I just finished building an amp from salvage parts designed to run triode strapped 6l6 tubes. After not being very impressed with the veiled sound of those tubes, I tried 6v6, 6f6, and 6k6 types. All sounded better, with the 6f6 and 6k6 taking the edge in terms of clarity. Then, I popped in some 6w6 types that I had. Much better. Remembering an old experiment with similar 6DG6GT tubes, I popped them in the amp and . . .WHOA! With a properly selected driver tube (an unknown 6SN7 tube), this is by far the best sounding amp I have ever heard. It is detailed without being clinical, has beautiful tone, a thick midrange, and three dimensional sound staging. The funny thing is that I have only used junk box parts, except for Auricaps that I had from other projects, and was planning on replacing the E caps and trannys with higher quality parts. I don't see the point now. What I don't understand is what it is that makes an amp "magical". That is, in terms of the circuit behavior, what is the defining property or properties? That term gets bandied about easily, but until I heard this amp, I did not know what it meant--instant relaxation and enjoyment of music with no desire to tweak anything.
I used some (common) design features that I have never tried before, like a separate final filter cap for each output tube, separate transformers for the filament and B+ supplies, extensive bypassing with high quality caps, and lifting the ground with well implemented star grounding (dead quiet). I also have very short signal paths from the input jack to the pot to the first tube with twisted pairs of teflon CAT 5 for signal wire. Also, the tubes are running way above spec but nothing is glowing red.
Anyway, I thought I would post the schematic to solicit comments on what might account for the special sound. I wired up the amp without one and reconstructed it today at work, so hopefully I didn't put anything explicitly wrong on it. Thanks for help on my previous few posts.
I used some (common) design features that I have never tried before, like a separate final filter cap for each output tube, separate transformers for the filament and B+ supplies, extensive bypassing with high quality caps, and lifting the ground with well implemented star grounding (dead quiet). I also have very short signal paths from the input jack to the pot to the first tube with twisted pairs of teflon CAT 5 for signal wire. Also, the tubes are running way above spec but nothing is glowing red.
Anyway, I thought I would post the schematic to solicit comments on what might account for the special sound. I wired up the amp without one and reconstructed it today at work, so hopefully I didn't put anything explicitly wrong on it. Thanks for help on my previous few posts.
Attachments
I built the thing on a DAT chassis for a weird look. The gunk on the front, that I am trying desparately to remove, is hot glue to hold on the drive cover. I forgot to connect the LED on the front to the filament supply. Output transformers are little 2 inch tall jobs that I pulled out of a console. They are mounted inside.
Attachments
Sasquatch said:
That's something that's impossible to answer. After all, "magical" isn't something that you can measure. However, there are a couple of points to keep in mind here. It comes as no surpride that the 6L6s didn't work out. Given the likely power supply, (and that 6DG6GTs are really a 6.3V version of the 50L6) 6L6s just were never made to operate properly at such low voltages. As for why that particular VT seems to work so well, this could be a function of having found a very good Q-Point, based on the load the xfmr gives it. Like some other BPPs, its pentode characteristics don't look so good, but it probably has an excellent characteristic when trioded. Other items that help considerably are the fixed bias, careful attention to constructional details, and decent components. Usually, the amps where you would normally find a 50L6 were not intended to really sound good, and so you wouldn't be seeing any AuriCaps used there, let alone decent iron. Those are all details that you can do something about.
As for making "magical" amps, if I knew how to do that I'd be doing it and putting the forum out of business. 😉
2x165, so about 350 unloaded for the PS. However, loaded takes the B+ at the tube to about 185. The primaries of the output iron are about 550 ohms DCR.
I suggest loading up something like AudioRightmark and magic hunting. In my experience different models of tubes, and different brands of the same model, each have their own distortion signature. You may have stumbled on a sweetheart model/op point. That you noted it's driver dependent also makes me wonder if a little second-order cancellation is happening.
6DG6GT/6W6GT
I've found that the 6W6GT and 6DG6GT (essentially identical) are real "sleepers" as audio tubes. I've used them as driver tubes in transformer-coupled single-ended and push-pull amps with good sucess.
I'm not quite sure why these tubes sound so good, but perhaps because these are identical to the highly-mass-produced 50L6GT, except for the heater, the internal construction is very consistent and not pushed too hard to get high transconductance. Both these factors tend to improve linearity. Just a guess. Anyway, luckily, these tubes are still quite available, since they were used in millions of TV sets from the 1950s and 60s.
I've found that the 6W6GT and 6DG6GT (essentially identical) are real "sleepers" as audio tubes. I've used them as driver tubes in transformer-coupled single-ended and push-pull amps with good sucess.
I'm not quite sure why these tubes sound so good, but perhaps because these are identical to the highly-mass-produced 50L6GT, except for the heater, the internal construction is very consistent and not pushed too hard to get high transconductance. Both these factors tend to improve linearity. Just a guess. Anyway, luckily, these tubes are still quite available, since they were used in millions of TV sets from the 1950s and 60s.
I measured the circuit better: Full behavior is -19.5 bias, 185 V plates, and 42 mA draw. I spent last night tube rolling and found that these output tubes were much more important than the driver. "Magic" happens with a few 6sn7s that I have and these 6dg6 tubes but none of the output tubes I have are magical with the current bias and B+. The best combo, though, is with the 6sn7 in the picture. I think it is a Raytheon. However, most of the output tubes I have with that pinout (6l6, 6w6, 6v6, 6f6), sound good and make decent amps. My statements about the 6l6 types being veiled apply more to the older globe types I have. In between those and the present setup are RCA 6l6gc blackplates and nearly as good are GE 6l6GC gray plates. As Miles Prower stated above, the plate voltage is below the design center of the 6L6 types, so I am not giving them a fair shake with this circuit.
The schematic I posted left out a heater lift on the audio tubes of about +70 volts. I don't know if this makes a difference.
I like the idea of sytematic magic chasing. One thing I have noticed is that really good amps always soundstage well. All amps that soundstage well do not sound great, however. This one has natural depth and height that I have never heard. In addition it has a believable tonal mixture, and avoids the slightly thin sound that tends to go along with single ended triodes. If it has any weakness at all, the bass response has a resonance (subjectively measured) at about 90hz, so some bass notes in that region boom a little too much. It's not unpleasant, just innacurate. Probably the unknown transformers. I think I pulled them from an old GE console.
The schematic I posted left out a heater lift on the audio tubes of about +70 volts. I don't know if this makes a difference.
I like the idea of sytematic magic chasing. One thing I have noticed is that really good amps always soundstage well. All amps that soundstage well do not sound great, however. This one has natural depth and height that I have never heard. In addition it has a believable tonal mixture, and avoids the slightly thin sound that tends to go along with single ended triodes. If it has any weakness at all, the bass response has a resonance (subjectively measured) at about 90hz, so some bass notes in that region boom a little too much. It's not unpleasant, just innacurate. Probably the unknown transformers. I think I pulled them from an old GE console.
Sasquatch said:
I find the same thing. If you're up for an unconventional recommendation, try movie DVDs which contain large amounts of surround information. Of course over two channels those surround effects get mixed back into the left and right channels. However they get mixed back with all the DSP jiggery-pokery Dolby and the rest use to enhance perception of above and behind localization. My system - when things are clicking - can reconstruct some of that and provide imaging above, in front and well to the side and forward of the only two speakers. I find this material extremely helpful in voicing whatever I'm working on. (Voices on movie soundtracks are also much, much less processed than typical of music production, also very helpful.)
Sasquatch said:
What you mention here is a consequence of not adding NFB. For whatever reason, NFB contracts the sound stage. The 807 amp I did had a wide open sound stage running pure open loop. Of course, there was also lots of pentode nastiness and woofer resonance. Including local and global feedback fixed that at the expense of a more contracted sound stage. An equitable design trade-off anyway.
If you don't add NFB, you're probably going to have woofer resonance problems since you have no control over speaker damping, and have to take what the OPTs'll give you, and hope it's good enough.
Miles Prower said:
It's good to hear that confirmed by someone with technical chops, I agree completely. I haven't discounted the possibility a little harmonic enhancement helps the ear lock on to subtle location cues.
rdf said:
No one really knows. What is known, however, as Crowherst demonstrated, is that NFB will clean up gross nonlinearities, but it also tends to fill the noise floor right at the threshold of audibility with uncorrolated harmonic distortion. The process of analog-to-digital conversion does the same thing (quantizing noise).
There is a theory out there that the information content right at the noise floor is necessary for determining sound direction and distance. Since this tends to be wiped out by quantizing noise and NFB, the theory goes that this is responsible for the negative reactions to both CDs and amps that use heavy gNFB. Not known if that's correct.
Although the more NFB you add, the narrower the sound stage becomes. You have to decide whether or not that's an acceptable design trade-off.
Hi Miles, I'm still waiting on the iron for that 1625 amp, it's been 3 weeks and no show yet, but I did build a Fender guitar amp clone in the meantime.
Anyway, I see you have a lot of know-how on the feedback issues, on the guitar amps they often have a "presence" control, which seems to me to be, in effect, a direct control of the NFB.
When completely turned down is seems to ground out the NFB, at least from my interpretation of the schematic.
It only affects one of the cathodes of the long-tailed phase splitter.
Can this type of control work for hi-fi units as well effectively?
Here's one analysis of that circuit (off the web):
At high frequencies with the presence control at maximum the input impedance gets as low as 1.9 megohms under open-loop conditions, which is still much higher than the output impedance of the tone stack. When the presence control is at minimum the input impedance increases to more than 2.7 megohms. Negative feedback increases these values even more. Based on the DC operating point we estimate the triode amplification factors to be 101 and the plate resistances to be 57.7k. The voltage gain for the inverted output (connected to the 82k plate resistor) is then -21.9, taking into account the additional load of the push-pull power amplifier input. For the in-phase output the gain is +22.6. Feedback from the output transformer is connected to the phase splitter via a 27k resistor. The input impedance presented to the feedback signal is 31k at minimum presence and 27k at maximum presence for high frequencies. The voltage gains for the feedback input are +3.4 and -3.5 to the inverted and in-phase outputs.
Anyway, I see you have a lot of know-how on the feedback issues, on the guitar amps they often have a "presence" control, which seems to me to be, in effect, a direct control of the NFB.
When completely turned down is seems to ground out the NFB, at least from my interpretation of the schematic.
It only affects one of the cathodes of the long-tailed phase splitter.
Can this type of control work for hi-fi units as well effectively?
Here's one analysis of that circuit (off the web):
At high frequencies with the presence control at maximum the input impedance gets as low as 1.9 megohms under open-loop conditions, which is still much higher than the output impedance of the tone stack. When the presence control is at minimum the input impedance increases to more than 2.7 megohms. Negative feedback increases these values even more. Based on the DC operating point we estimate the triode amplification factors to be 101 and the plate resistances to be 57.7k. The voltage gain for the inverted output (connected to the 82k plate resistor) is then -21.9, taking into account the additional load of the push-pull power amplifier input. For the in-phase output the gain is +22.6. Feedback from the output transformer is connected to the phase splitter via a 27k resistor. The input impedance presented to the feedback signal is 31k at minimum presence and 27k at maximum presence for high frequencies. The voltage gains for the feedback input are +3.4 and -3.5 to the inverted and in-phase outputs.
frank754 said:
It's usually a bit more complex than that, since the idea behind the "presence" control is to manipulate the frequency response at the upper end, as opposed to varying the NFB at all frequencies. I've seen a few Hi-Fi units that included this.
I must agree with Miles & RDF and add that local feedback does the same contraction of spatial cues (to my ears at least), sometimes even harder so than GNFB, when its about some strong degeneration that you can't actively moderate having to forgo good operation points and the like instead.
- Status
- Not open for further replies.