OK, here is my take on the circuit and the resons for my design choices. There are several places that GNFB can be applied to an input stage. In a single ended input stage the most common point is the cathode of the input tube. G2 and G3 on pentodes are other possibilities but seldom used. It is also possible to sum the GNFB with the input signal directly at the grid of the first stage as Fred did. This has one disadvantage in some cases. A small anount of the GNFB signal will be present at the input terminals of the amplifier. This GNFB signal will be imposed on the output of the driving circuit. In most cases this will not be an issue, but if the driving stage is a poorly designed silicon based life form IMD could result.
In a LTP there is another input available for GNFB, the tube on the right. The two inputs do not couple to each other, therefore the tube on the right seems like a logical place to inject GNFB. This is how most SS opamps work and is why I use that approach.
We also need a way to compensate for the DC offsets that accumulate due to mismatched devices in a DC coupled amplifier. As more stages are DC coupled together the problem compounds. A small DC voltage applied to any of the same points that respond to GNFB will work, so I simply sum the DC offset voltage and the GNFB and apply it to the same input.
Many of my applications for these driver boards will not use GNFB so only the DC offset correction is active.
I have worked on an automatic DC offset correction system and it will be implemented by connecting it to the same points once it works right.
In a LTP there is another input available for GNFB, the tube on the right. The two inputs do not couple to each other, therefore the tube on the right seems like a logical place to inject GNFB. This is how most SS opamps work and is why I use that approach.
We also need a way to compensate for the DC offsets that accumulate due to mismatched devices in a DC coupled amplifier. As more stages are DC coupled together the problem compounds. A small DC voltage applied to any of the same points that respond to GNFB will work, so I simply sum the DC offset voltage and the GNFB and apply it to the same input.
Many of my applications for these driver boards will not use GNFB so only the DC offset correction is active.
I have worked on an automatic DC offset correction system and it will be implemented by connecting it to the same points once it works right.
It's the same difference between an inverting connection, and a non-inverting connection when using solid state op-amps.
When applying your gNFB to the opposite side of an LTP, you gain the advantage of not dragging down the input impedance of the input side since you don't have that resistor connected between the signal source and what is (more or less) a virtual AC ground connection.
When building an amp that incorporates an OPT, it doesn't make any difference, since positive feedback becomes negative feedback simply by reversing the OPT primary connection(s) to the final(s).
In an OPT-less design like Fred's, you don't have that possibility. Therefore, your only choices as for getting the phase correct is to eliminate a gain stage, reducing the open and closed loop gain to an unworkable lack of input sensitivity. Or you can add another inverting stage, meaning another hole to knock in the chassis, another source of open loop distortion, increased difficulties with DC offset, more heater and plate power demand from the PS.
I don't particularly like this sort of feedback connection since the gain will vary with changes in input impedance, since the source impedance is in series with the feedback sampling resistor. That needs to be isolated from the source with a cathode follower. (This is not a prob when using op-amps since the Zo of an op-amp is very small compared to the sampling resistors usually used.)
thanks tubelab. i'm still unsure of where the feedback signal acually comes from in your schem. i see the set up for the dc bias with the LEDS using the filament winding, but what are T3-PRI1 1, 2 and 3? it looks like the feedback signal should be applied to T3-PRI1 3 and C4 maintains the DC bias voltage, is that about right?
First off, I must state that you should ignore all of the labels on the connectors in that schematic. Being lazy I copied another schematic (the dual 6SN7 design that I started this thread with) and just changed the input tubes. That schematic was copied from another......The labels come from the Simple SE which bears no resemblance to what is going on here.
The normal signal input is on the left and is labeled T2-SEC. The inverting input is connected to the grid circuit of the right pentode. T3-PRI1-1 is the input or feedback ground connection. T3-PRI1-2 is the inverting input hot connection. If you wanted a fully differential (or balanced) input it would be connected betweenT3-PRI1-2 and T2-SEC2-1. C2 and R47 are the usual feedback compensation components. For GNFB the secondary of the OPT is connected with the 0 ohm wire on T3-PRI-1 and the speaker lead on T3-PRI-3.
The LEDs are simply used as 2 volt zener diodes. They are fed from the negative bias voltage supply, T1-RED-1 (about -100 volts) and the positive PowerDrive supply, T1-RED-3 (about +25 to 50 volts). The PowerDrive circuitry is not shown on this schematic since it was moved to the output PCB for this experiment. C4 is a bypass cap to stop any zener (LED) noise from getting into a sensitive input. The pot R14 provides a small voltage that is adjustable from -2 to +2 volts to the grid through R11.
T1-RED1-2 is the screen voltage for the input tubes. It is about +125 volts. T1-RED1-1 is the plate supply for the driver tubes. I was experimenting in the 400 to 600 volt range for maximum drive voltage!
It must be stressed that this PC board was never completed. Life has a way of interrupting fun projects and this one sits among others looking for attention. Breadboard experiments were positive but that's as far as it got. A second breadboard is nearly done which may result in some circuit changes whenever I get time to revisit it.
Guys (tubelab maybe?),
Looking at the FET powergrid design as a source follower, I've just seen another similar circuit but the source of the FET was just tied to ground (with a resistor of course) instead of a negative voltage applied.
What advantage does the negative voltage have and is it necessary?
Can we just have the B+ applied to the Drain and a sink resistor on the Source to set the current through the FET?
Looking at the FET powergrid design as a source follower, I've just seen another similar circuit but the source of the FET was just tied to ground (with a resistor of course) instead of a negative voltage applied.
What advantage does the negative voltage have and is it necessary?
Can we just have the B+ applied to the Drain and a sink resistor on the Source to set the current through the FET?
George, why a cuarrent balance pot (R24) and a screen adjust pot (R5) on the two input pentodes?
I presume R5 is for AC symetry so that tubes don't have to be matched.
I though an LTP would provide symetry if driven from a current source in the cathodes as you have done.
It seems that if you adjust R24 for balance, then adjust R5 you will upset the DC balance again. If you adjust R5 first for AC symetry first, then adjust R24, you will upset the AC symetry (although not as much as the former would). Is R5 indeed for AC symetry?
I presume R5 is for AC symetry so that tubes don't have to be matched.
I though an LTP would provide symetry if driven from a current source in the cathodes as you have done.
It seems that if you adjust R24 for balance, then adjust R5 you will upset the DC balance again. If you adjust R5 first for AC symetry first, then adjust R24, you will upset the AC symetry (although not as much as the former would). Is R5 indeed for AC symetry?
I use the dumb blonde approach when trying something new. I make everything adjustable.
In theory the pot in the cathodes can cause a slight amplitude imbalance at the outputs since there is an AC voltage drop across the pot. I wanted the ability to try both methods of achieving balance by shorting out one pot or the other. Layziness took over when I made the breadboard and the screen pot never got implemented. A PC board layout is mostly finished that includes both pots. I have not had time to make one yet.
If you are using fixed bias the grid of the output tube must have a negative voltage applied to it. This negative voltage has to come from somewhere. In a normal fixed bias circuit the negative voltage EQUALS the needed grid bias and gets modulated up and down with the drive signal such that under ideal maximum output conditions in AB1 the grid voltage will swing from negative twice the bias value to zero. So for a hypothetical circuit that needs -25 volts on the grid the grid will swing from -50 volts to zero volts. Under AB2 conditions the grid could swing from -70 to +20 volts.
The PowerDrive is DC coupled. The advantage is the capacitor and its long time constant is moved away from the output tube grid. Since the grid - cathode of the output tube by definition form a diode, the diode - capacitor combination form an AM detector. This will force the charge on the capacitor to vary with the average value of the audio signal upsetting the bias at a low frequency rate. This eliminates blocking distortion under overload conditions, and also eliminates the audible artifacts of slowly varying bias. The disadvantage is that a negative supply that is equal to the maximum negative grid swing plus some headroom is needed. In this case (-25 volts grid bias) I would use a -80 to -100 volt supply.
I am not familiar with the circuit that you saw but a negaive supply is needed unless the cathode operates at a positive voltage.
thxs tubelab.
I got a little lost in your second paragraph
I have the FET DC coupled. I have a fixed bias applied before a 1K resistor before the Gate of the FET.
Then I have around +120 volts on the Drain and about -85 on the source of the FET.
So the neg. voltage on the source of the FET eliminates blocking distortion under overload conditions, and also eliminates the audible artifacts of slowly varying bias?
Is this noticeable so much or could the FET be grounded (with a suitable resistor)?
I got a little lost in your second paragraph
I have the FET DC coupled. I have a fixed bias applied before a 1K resistor before the Gate of the FET.
Then I have around +120 volts on the Drain and about -85 on the source of the FET.
So the neg. voltage on the source of the FET eliminates blocking distortion under overload conditions, and also eliminates the audible artifacts of slowly varying bias?
Is this noticeable so much or could the FET be grounded (with a suitable resistor)?
To restate a bit, the driver tube is AC coupled to the fet. Since the fet is very high Z, the capacitor can be small, and the fet does not put out grid current .
The FET follower is DC coupled to the output grid. This removes several issues that George explained. The output tube grid current cannot affect the bias because there is negligible capacitance to charge and the fet supplies a very low Z path to AC ground.
If the FET had no negative supply, it would need to be AC coupled to the output grid, loosing many of powerdrive's advantages. To see a circuit example, find Tubelab SE on George's site.
HTH
Doug
.The FET follower is DC coupled to the output grid. This removes several issues that George explained. The output tube grid current cannot affect the bias because there is negligible capacitance to charge and the fet supplies a very low Z path to AC ground.
If the FET had no negative supply, it would need to be AC coupled to the output grid, loosing many of powerdrive's advantages. To see a circuit example, find Tubelab SE on George's site.
HTH
Doug
Thxs Doug. That helps.
Yep mine is AC coupled from the driver to FET and DC coupled from FET to output tube's grid.
I assume that this is why the atmasphere M-60 design has it's CF driving the output tubes grids (AC coupled) powered with a negative supply on the cathode resistors. -Same principle no?
Yes, using a cathode follower is essentially the same as using an FET to drive the output tubes, but all CFs will have a significantly higher output Z than a source follower.
The only situation I can think of that would involve a source follower grounded rather than connected to a negative voltage is if you were running the output tubes entirely in the positive grid region. I think that idea is mentioned in morgan jones' book, to avoid distortion caused by the transition between the negligible grid current present when grids are negative to the substantial grid current when grids are driven in to AB2. With a source follower (as opposed to a cathode following tube), its pretty much a moot point because the output Z of the source follower is so low that this distortion is negligible.
feel free to correct me if i've led anyone down the wrong track here!
The only situation I can think of that would involve a source follower grounded rather than connected to a negative voltage is if you were running the output tubes entirely in the positive grid region. I think that idea is mentioned in morgan jones' book, to avoid distortion caused by the transition between the negligible grid current present when grids are negative to the substantial grid current when grids are driven in to AB2. With a source follower (as opposed to a cathode following tube), its pretty much a moot point because the output Z of the source follower is so low that this distortion is negligible.
feel free to correct me if i've led anyone down the wrong track here!
Chris/George,
I've been working on restoring this other old amp project to get it off my bench (I'll start a thread at some point). Down the road I'm probably going to go to a MOSFET stage like this one. I planned ahead and already have a +/- 100VDC power supply in there. I've been taking a more in-depth look at this design and I really like it a lot.
Just to be clear, there are no DC-blocking capacitors between the two differential stages, right? I see that the plate loads are such that the CCS can hang the second cathode pair at a higher voltage to get the right current flow for whatever the grids end up at. Am I seeing this right?
I've been working on restoring this other old amp project to get it off my bench (I'll start a thread at some point). Down the road I'm probably going to go to a MOSFET stage like this one. I planned ahead and already have a +/- 100VDC power supply in there. I've been taking a more in-depth look at this design and I really like it a lot.
Just to be clear, there are no DC-blocking capacitors between the two differential stages, right? I see that the plate loads are such that the CCS can hang the second cathode pair at a higher voltage to get the right current flow for whatever the grids end up at. Am I seeing this right?
Neither of our amps use caps between the tubes. I have a version of the PC board that still has the caps for cases when I want to extract the maximum amount of drive voltage for a cathode follower output stage. This allows the second stage to operate with its cathode at or below ground potential. The capless version just works better for all "normal" amp designs.
The CCS sets the current for the diff pair. The plate load resistor sets the plate voltage at that current. Pick a current, then set the plate load for maximum undistorted plate voltage swing for a given B+, OR pick a current and pick a needed amount of drive voltage based on the output tube choice and set the plate load resistance for the lowest distortion (or best distortion spectra) at that voltage. In other words you don't need to be able to crank out 200 volts P-P for 6L6GC's!
We discovered that -100 volts may not be quite enough for KT88's in triode at high power and will not be enough for big sweep tubes. You need at least twice the bias voltage and some bigh sweep tubes need -60 volts or so.
Someone woke this thread up from its sleep, so I got my breadboard back out of the closet. I might get to fire it up this weekend depending on weather. I got the itch to make more speakers. Down here winter is the time to go outside and play with power tools.
Right OK. That answers my question. I guess the point I was trying to make was that this setup is more imune to the typical touchiness you have with directly-coupled stages, since the CCS will effectively move the cathode voltage around to maintain whatever Vgk will flow the current that it wants. If the plate voltages of the first stage aren't exactly where you expected them to be, this is handy. I just find that to be rather elegant and I hadn't noticed it before.
I did catch that. It's more like +/- 130V without a load, but I expect it to sag some since it is a small CRC filter. We'll see how it plays out. The chassis has octal sockets. It originally used 8417s, which I still have. The goal of the project is to let me play around with some of the ideas I keep coming across here.
That's nice. Up here we have to dig through 4 feet of snow to get to our garage.
Power driving KT88's in triode
I'm considering using this 6SN7 universal driver topology to drive trioded KT88's (or 6550's) with a 5K OT primary. The TubeCad PP calculator shows around 28W or so total and 20 watts class A for KT88's depending on B+ (460V or so).
Are 6SN7's still an appropriate choice if I'm planning on no feedback and using mosfets? Seems like lots of voltage gain when I only need 50V PP for AB1. I'm liking the low noise and CCS adjustability aspects of the design though.....
I have some small hammond "flat pack" 120:120V transformers to power the mosfets and I'm planning on using the 2SK3563 mosfets.
I'm considering using this 6SN7 universal driver topology to drive trioded KT88's (or 6550's) with a 5K OT primary. The TubeCad PP calculator shows around 28W or so total and 20 watts class A for KT88's depending on B+ (460V or so).
Are 6SN7's still an appropriate choice if I'm planning on no feedback and using mosfets? Seems like lots of voltage gain when I only need 50V PP for AB1. I'm liking the low noise and CCS adjustability aspects of the design though.....
I have some small hammond "flat pack" 120:120V transformers to power the mosfets and I'm planning on using the 2SK3563 mosfets.
Somewhere back in this thread there should be a post or two of the KT88 itteration that I did. It ran a pair of 6SN7's per channel and worked out quite nice. That version still exists....in my closet. I don't think I have robbed it for parts and it worked very well. No feedback...trioded KT88's.... really good sound.....I used a 6600 ohm OPT which I probably wired for 3300 ohms for the higher powered tests. I am at work now so i can't search for the posts.
I can't think of any better choice of tubes than the 6SN7 for this duty. You only need a 6SL7 in the first tube when you need to drive big tubes, or when you want to use some feedback.
It has been branded impossible and foolish by many, but this amp responds well to Schade feedback. The usual resistor from the plate of the output tube to the plate of the driver works good, but the top secret tweak is to cross couple feedback with a resistor from the plate of the output tube to the plate of the opposite input tube. Here a 6SL7 may be the hot ticket for an input tube even with KT88 outputs.
Is there any particular reason that the bleed R for the -105V rail is 1M and the bleed R for the +75V rail is 220K?.......is it to allow the -105V rail to come up a bit faster?
I'm slowly (very slowy..) breadboarding this.
This is the latest schemo from Chrish in post 460 in this thread....
I'm slowly (very slowy..) breadboarding this.
This is the latest schemo from Chrish in post 460 in this thread....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.