ampage Tube Amps / Music Electronics	For current discussions, please visit Music Electronics Forum.

ampage archive

Vintage threads from the first ten years

Preliminary results of experiments

10/24/1999 7:55 PM
Ken Gilbert	Preliminary results of experiments Well I bit the bullet and ripped out everything from the preamp, which resides in an old Marshall head chassis. Wiped it clean--nothing was left except the iron, and the two LCR can caps above the deck. Here's a list, in no particular order. Changed all the sockets to ceramic. Finally pulled out that butt-nasty silly phenolic sockets that came stock on the Marshall JCM900's. They had no spring tension, and flimsy lugs on the backside. This required a bit of socket hole punching (thank God for the Greenlee) and some filing, but it was well worth it. All sockets are from New Sensor, and have gold-plated lugs. They seem to stay pretty clean, and they aren't that much more expensive. Added a common mode choke to the power supply. This turned out very interesting indeed. I put a small toroidal core CMC right after the diode bridge. Both + and - leads go through the choke, which has two separate windings on the same core. This CMC was scavenged from a computer power supply, and had very low DCR. Inductance was pretty low, about 150 mH per winding. With a scope probe, looking at both the + and - leads from the bridge BEFORE the CMC showed quite a bit of hash and other shit from the recs--and these has snubbers attached at the time! I would hate to have seen it without snubbers. It probably would have been in the tens of volts. As it was, it was under a volt or so, but it was particularly nasty looking--way up there in frequency. AFTER the CMC, the rails were DEAD FLAT. It was unbelievable. Totally suggested. Snubber networks across all B+ diodes. Nothing more than 150R and 1nF in series across diodes, with R's towards cathodes of 1N4007's. I did not compare before and after, although I KNOW they cannot hurt and are likely doing quite a bit of good. Switched to a discreet darlington connected heater series pass regulator. This enabled me to increase the time constant of the base capacitance, which brought ripple on the heater rail down to 2 mV. I needed more voltage headroom because of the double diode drops through the transistors, so I did the following: Switched over to Schottky diodes in the DC heater supply. These work cooler, quieter, and with less voltage drop than regular PN diodes. I did not need snubbers here for that reason. 'Nuff said. Implimented a cascode input stage using the Svetlana 6N1P duo-triode. This was by far the most exciting mod yet. B+ is about 350V, the two tubes are stacked on top of one another in cascode fashion. In this way the bottom tube is operating into a rather low impedance, but in CURRENT MODE, since it's plate voltage (which is the top tube's cathode) does not change. Rl was set to 100K, and the voltage on the top tube's grid was set to about 100VDC through a resistive divider from B+. This floated the cathode a few volts higher at 105 to 110VDC. (NOTE: this exceeds data sheet rec's on Vhk. Final judgement has not been made as to reliability concerns. They're cheap, though, so I'm letting fly.) The bottom tube's cathode was connected directly to the input jack ground, which was then grounded to the buss. Rk was 1K, chosen somewhat arbitrarily. Grid had a 68K stopper attached RIGHT AT THE LUG, and went straight to the input jack hot. From there the grid leak R of 500K was grounded to the input jack ground as well. Top plate sat a little lower than 200VDC--a little low, but reasonably centered around B+. Sonic impressions? Shit, I can't really begin to explain. Let me tell you something I was amazed at: this single stage--when hooked up to a small power amp--provided a line level output that was absolutely BEAUTIFUL in tone. It was very clean but it had unbelieveable sparkle and tone. Output coupling cap was a 2n7. Honestly, I have never heard a clean quite like this clean before. It was incredibly seductive. In fact, I use this unfettered output with NO OTHER TONE CONTROLS as one of the channels of the amp... It's that good. To figure on a gain number, we must use the equation of A = Rl * gm. With these high transconductance (7500 mS or so) 6N1P's, and a 100K Rl, that works out to be really fucking HIGH. Now, I must say the tube is loafing along at only 1-2 mA, so the gm will be a bit lower--say 5000 mS at these plate currents. Still, that's .005 mA/V * 100,000 ohms or a voltage gain of _500_ from one stage. The output impedance is about twice the plate resistance of the tube, which in this case is 2 * 5K or so, or 10K total. This output Z is still far far lower than a 12AX7, and orders of magnitude lower than a similar pentode stage's (like an EF86) output Z. And even an EF86 would max out at voltage gains of 250 or so... nowhere near what I'm getting out of this stage. It's almost too good to be true. I spent very little time optimizing the balance of the cascode's voltages. For hifi use, you would want to. When I looked at the output with a scope I noticed a very smooth even harmonic distortion being added--no doubt to the lack of voltage headroom symmetry. This took nothing away from the sound... in fact, I think that's what made it soooo sexy. I had at least 100Vpk signal output from this single stage. The only shunt capacitance on the input grid is 4.8 pF, and there is NO Miller effect. Admittedly the very great HF performance of this stage absolutely mandates a decent scope to seek out ultrasonic oscillations, at least if there's any more gain to be added on. Some pics and other info on other mods will follow shortly. I need to take a little break for a while... too much thinking lately. I'm FRIED. KG

10/24/1999 7:55 PM

Ken Gilbert

Preliminary results of experiments

Well I bit the bullet and ripped out everything from the preamp, which resides in an old Marshall head chassis. Wiped it clean--nothing was left except the iron, and the two LCR can caps above the deck. Here's a list, in no particular order.

Changed all the sockets to ceramic.

Finally pulled out that butt-nasty silly phenolic sockets that came stock on the Marshall JCM900's. They had no spring tension, and flimsy lugs on the backside. This required a bit of socket hole punching (thank God for the Greenlee) and some filing, but it was well worth it. All sockets are from New Sensor, and have gold-plated lugs. They seem to stay pretty clean, and they aren't that much more expensive.

Added a common mode choke to the power supply.

This turned out very interesting indeed. I put a small toroidal core CMC right after the diode bridge. Both + and - leads go through the choke, which has two separate windings on the same core. This CMC was scavenged from a computer power supply, and had very low DCR. Inductance was pretty low, about 150 mH per winding. With a scope probe, looking at both the + and - leads from the bridge BEFORE the CMC showed quite a bit of hash and other shit from the recs--and these has snubbers attached at the time! I would hate to have seen it without snubbers. It probably would have been in the tens of volts. As it was, it was under a volt or so, but it was particularly nasty looking--way up there in frequency. AFTER the CMC, the rails were DEAD FLAT. It was unbelievable. Totally suggested.

Snubber networks across all B+ diodes.

Nothing more than 150R and 1nF in series across diodes, with R's towards cathodes of 1N4007's. I did not compare before and after, although I KNOW they cannot hurt and are likely doing quite a bit of good.

Switched to a discreet darlington connected heater series pass regulator.

This enabled me to increase the time constant of the base capacitance, which brought ripple on the heater rail down to 2 mV. I needed more voltage headroom because of the double diode drops through the transistors, so I did the following:

Switched over to Schottky diodes in the DC heater supply.

These work cooler, quieter, and with less voltage drop than regular PN diodes. I did not need snubbers here for that reason. 'Nuff said.

Implimented a cascode input stage using the Svetlana 6N1P duo-triode.

This was by far the most exciting mod yet. B+ is about 350V, the two tubes are stacked on top of one another in cascode fashion. In this way the bottom tube is operating into a rather low impedance, but in CURRENT MODE, since it's plate voltage (which is the top tube's cathode) does not change. Rl was set to 100K, and the voltage on the top tube's grid was set to about 100VDC through a resistive divider from B+. This floated the cathode a few volts higher at 105 to 110VDC. (NOTE: this exceeds data sheet rec's on Vhk. Final judgement has not been made as to reliability concerns. They're cheap, though, so I'm letting fly.) The bottom tube's cathode was connected directly to the input jack ground, which was then grounded to the buss. Rk was 1K, chosen somewhat arbitrarily. Grid had a 68K stopper attached RIGHT AT THE LUG, and went straight to the input jack hot. From there the grid leak R of 500K was grounded to the input jack ground as well. Top plate sat a little lower than 200VDC--a little low, but reasonably centered around B+.

Sonic impressions? Shit, I can't really begin to explain. Let me tell you something I was amazed at: this single stage--when hooked up to a small power amp--provided a line level output that was absolutely BEAUTIFUL in tone. It was very clean but it had unbelieveable sparkle and tone. Output coupling cap was a 2n7. Honestly, I have never heard a clean quite like this clean before. It was incredibly seductive. In fact, I use this unfettered output with NO OTHER TONE CONTROLS as one of the channels of the amp... It's that good.

To figure on a gain number, we must use the equation of A = Rl * gm. With these high transconductance (7500 mS or so) 6N1P's, and a 100K Rl, that works out to be really fucking HIGH. Now, I must say the tube is loafing along at only 1-2 mA, so the gm will be a bit lower--say 5000 mS at these plate currents. Still, that's .005 mA/V * 100,000 ohms or a voltage gain of _500_ from one stage. The output impedance is about twice the plate resistance of the tube, which in this case is 2 * 5K or so, or 10K total. This output Z is still far far lower than a 12AX7, and orders of magnitude lower than a similar pentode stage's (like an EF86) output Z. And even an EF86 would max out at voltage gains of 250 or so... nowhere near what I'm getting out of this stage. It's almost too good to be true.

I spent very little time optimizing the balance of the cascode's voltages. For hifi use, you would want to. When I looked at the output with a scope I noticed a very smooth even harmonic distortion being added--no doubt to the lack of voltage headroom symmetry. This took nothing away from the sound... in fact, I think that's what made it soooo sexy.

I had at least 100Vpk signal output from this single stage. The only shunt capacitance on the input grid is 4.8 pF, and there is NO Miller effect. Admittedly the very great HF performance of this stage absolutely mandates a decent scope to seek out ultrasonic oscillations, at least if there's any more gain to be added on.

Some pics and other info on other mods will follow shortly. I need to take a little break for a while... too much thinking lately. I'm FRIED.

KG

Replies:

Randall Aiken Finally! Someone who agrees with m... -- 10/24/1999 8:57 PM
Marc Ken sounds cool. I ... -- 10/25/1999 2:38 AM
Fuzzman q{A close up of a new design produc... -- 10/26/1999 2:54 PM
R.G. Wow, Ken. Very impressive results. ... -- 10/26/1999 3:37 PM