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

ampage archive

Vintage threads from the first ten years

preamp design ruminations

6/29/1999 9:12 PM
ken gilbert	preamp design ruminations Thinking hard about preamp distortion lately, and haveing some success on that front as well... I heard from an old guru friend of mine, a while back, that he didn't see the wisdom of low current stages, ie. like somthing you'd see in Soldano's designs, with an Rk of something like 25K or so, and a normal Rl. He didn't appreciate the usefulness of a stage that didn't add any voltage gain, and to him, if the tube was off, it wasn't worth putting in the amp at all. I have begun to see some wisdom in the practice. I myself am a proponent of smaller coupling caps, especially in the dirty channel, of something like 3 nF or so. This is way off the normal practice of 22 nF or more. I found this practice to reduce the effects of grid blocking on the following stage, since the time constant formed by the grid leak resistor and C was shorter. Along with that, I definitely heard the effect of cascading a few stages--maybe even up to 4. Although you'll get PLENTY of voltage gain from two complete 12AX7's, there's also the effect of the clipping provided by running into either a) the onset of grid current or b) the plate slamming into the Ebb rail. By cascading stages you end up with a more drastic clipping characteristic, and this gets that "saturated" sound. Then something occured to me about cathode degenerative amplifiers (which is any cathode biased stage with incomplete cathode bypassing, or any stage which has a signal dependant voltage on it's cathode)... There's almost no chance of grid blocking. In the extreme case, you've got very low current stages, like with a 10K Rk, and no bypass cap. K floats at say 7VDC or so, G is at OVDC. Now, with a Ck of say 25 mics, that cathode voltage is going to STAY at 7VDC (the break frequency is under 1Hz), no matter where the grid goes. Suddenly, Joe Guitar player slams the strings. The grid gets a swing of 20Vpp, goes positive, up to the 7VDC limit, and the tube begins to attempt to suck current into the grid. We've got the classic beginnings of grid blocking going on. The grid circuit is high impedance, can't supply the current, so instead gets negatively charged. The coupling C will retain this charge for some finite time, and the output during this time will sound like farting asses. Now, the normal design practice is to limit the grid drive voltage, usually with a voltage divider. I hate that--I hate throwing away gain when I don't have to, and I ESPECIALLY hate shunting stuff to ground, like the "boogie style" HF rolloff oscillation preventers. So, instead, I pop off the Ck. Now, the DC condition of the tube is completely unchanged. But when that same 20Vpp signal flies down the grid, the grid goes up in voltage, AND SO DOES THE CATHODE. It is very clear and easy to see here how the simple cathode degeneration can limit (if not irradicate) grid current flow, depending on the size of the Rk. It does this instantaneously, since there are no time constants or capacitances to charge/discharge. To make a long story short, I made great headway with my third channel, the "ultragain" channel, by not forming heavily attenuating voltage dividers between stages cranked for maximum gain, but by adding cathode resistance, and eliminating cathode bypass caps. In this way I was able to eliminate series resistances, essentially keeping the design "lower impedance," which also helped with hum and noise. I also used lower than normal grid leak resistors to keep the TC's down, along with my traditionally small coupling caps. The response to transients was FAST and CLEAN, and it had so much gain it was ridiculous. There were two cascaded stages with high Rk's, one 10K and one 22K. Both were pots for easy tweaking. Anyway, just sharing some empirical evidence with proposed theory of operations. I found it interesting since it is so antithecal to the traditional Fender heavily bypassed, heavily voltage divided design. It sure don't sound like a Fender. ~KG~

6/29/1999 9:12 PM

ken gilbert

preamp design ruminations

Thinking hard about preamp distortion lately, and haveing some success on that front as well...

I heard from an old guru friend of mine, a while back, that he didn't see the wisdom of low current stages, ie. like somthing you'd see in Soldano's designs, with an Rk of something like 25K or so, and a normal Rl. He didn't appreciate the usefulness of a stage that didn't add any voltage gain, and to him, if the tube was off, it wasn't worth putting in the amp at all.

I have begun to see some wisdom in the practice. I myself am a proponent of smaller coupling caps, especially in the dirty channel, of something like 3 nF or so. This is way off the normal practice of 22 nF or more. I found this practice to reduce the effects of grid blocking on the following stage, since the time constant formed by the grid leak resistor and C was shorter.

Along with that, I definitely heard the effect of cascading a few stages--maybe even up to 4. Although you'll get PLENTY of voltage gain from two complete 12AX7's, there's also the effect of the clipping provided by running into either a) the onset of grid current or b) the plate slamming into the Ebb rail. By cascading stages you end up with a more drastic clipping characteristic, and this gets that "saturated" sound.

Then something occured to me about cathode degenerative amplifiers (which is any cathode biased stage with incomplete cathode bypassing, or any stage which has a signal dependant voltage on it's cathode)... There's almost no chance of grid blocking.

In the extreme case, you've got very low current stages, like with a 10K Rk, and no bypass cap. K floats at say 7VDC or so, G is at OVDC. Now, with a Ck of say 25 mics, that cathode voltage is going to STAY at 7VDC (the break frequency is under 1Hz), no matter where the grid goes.

Suddenly, Joe Guitar player slams the strings. The grid gets a swing of 20Vpp, goes positive, up to the 7VDC limit, and the tube begins to attempt to suck current into the grid. We've got the classic beginnings of grid blocking going on. The grid circuit is high impedance, can't supply the current, so instead gets negatively charged. The coupling C will retain this charge for some finite time, and the output during this time will sound like farting asses.

Now, the normal design practice is to limit the grid drive voltage, usually with a voltage divider. I hate that--I hate throwing away gain when I don't have to, and I ESPECIALLY hate shunting stuff to ground, like the "boogie style" HF rolloff oscillation preventers.

So, instead, I pop off the Ck. Now, the DC condition of the tube is completely unchanged. But when that same 20Vpp signal flies down the grid, the grid goes up in voltage, AND SO DOES THE CATHODE. It is very clear and easy to see here how the simple cathode degeneration can limit (if not irradicate) grid current flow, depending on the size of the Rk. It does this instantaneously, since there are no time constants or capacitances to charge/discharge.

To make a long story short, I made great headway with my third channel, the "ultragain" channel, by not forming heavily attenuating voltage dividers between stages cranked for maximum gain, but by adding cathode resistance, and eliminating cathode bypass caps. In this way I was able to eliminate series resistances, essentially keeping the design "lower impedance," which also helped with hum and noise. I also used lower than normal grid leak resistors to keep the TC's down, along with my traditionally small coupling caps. The response to transients was FAST and CLEAN, and it had so much gain it was ridiculous. There were two cascaded stages with high Rk's, one 10K and one 22K. Both were pots for easy tweaking.

Anyway, just sharing some empirical evidence with proposed theory of operations. I found it interesting since it is so antithecal to the traditional Fender heavily bypassed, heavily voltage divided design. It sure don't sound like a Fender.

~KG~

6/29/1999 11:31 PM
R.G.	With reference to our friend Joe Guitar and his trademark string slam... I don't have a good tube testbed set up right now, or I'd go hack this one in and try it. What happens if you put a semiconductor diode with a large resistor parallel to it in series with the grid of the tube that's being pummeled? Can the diode block the grid current flow without blenderizing the sound? I can think of arguments for and against it, and that's where I usually drift off to the workbench in the garage...

6/29/1999 11:31 PM

R.G.

With reference to our friend Joe Guitar and his trademark string slam...

I don't have a good tube testbed set up right now, or I'd go hack this one in and try it. What happens if you put a semiconductor diode with a large resistor parallel to it in series with the grid of the tube that's being pummeled? Can the diode block the grid current flow without blenderizing the sound? I can think of arguments for and against it, and that's where I usually drift off to the workbench in the garage...

6/30/1999 7:04 AM
Jamie Tyson	Wow! I never thought of that but it's a cool concept. I love the 4-knob tube driver that I have but it is notorious for its farting sound, especially when using the neck pickup. I tried changing coupling cap values to reduce the blocking but it did'nt stop. This may be the solution!! Actually- I'll have to take a closer look at the Tube Driver- the cathode may be tied to ground. Thanks- I'll let you know. JT

6/30/1999 7:04 AM

Jamie Tyson

Wow! I never thought of that but it's a cool concept. I love the 4-knob tube driver that I have but it is notorious for its farting sound, especially when using the neck pickup. I tried changing coupling cap values to reduce the blocking but it did'nt stop. This may be the solution!! Actually- I'll have to take a closer look at the Tube Driver- the cathode may be tied to ground.

Thanks- I'll let you know.

JT

6/30/1999 6:26 PM
Craig	I have heard that one of the primary flaws in the Tube Driver design is its use of very low B+ supply, which does not allow the tube to operate within its optimal range, and hence the fartiness.

7/1/1999 11:43 AM
GFR	The tube driver requires grid current to bias. Altough they patented this, it is an old technique known as "clamped grid" operation. RG has some stuff about it (from a Milman book) on his site. If you prevent grid current from flowing it will stop working.

6/30/1999 2:38 PM
Mike D.	Ken, Using the Ampage "Marshall Master volume" schematic as a reference: 1. Would you modify the first stage? Is there a problem with grid blocking on the second stage? 2. Between 2nd and 3rd stage: change 22n to 3n3, remove 470k series resistor and 470p cap. change 470k grid leak resistor to 220k or 150k. 3. on 3rd stage, increase Rk to 22k. 4. How useful is it to have variable Rk on 2nd and 3rd gain stage? Thanks, Mike D.

6/30/1999 2:38 PM

Mike D.

Ken,

Using the Ampage "Marshall Master volume" schematic as a reference:

1. Would you modify the first stage? Is there a problem with grid blocking on the second stage?

2. Between 2nd and 3rd stage: change 22n to 3n3, remove 470k series resistor and 470p cap. change 470k grid leak resistor to 220k or 150k.

3. on 3rd stage, increase Rk to 22k.

4. How useful is it to have variable Rk on 2nd and 3rd gain stage?

Thanks,

Mike D.

6/30/1999 8:13 PM
Ken Gilbert	1. Would you modify the first stage? Is there a problem with grid blocking on the second stage? Looking from the component values, I would say no, there isn't a problem with blocking, with one provision--the amp is driven by a guitar level signal. If you've got a tube screamer or the like in front of the whole shebang, then you're talking much greater signal amplitudes through the front end. Notice how the only stage with a cathode bypass cap is the input. The other following stages do NOT, and I suspect that's one of the reasons that the Marshall has the crunch it does... the lack of Ck makes it that much harder to run into grid conduction. That 10K Rk on the second stage is running that tube pretty cold--I'd say half a milliamp or less. At first glance, you'd look at that stage and as a quick approximation, the gain can only be 10 or less--what good is that? But try wedging a bypass cap across that 10K and see what happens--I'd almost guarantee the amp won't sound as good! That first stage's bypass cap is pretty small, too, with an approximate break frequency of 86Hz or so... that means any signal with a frequency lower than that will cause degeneration (negative current feedback) across that Rk, lessening the gain from 100 (or so) down to 37 (or so). [These figures are first approximations, without taking into account the internal losses in the tube itself.] This limit of LF helps to avoid the grid blocking. 2. Between 2nd and 3rd stage: change 22n to 3n3, remove 470k series resistor and 470p cap. change 470k grid leak resistor to 220k or 150k. That's exactly what I would try. I really have no idea what the outcome will be--the subsequent EQ has a LOT to do with the sound. As mods go, this could be achieved in a couple of seconds--just jumper the series 470K, don't worry about the 470p, pull the 22n and replace with a 3n3. You might even try gator clipping a pot for the grid leak resistor, so that you can tweak the values here. Incidentally, I like the sound of a heavy AC shunt loaded tube stage... ie. putting a fairly low value grid leak resistor in there. 3. on 3rd stage, increase Rk to 22k. Now, the only thing you've got to keep in mind here is that the CF is direct coupled... if you change the plate current of the 3rd stage, then you're also changing the voltage drop across Rl--thereby screwing with the grid voltage of the CF. That's not inherently a "bad thing," but it's something to be aware of. If you run the CF grid too close to B+, you're going to stress the tube/Rk with too much current, and limit it's pk-pk Vo. One thing I've done with much success in these direct coupled CF stages is to return the cathode to a NEGATIVE voltage supply (like the bias supply), increase the Rk to a value suitable for the desired plate current (like Rk=V-/Ip), and then bypass to GROUND with a non-polar cap. By doing so, you increase the dynamic range of the stage, and with a careful choice of Ck, get you the frequency response you want. I usually try to get the most peak to peak signal swing out of my CF's, which means situating their grids at about B+/2. 4. How useful is it to have variable Rk on 2nd and 3rd gain stage? I find that the variable Rk is one of the most versatile resistances to make variable, since it affects both stage gain AND bias. If you've got the pots, and you've got the chassis space, throw 'em in there--just hot glue 'em down. You can always yank 'em out later, once you know how varying their values alters the tone and response. Very educational. Hell, just use gator clips everywhere, and turn all the knobs every which way. Keep the guitar strapped to your back, of course, and keep whanging on the strings every now and then. Make sure you're running the amp into the rig that you normally would--or if you're not, you're periodically checking it. It's amazing how much the speakers/cabinet/power amp can change how you think of the tone. One thing about very low current stages--it's almost impossible to get a VERY clean sound out of them--there's always some hash in the background as the tube cuts off sharply... but then again I think they can actually act like a volume EXPANSION device, where a small input nets a very small output, but a slightly larger input (past a threshold of sorts) nets a much larger Vo. Makes a little sense if you look at the curve of the transfer characteristics--the gm goes way down as the tube gets pinched off. As of yet unverified by anything other than my ears though. ~KG~

6/30/1999 8:13 PM

Ken Gilbert

1. Would you modify the first stage? Is there a problem with grid blocking on the second stage?

Looking from the component values, I would say no, there isn't a problem with blocking, with one provision--the amp is driven by a guitar level signal. If you've got a tube screamer or the like in front of the whole shebang, then you're talking much greater signal amplitudes through the front end.

Notice how the only stage with a cathode bypass cap is the input. The other following stages do NOT, and I suspect that's one of the reasons that the Marshall has the crunch it does... the lack of Ck makes it that much harder to run into grid conduction. That 10K Rk on the second stage is running that tube pretty cold--I'd say half a milliamp or less. At first glance, you'd look at that stage and as a quick approximation, the gain can only be 10 or less--what good is that? But try wedging a bypass cap across that 10K and see what happens--I'd almost guarantee the amp won't sound as good!

That first stage's bypass cap is pretty small, too, with an approximate break frequency of 86Hz or so... that means any signal with a frequency lower than that will cause degeneration (negative current feedback) across that Rk, lessening the gain from 100 (or so) down to 37 (or so). [These figures are first approximations, without taking into account the internal losses in the tube itself.] This limit of LF helps to avoid the grid blocking.

2. Between 2nd and 3rd stage: change 22n to 3n3, remove 470k series resistor and 470p cap. change 470k grid leak resistor to 220k or 150k.

That's exactly what I would try. I really have no idea what the outcome will be--the subsequent EQ has a LOT to do with the sound. As mods go, this could be achieved in a couple of seconds--just jumper the series 470K, don't worry about the 470p, pull the 22n and replace with a 3n3. You might even try gator clipping a pot for the grid leak resistor, so that you can tweak the values here.

Incidentally, I like the sound of a heavy AC shunt loaded tube stage... ie. putting a fairly low value grid leak resistor in there.

3. on 3rd stage, increase Rk to 22k.

Now, the only thing you've got to keep in mind here is that the CF is direct coupled... if you change the plate current of the 3rd stage, then you're also changing the voltage drop across Rl--thereby screwing with the grid voltage of the CF. That's not inherently a "bad thing," but it's something to be aware of. If you run the CF grid too close to B+, you're going to stress the tube/Rk with too much current, and limit it's pk-pk Vo.

One thing I've done with much success in these direct coupled CF stages is to return the cathode to a NEGATIVE voltage supply (like the bias supply), increase the Rk to a value suitable for the desired plate current (like Rk=V-/Ip), and then bypass to GROUND with a non-polar cap. By doing so, you increase the dynamic range of the stage, and with a careful choice of Ck, get you the frequency response you want.

I usually try to get the most peak to peak signal swing out of my CF's, which means situating their grids at about B+/2.

4. How useful is it to have variable Rk on 2nd and 3rd gain stage?

I find that the variable Rk is one of the most versatile resistances to make variable, since it affects both stage gain AND bias. If you've got the pots, and you've got the chassis space, throw 'em in there--just hot glue 'em down. You can always yank 'em out later, once you know how varying their values alters the tone and response. Very educational. Hell, just use gator clips everywhere, and turn all the knobs every which way. Keep the guitar strapped to your back, of course, and keep whanging on the strings every now and then. Make sure you're running the amp into the rig that you normally would--or if you're not, you're periodically checking it. It's amazing how much the speakers/cabinet/power amp can change how you think of the tone.

One thing about very low current stages--it's almost impossible to get a VERY clean sound out of them--there's always some hash in the background as the tube cuts off sharply... but then again I think they can actually act like a volume EXPANSION device, where a small input nets a very small output, but a slightly larger input (past a threshold of sorts) nets a much larger Vo. Makes a little sense if you look at the curve of the transfer characteristics--the gm goes way down as the tube gets pinched off. As of yet unverified by anything other than my ears though.

~KG~

Page 1 of 4

Next>

Last Page>>