The bias on the 2A3 is -42V. I need to reach 42Vpk input signal on the 2A3 grid to drive it to clipping.
6N30P mu is about 14.
42/14 = 3
Does that mean 3Vpk input to the 6N30P should drive the 2A3 to full power/clipping? If yes, then that equals 2.12Vrms input to 6N30P.
Now let's say I want 6dB of extra gain to account for weaker sources than the CD player that might need a boost...
Add an extra 2Vrms of input, for a total of 4Vrms, or 5.6Vpk required at the 6N30P input.
Now I need a gain of 2 from somewhere. That would be the line stage. Of course, this is exactly the problem that MJ describes in his discussion of line stages in "Valve Amplifiers" where he explains why line preamps are undesirable. What triode gives you a mu < 5 with low input C, low rp and low distortion? None that I know of.
But a buffer (cathode follower) has gain of ~1. If I use a CF, I won't have any extra boost should I need it.
What would you think of a 5687 LTP with the output taken from one side only (mu of about 7)? Would the input C be too high? Distortion would be low, would it not? That might be Plan A.
How's this for a Plan B...
I'm going to put together a 5-position input switch and Noble 100k stereo pot in a box ("passive preamp") and use that as a module for experimenting. Then I'll build a cathode follower and plug the passive pre into that. If it floats my boat, then that's the solution.
Sound like a reasonable plan?
--
More thinking...
A 5687 or 6SN7 CF can take an awful lot of signal at its input before overload or grid current. Way more than the 2Vrms it's likely to see. Why wouldn't it be a good idea to put the volume pot on the output of the CF as opposed to the input?
My (probably muddled) thinking on this is that the output Z of the sources (FM tuner, CD player and phono preamp) are relatively high, and would be driving the selector switch and pot before reaching the input of the CF. Is there a chance that the sources could be loaded down by the switch and pot? The input switch is bad enough by itself. Wouldn't it be a good idea to move the pot away from the outputs of the audio sources?
A CF has very high input Z, so should present a much easier load to the sources. It also has a very low output Z, so the CF should be able to send the audio signal through a potentiometer with less losses.
Perhaps the solution is a ladder attenuator instead of a pot.
At any rate, if I build a separate little switch/pot box, I can put the CF before or after the box, and compare results. But I'd like to know if there is a good engineering reason why the pot belongs at the input of the CF buffer stage.
Thanks again...
A 5687 or 6SN7 CF can take an awful lot of signal at its input before overload or grid current. Way more than the 2Vrms it's likely to see. Why wouldn't it be a good idea to put the volume pot on the output of the CF as opposed to the input?
My (probably muddled) thinking on this is that the output Z of the sources (FM tuner, CD player and phono preamp) are relatively high, and would be driving the selector switch and pot before reaching the input of the CF. Is there a chance that the sources could be loaded down by the switch and pot? The input switch is bad enough by itself. Wouldn't it be a good idea to move the pot away from the outputs of the audio sources?
A CF has very high input Z, so should present a much easier load to the sources. It also has a very low output Z, so the CF should be able to send the audio signal through a potentiometer with less losses.
Perhaps the solution is a ladder attenuator instead of a pot.
At any rate, if I build a separate little switch/pot box, I can put the CF before or after the box, and compare results. But I'd like to know if there is a good engineering reason why the pot belongs at the input of the CF buffer stage.
Thanks again...
High is a relative term.
The output of standard FM tuners, CD players, Phono Preamps, etc expect to see a load of 47K ohms or more recently 10K ohms. Those are high values compared to power amp output impedance.
However compared to tube input impedances they are very low. The output impedance to drive these inputs should be 10X lower, so in most cases 1K output impedance or less is strived for.
Now your typical tube volume control is in the range of 100K to 1M. I've been using the Radio shack 100K dual pot.
Drive the 1K output into the 100K pot and you have no problems with loading.
Now move the pot to the output tube as you propose. You need a low impedance pot to drive your next stage input, and you want it to be 10X the output impedance of the drving stage, so 10K would do. As you turn down the volume you increase the output impedance by the resistance value, so your less than 1K output is now going up to as much as 11K in the full off position. Capacitance on the input of the next stage combines with the output impedance of yoru driver stage and changes frequency response with volume setting.
The output of standard FM tuners, CD players, Phono Preamps, etc expect to see a load of 47K ohms or more recently 10K ohms. Those are high values compared to power amp output impedance.
However compared to tube input impedances they are very low. The output impedance to drive these inputs should be 10X lower, so in most cases 1K output impedance or less is strived for.
Now your typical tube volume control is in the range of 100K to 1M. I've been using the Radio shack 100K dual pot.
Drive the 1K output into the 100K pot and you have no problems with loading.
Now move the pot to the output tube as you propose. You need a low impedance pot to drive your next stage input, and you want it to be 10X the output impedance of the drving stage, so 10K would do. As you turn down the volume you increase the output impedance by the resistance value, so your less than 1K output is now going up to as much as 11K in the full off position. Capacitance on the input of the next stage combines with the output impedance of yoru driver stage and changes frequency response with volume setting.
This approach, though sound in theory, has a non-obvious caveat - the emitter resistor in the mirror 'CCS' needs to have as low as practical value, in order not to take up the available voltage - with monolitic double transistors you might even get away with no resistor at all, using only the emitter bulk resistance.
If you do this, expect an unexpectedly high noise level! The problem is, that the mirror transistor in the 'CCS' forms a cascode with the triode, and since the gm of a BJT is orders of magnitude larger than that of the tube, the whole circuit will amplify any noise at the mirror transistor's base, including that of the transistor itself, roughly Ra/Re times, where Ra is the plate resistor of the triode, and Re is the (bulk) emitter resistance of the 'CCS' transitor. In practice, this may well be on the order of 1000 - while the amplification factor of the whole amplifier may well be less than 10!
The simplest solution is to use a negative rail. This gives you a wide choice of Re versus early effect, whatever type of CCS is used. Otherwise, expect a battle to find the compromise solution, where neither the noise nor the early effect will be optimal. paralleling multiple low nosie transitors may help by alowing each transistor to pass a current better suited for a lower noise figure.
I've never heard of noisy current mirrors. Where do you propose this noise comes from? It certainly isn't the Johnson noise of the effective emitter resistance, which is a few ohms.
If noise were as serious a problem as you seem to suggest, there would be an awful lot of low-noise integrated circuits that would never work.
Tim
If noise were as serious a problem as you seem to suggest, there would be an awful lot of low-noise integrated circuits that would never work.
Tim
Higher distortion (the gain stage is always running wide open and distortion is roughly proportional to stage output level), and higher/variable source impedance (with a perfect zero ohm driving source, the impedance would be as high as 1/4 the potentiometer resistance and would vary with control setting). Other than that, how did you enjoy the play, Mrs. Lincoln? 😀
The problem is not alarming in most cases, especially if you are using differential outputs, the noise would be common mode.
What I was pointing out was that a voltage noise source presented to the base of the current mirror might get amplified quite a lot because of the Ra/Re ratio.
If I remember correctly, a few years ago someone made a case for using higher reference voltages for standard single transistor current sources, where current is (Vref-Vbe)/Re, to exploit a higher Re in order to increase the CCS impedance, and to reduce the amplification factor with respect to the reference source, this is just another example of the same principle.
Hey y'all...
Well, I cleaned up my line preamp over the weekend.
1. Moved volume pot (50k log, it's all I have at the moment) to after the selector switch, to the grid circuit of the 5687 triode (the usual place).
2. Added a 100nF mylar cap from B+ to ground, right at the tube.
I hate to say it, but... What was I thinking????
I obviously had the volume control located in the circuit bass-ackwards. I can definitely hear the reduction in overload on transients. I'm sure program peaks were driving the 5687 into grid current. Most immediately noticeable improvement is on orchestral strings (believe it or not). Sound is generally much clearer now.
But of course I have a question...
I fixed a different line amp I built a while ago (put the pot in the correct place). This preamp only has a +255V B+, so my choices for the size of the plate load resistor are limited.
I set it up for a 6DJ8 with a 21k ohm plate resistor, 2.6V cathode bias. That yields Ea = 100V, Ia = 7.5mA (Ek = 2.5V). Should be a decent place for a 6DJ8. But I find I don't like the sound. It sounds like it's making too much 2nd harmonic distortion (rich and thick, too "tubey" sounding).
So I tried a 6GU7 in its place. That I like. Its ra should be about 5.5k ohms, so should work fine with the 22k ohm plate load. Ia is about 7.5mA, pretty much the same as the 6DJ8. But this tube sounds completely different. Much less gain, of course, but definitely clearer than the 6DJ8.
Next I tried a 6CG7, just because I have a few. I like this one best of all. But it's totally wrong. It biases up at Ia = 5.5mA, so its ra = 8k ohm. That ra is way too high to work right into a 21k ohm plate load. Yet the 6CG7 sounds the best to me, extremely clear, like it has *less* distortion, not more.
I was reading the discussion of line level preamps in "Valve Amplifiers" last night, and came upon a passage where Morgan briefly describes running a common cathode triode with lower value plate load to reduce both input capacitance and gain. Could this be what I'm hearing with the 6CG7? Or do you think I'm hearing extra 3rd harmonic distortion, which I'm mistaking for "detail"?
Thanks again.
--
Well, I cleaned up my line preamp over the weekend.
1. Moved volume pot (50k log, it's all I have at the moment) to after the selector switch, to the grid circuit of the 5687 triode (the usual place).
2. Added a 100nF mylar cap from B+ to ground, right at the tube.
I hate to say it, but... What was I thinking????
I obviously had the volume control located in the circuit bass-ackwards. I can definitely hear the reduction in overload on transients. I'm sure program peaks were driving the 5687 into grid current. Most immediately noticeable improvement is on orchestral strings (believe it or not). Sound is generally much clearer now.
But of course I have a question...
I fixed a different line amp I built a while ago (put the pot in the correct place). This preamp only has a +255V B+, so my choices for the size of the plate load resistor are limited.
I set it up for a 6DJ8 with a 21k ohm plate resistor, 2.6V cathode bias. That yields Ea = 100V, Ia = 7.5mA (Ek = 2.5V). Should be a decent place for a 6DJ8. But I find I don't like the sound. It sounds like it's making too much 2nd harmonic distortion (rich and thick, too "tubey" sounding).
So I tried a 6GU7 in its place. That I like. Its ra should be about 5.5k ohms, so should work fine with the 22k ohm plate load. Ia is about 7.5mA, pretty much the same as the 6DJ8. But this tube sounds completely different. Much less gain, of course, but definitely clearer than the 6DJ8.
Next I tried a 6CG7, just because I have a few. I like this one best of all. But it's totally wrong. It biases up at Ia = 5.5mA, so its ra = 8k ohm. That ra is way too high to work right into a 21k ohm plate load. Yet the 6CG7 sounds the best to me, extremely clear, like it has *less* distortion, not more.
I was reading the discussion of line level preamps in "Valve Amplifiers" last night, and came upon a passage where Morgan briefly describes running a common cathode triode with lower value plate load to reduce both input capacitance and gain. Could this be what I'm hearing with the 6CG7? Or do you think I'm hearing extra 3rd harmonic distortion, which I'm mistaking for "detail"?
Thanks again.
--
Since a line amp only has to swing a few volts, you can make up for the low B+ by using a CCS load for the plate. That drops the distortion to its lowest possible value (I think Morgan called it "irreducible distortion.").
Thanks SY, that is on my to-do list for this preamp. Based on my experience this weekend, I think I want to use the 6CG7, or better yet, change the whole thing to a 6SN7.
In the meantime, and in the interest of educating myself about the sound of low distortion vs. higher distortion, I was wondering if anyone had noticed something like I encountered...
- Status
- Not open for further replies.