First off the datasheet for the 6AU8A is here http://www.mif.pg.gda.pl/homepages/frank/sheets/093/6/6AU8A.pdf Ok I want to use this tube as a front end for a pushpull hybrid circuit. Attached is the circuit I came up with, pentode section as the voltage gain stage dc coupled to the triode section as a split load PI. The PI will be driving about 9Vrms into a pair of power followers with worst case input impendance of 100k ohm for full output of about 18 watts. I have added expected voltages and currents to the schematic.
My main concerns are that I have not built anything with this high of a B+ rail before and I would like to know if I did the dc coupling correctly as I don't want the triode to self destruct. Also will I need to reference the heater supply to a higher voltage because of the 200v cathode to heater limit on the triode section? The cathode of the triode will be sitting somewhere near 150V if I got it correct. Could someone show an example of how to reference the heater supply if it needs to be?
I could use rc coupling for the PI and that leads to my next question. How many caps can the signal go through before phase shift has a bad effect on gNFB? If I were to rc couple there would be 3 caps in all before returning to the gNFB point, 2 if dc coupled.
Would the use of a diode such as used by John Broskie between the grid and cathode to keep the the grid from being pulled to B+ be of any benefit here or would it interfear with the PI's operation?
And the last question for now. Does my chosen operating point for the pentode section look correct and fairly linear? Operating point will be about 142 volts from plate to cathode with a bias voltage of -3.5V and an idle current around 6 - 6.5mA , going by the datasheet anyway.
My main concerns are that I have not built anything with this high of a B+ rail before and I would like to know if I did the dc coupling correctly as I don't want the triode to self destruct. Also will I need to reference the heater supply to a higher voltage because of the 200v cathode to heater limit on the triode section? The cathode of the triode will be sitting somewhere near 150V if I got it correct. Could someone show an example of how to reference the heater supply if it needs to be?
I could use rc coupling for the PI and that leads to my next question. How many caps can the signal go through before phase shift has a bad effect on gNFB? If I were to rc couple there would be 3 caps in all before returning to the gNFB point, 2 if dc coupled.
Would the use of a diode such as used by John Broskie between the grid and cathode to keep the the grid from being pulled to B+ be of any benefit here or would it interfear with the PI's operation?
And the last question for now. Does my chosen operating point for the pentode section look correct and fairly linear? Operating point will be about 142 volts from plate to cathode with a bias voltage of -3.5V and an idle current around 6 - 6.5mA , going by the datasheet anyway.
Attachments
First off the datasheet for the 6AU8A is here http://www.mif.pg.gda.pl/homepages/frank/sheets/093/6/6AU8A.pdf Ok I want to use this tube as a front end for a pushpull hybrid circuit. Attached is the circuit I came up with, pentode section as the voltage gain stage dc coupled to the triode section as a split load PI. The PI will be driving about 9Vrms into a pair of power followers with worst case input impendance of 100k ohm for full output of about 18 watts. I have added expected voltages and currents to the schematic.
My main concerns are that I have not built anything with this high of a B+ rail before and I would like to know if I did the dc coupling correctly as I don't want the triode to self destruct. Also will I need to reference the heater supply to a higher voltage because of the 200v cathode to heater limit on the triode section? The cathode of the triode will be sitting somewhere near 150V if I got it correct. Could someone show an example of how to reference the heater supply if it needs to be?
I could use rc coupling for the PI and that leads to my next question. How many caps can the signal go through before phase shift has a bad effect on gNFB? If I were to rc couple there would be 3 caps in all before returning to the gNFB point, 2 if dc coupled.
Would the use of a diode such as used by John Broskie between the grid and cathode to keep the the grid from being pulled to B+ be of any benefit here or would it interfear with the PI's operation?
And the last question for now. Does my chosen operating point for the pentode section look correct and fairly linear? Operating point will be about 142 volts from plate to cathode with a bias voltage of -3.5V and an idle current around 6 - 6.5mA , going by the datasheet anyway.
I substituted the 6GH8A in a Dynaco MKIV with 6AU8A using an adapter socket, it did a marvelous job!!
My main concerns are that I have not built anything with this high of a B+ rail before and I would like to know if I did the dc coupling correctly as I don't want the triode to self destruct.
450V isn't all that high. DC coupling can be a problem in hollow state designs. I took care of the problem in two ways for two different projects. The one used a SS power supply with a separate heater supply. That way, all the heaters can come up to operating temp before the HV is switched on. That way, they start conducting at once, like transistors, and excessive Vgk's simply don't appear. For another project, I simply floated the heater supply, that way, Vhk doesn't become a problem. I didn't have any issues with hum, but that was for that particular project. No idea how a floating heater supply would fare here.
Also will I need to reference the heater supply to a higher voltage because of the 200v cathode to heater limit on the triode section? The cathode of the triode will be sitting somewhere near 150V if I got it correct. Could someone show an example of how to reference the heater supply if it needs to be?
If you don't float the heater supply, you'll need to bias the heater supply to about +75Vdc above ground to stay within the Vhk spec. That can be a voltage divider across the main DC rail, connected either to the heater center tap, or a tap fabricated from a couple of resistors across the 6.3V secondary, bypassed to ground with a suitable capacitor.
I could use rc coupling for the PI and that leads to my next question. How many caps can the signal go through before phase shift has a bad effect on gNFB? If I were to rc couple there would be 3 caps in all before returning to the gNFB point, 2 if dc coupled.
As few RC couplings as you can manage. Be sure to stagger the cutoff frequencies so the phase shifts don't become excessive.
Would the use of a diode such as used by John Broskie between the grid and cathode to keep the the grid from being pulled to B+ be of any benefit here or would it interfear with the PI's operation?
It's a good solution to the excessive grid voltage problem. That diode drops out when the voltages stabilize since it's reverse biased. Its internal capacitance shouldn't be a problem since the cathode voltage follows the grid voltage, reducing the current through that capacitance, effectively reducing it below its nominal static value.
And the last question for now. Does my chosen operating point for the pentode section look correct and fairly linear? Operating point will be about 142 volts from plate to cathode with a bias voltage of -3.5V and an idle current around 6 - 6.5mA , going by the datasheet anyway.
That's more of a problem since designing pent stages isn't so simple as triode stages. With triodes, you have just the one set of plate characteristics. Pents have a potentially infinite number of plate characteristics, one for each screen voltage. This type of design is therefore more empirical to determine the harmonic distortion, and its content. The usual rule is to get that screen voltage as low as you can manage. Given that you have -3.5Vdc of grid bias, you might have room to drop the screen voltage, depending on what your input voltage is, of course.
My main concerns are that I have not built anything with this high of a B+ rail before and I would like to know if I did the dc coupling correctly as I don't want the triode to self destruct.
You have given the solution alredy - a diode between G1 and K. This will limit the inital voltage while the tube is cold, to a sensible voltage. Other than that, the DC coupling is OK, but do check if you can get the proper swing. A good rule of thumb for low distortion is to have about 3x swing as the DC voltage on the anode and cathode resistor of the phase splitter, and about the same voltage across the resistors and the triode. There is a good amount of 'play' in this, for instance the DC voltage across the resistors can be about 2x the swing (one way), and the voltage caross the triode somewhat more than on the resitors. Or, look at it in another way, with a predetermined PSU voltage, you get the required DC voltage on the pentode plate. In general the pentode can 'pull' it's plate to much lower voltages compared to a similar size triode, so it's rarely the limiting factor when it comes to maximum output swing.
Here is an example - given a 300V PSU voltage, you decide on ~100V DC on the phase splitter resistors and triode. This will put the triode cathode at about 100V, the plate at 200V. And, since G1 voltage is only a few V below cathode, this also gives you a bit less than 100V on the plate of the pentode. And, with that you can expect about +-30V output swing with little distortion.
For a 145V plate voltage of the pentode, you are looking at a PSU voltage of 450V for the phase splitter, maybe 400V if you don't need to squeeze the maximum swing out of it.
Also will I need to reference the heater supply to a higher voltage because of the 200v cathode to heater limit on the triode section? The cathode of the triode will be sitting somewhere near 150V if I got it correct. Could someone show an example of how to reference the heater supply if it needs to be?
I'm sure you can figure it out simply make a resistor divider from B+ with about 1mA or so current going through it, to a required voltage, then separate the resistors and insert the heater supply into their former connection point. In some cases you will have to block the reference voltage to ground using a capacitor to prevent signal transfer within the pentode-triode through the internal capacitances between the cathodes and the heater. In this case the divider stays connected normally with a cap between the reference voltage point and ground, and two small-ish resistors (few hundred ohms) are connected across the heater supply to create a center point of the heater AC at their joining point. This you connect to the heater reference voltage.
As for what the voltage would be, it's very desirable for the pentode cathode to have a lower DC voltage than the heater reference. This is because the cathode - heater system forms a (rather bad but still) diode so you can get leakage current from cathode to the heater - the heater being usually AC supplied, you would get a hum injection point which the pentode would gladly amplify. On the other hand it's not so important for the triode since the phase splitter gain is ~1 and it's operating with an already amplified input signal. Therefore, you would lwant to chose a heater reference voltage about right in the middle between the pentode and triode cathode voltage, so at about 70V or so in your example.
I could use rc coupling for the PI and that leads to my next question. How many caps can the signal go through before phase shift has a bad effect on gNFB? If I were to rc couple there would be 3 caps in all before returning to the gNFB point, 2 if dc coupled.
In principle it depends on the size of the coupling caps, i.e. the time constants. But there is a catch. These are not the only ones in the system - there is a hidden loop in there, namely, you also have the power supply filtering chain constants. So, even though you could get phase shift at audio frequencies down to negligible value by choosing relatively high capacitance coupling caps (and do not forget cathode caps and G2 caps are also coupling caps!), you may create phase shift loops at very low frequencies formed by the signal coupling and the PSU filter components (the RC filtering going from the B+ point in your schematic through a succession of RC voltage drop / filtering reistors and caps). Since the signal generates a current swing in the power supply, the power supply filter chain also represents a feedback loop of sorts between stages. Not being careful with the time constants can result in this loop becoming a positive feedback at nearly the full gain of the circuit, and you get LF oscillations, or in extreme conditions, this can drive the stages into highly nonlinear behavior and you get a phenomenon called motorboating. For this reason it's advisable to avoid adding a time constant into circuit with a loop in it whenever you can.
Would the use of a diode such as used by John Broskie between the grid and cathode to keep the the grid from being pulled to B+ be of any benefit here or would it interfear with the PI's operation?
It would be of great benefit. In the grand majority of cases (in virvuits in general) it does not interfere as the diode stays comfortably reverse biased. The exception would be circuits where positive G1 voltage (WRT cathode, of course) is expected and desired, however this most certainly is NOT in this sort of PI. In fact, it's often desirable to increase the value of the grid stopper resistor to a double digit k ohm value to somewhat reduce the problem when it happens (and it can when a very high amplitude input signal is supplied)in which case grid current will flow and end up going through the cathode resistor (but obviously not the plate resistor) of the PI, and it's output will not be symmetric any more. This is always a problem in PP amps, as it results in appearance of un-cancelled DC current or voltage in the output stage (normally transformer coupled so it does not go out to the load but saturates and sometimes magnetizes the transformer core). Much more dangerous in hybrids where the speaker is directly coupled so in your case it's something to think about.
And the last question for now. Does my chosen operating point for the pentode section look correct and fairly linear? Operating point will be about 142 volts from plate to cathode with a bias voltage of -3.5V and an idle current around 6 - 6.5mA , going by the datasheet anyway.
The question here is what Vg2 will be so that you can extrapolate a proper set of plate curves for the pentode. There are many combinations which give the same DC operating point, but very different gain and linearity. For this particular pentode look at operating points and load lines down at the bottom of the plate curves. You will see that the distance between them for a given step of G1 voltage can increase dramatically as Vg1 approaches zero, which translates into fairly non-linear operation. But, there is an area, and at G2 voltages of around 100V or so, you would be there with Vg1=-3V or so, where the plate curves are fairly evenly spaced. The gain is around 100-150 (depending on B+ available and plate resistor), and quite linear. So, I'd say you are thereabouts now.
One more thing - using bootstraping with a pentode is certainly possible and you can get very high gains (even 4 digit figures) but it comes at the price of distortion, and mostly high-order at that. It's not obvious because the curves are quite imprecise, usually you are working in an area where the lines used to draw the curves and the scale of the graph suggest nice smooth and evenly spaced curves, but it might not be so. However, you can have serious benefits from partial bootstrapping (there is a resistor in series with the bootstrap cap) especially when there is not enough B+ available for the pentode to get a sufficiently flat load line to get high gain and high swing. This bootstrapping can even be done off the cathode output of the PI adjusting the resistors so that the cathode and plate see the same load (to insure symmetry) but I'll leave this as an idea for you to try and figure out
Here are some suggestions, i.e. answers to unasked questions about your circuit
Think about connecting the input resistor (220k from G1) to ground. You will need to adjust the values of the cathode resistors to get the same bias voltage and ratio for feedback but this should not be a problem. Also consider removing the capacitor in the feedback loop. You will have a relatively low voltage on the feedback insertion point to the cathode, say 1V or so, but it will be connected to the output of your circuit via a fairly large resistor. Since your output circuit is a follower of some kind and is no doubt referenced to 0V so that there is no DC on the output, it will try to keep this conition even with the minute amount of current flowing through the feedback resistor. This will slightly upset the DC bias of the pentode, maiking it a bit lower (easy to calculate from the ratios of the resistors) but it'e easily fixed by a small correction of the value of the bypassed portion of the cathode resistor to get the same DC voltage on G1. And as I said, getting a time constant out of a loop is always desirable.
That being said, keep the input cap. While it's theoretically not needed, it serves a potentially important role - because your PI outputs are at vastly different voltages they will need to be capacitor coupled to the output stage. This adds an LF cut, i.e. a high pass. The problem here is that this makes your NFB behave as a low pass, i.e. there is less and less NFB at frequencies below the LF cut point, so gain of the amp rises, and in fact is the highest possible for subsonics. The problem here is not obvious as the amplified subsonic signal cannot get to the output stage due to the same coupling caps that create the problem, BUT it is still amplified, and by the full gain of the circuit, which implies that it might drive the tube stages into clipping. For instance, expecting gain so that you get the full power output with 1Veff at the input will mean that the closed loop gain of your circuit will be about 14 or so. But open loop gain will be >100, say about 6-7 times higher. This is how much more a subsonic signal may be amplified, and in some cases you may get surprising distortion apparently 'without cause'. This is where your input cap comes in. It should be chosen to have the same or smaller time constant as your coupling caps. It's own high-pass characteristic will compensate the gain rise due to NFB falling out at LF and you can have your cake and eat it too.
There is also an apparent alternative - you can dispense with the input cap but manipulate the value of the cathode cap to get approximately the same result. Aproximate because the gain of the whole amp does not fall to zero for a DC input, but you would never expect that anyway, and it might drop to acceptable levels. However, this is a tricky proposition as it lowers OLG and therefore there is less and less of gain for the NFB to work with in TWO places (coupling of the PI with oiuput stage and cathode degeneration). It is possible but the turnover frequencies have to be set very low (think 1Hz or lower), so that even at the lowest audio frequency you have sufficient NFB action (including the available gain). Also, there is a hidden trap - if the cathode cap ends up being sufficiently large vakue so that an electrolytic cap must be used, you will run into serious trouble with capacitor distortion, which happens whenever an AC component appears on a cap which is not negligible to the DC component of the cap voltage. Foil caps are MUCH better at this than any elecrolytic but may still have problems, why it's usually more desirable to keep the input cap because it's within the usual values for foil caps and usually cheap enough to insure very good quality - and it's not in any loop. Capacitor distortion is, BTW, one good reason why coupling time constants of various kinds are chosen such that the resulting LF cut-off is at first glance needlessly low - i.e. the capacitance values seem exaggerated. This works particulairly well with electrolytics assuming their leakage current is not a problem (larger values often have larger leakage).
Think about connecting the input resistor (220k from G1) to ground. You will need to adjust the values of the cathode resistors to get the same bias voltage and ratio for feedback but this should not be a problem. Also consider removing the capacitor in the feedback loop. You will have a relatively low voltage on the feedback insertion point to the cathode, say 1V or so, but it will be connected to the output of your circuit via a fairly large resistor. Since your output circuit is a follower of some kind and is no doubt referenced to 0V so that there is no DC on the output, it will try to keep this conition even with the minute amount of current flowing through the feedback resistor. This will slightly upset the DC bias of the pentode, maiking it a bit lower (easy to calculate from the ratios of the resistors) but it'e easily fixed by a small correction of the value of the bypassed portion of the cathode resistor to get the same DC voltage on G1. And as I said, getting a time constant out of a loop is always desirable.
That being said, keep the input cap. While it's theoretically not needed, it serves a potentially important role - because your PI outputs are at vastly different voltages they will need to be capacitor coupled to the output stage. This adds an LF cut, i.e. a high pass. The problem here is that this makes your NFB behave as a low pass, i.e. there is less and less NFB at frequencies below the LF cut point, so gain of the amp rises, and in fact is the highest possible for subsonics. The problem here is not obvious as the amplified subsonic signal cannot get to the output stage due to the same coupling caps that create the problem, BUT it is still amplified, and by the full gain of the circuit, which implies that it might drive the tube stages into clipping. For instance, expecting gain so that you get the full power output with 1Veff at the input will mean that the closed loop gain of your circuit will be about 14 or so. But open loop gain will be >100, say about 6-7 times higher. This is how much more a subsonic signal may be amplified, and in some cases you may get surprising distortion apparently 'without cause'. This is where your input cap comes in. It should be chosen to have the same or smaller time constant as your coupling caps. It's own high-pass characteristic will compensate the gain rise due to NFB falling out at LF and you can have your cake and eat it too.
There is also an apparent alternative - you can dispense with the input cap but manipulate the value of the cathode cap to get approximately the same result. Aproximate because the gain of the whole amp does not fall to zero for a DC input, but you would never expect that anyway, and it might drop to acceptable levels. However, this is a tricky proposition as it lowers OLG and therefore there is less and less of gain for the NFB to work with in TWO places (coupling of the PI with oiuput stage and cathode degeneration). It is possible but the turnover frequencies have to be set very low (think 1Hz or lower), so that even at the lowest audio frequency you have sufficient NFB action (including the available gain). Also, there is a hidden trap - if the cathode cap ends up being sufficiently large vakue so that an electrolytic cap must be used, you will run into serious trouble with capacitor distortion, which happens whenever an AC component appears on a cap which is not negligible to the DC component of the cap voltage. Foil caps are MUCH better at this than any elecrolytic but may still have problems, why it's usually more desirable to keep the input cap because it's within the usual values for foil caps and usually cheap enough to insure very good quality - and it's not in any loop. Capacitor distortion is, BTW, one good reason why coupling time constants of various kinds are chosen such that the resulting LF cut-off is at first glance needlessly low - i.e. the capacitance values seem exaggerated. This works particulairly well with electrolytics assuming their leakage current is not a problem (larger values often have larger leakage).
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.