On the left is the input stage that I am currently using in my SE amp. The input stage drives a MOSFET source follower that drives the grid of a transmitter triode. I get about 2.5%THD at 75Vrms with about 65dB of gain in that stage. Distortion is a very low order waterfall. The stage is simply a CCS-loaded pentode with a p-channel follower driving the cathode from a feedback divider from the output tube plate.
I'd like to see if I can maybe do a bit better with a folded cascode or shunt cascode as some here call it. I ordered some 6E5P on Ebay and I think I'll probably pick up a pair of Rod Coleman's boards to play around with. I need to get ~75Vrms out as I want to test out a 211 but I have many other output tubes that don't require that much swing. The main twist I'm adding is the p-channel drive to the cathode of the tube. I've simulated the circuit the cathode drive with common mode signal seems to add a little distortion (from the Va-k variations that the common mode swing induces) but I still want to give this a shot and breadboard it.
I'm just looking for advice on this particular circuit as I've never used it. Is this a dumb idea or will I maybe get what I'm looking for with this approach?
Hello Heath,
Yes, the anode voltage for the 6E5P, and the stress it places on the PNP cascode transistor need careful consideration. A p-channel FET will probably work, and allow higher voltage, but the distortion increases from higher capacitance and lower gm. Still, you can get plenty of gain, and with loop feedback, should be able to iron it out.
The other way to get more swing is to return the load resistor to a negative voltage. The voltage must be stable and low noise though.
This is exactly the approach I have tried, but for lack of time, I have not yet perfected it. With loop feedback, the harmonic cadences did not look so favourable, and the sound not so good as when the feedback arises from the tube's cathode resistor, so I am planning to look harder at it when I get a moment.
I found that the 6E5P needs simple regulation of the heater to maintain stable anode current; and the current sources should also be as stable as you can get them, for a stable dc output. But other than that there is no good reason why it will not be able to drive 75 Vrms.
The great advantage of Shunt Cascode is that you get the high gain accompanied by low open-loop distortion; Also, with the designs I have used the bandwidth runs to 100-800kHz (depending on capacitive load) with minimal phase shift - so closed-loop-feedback is easy to apply.
In the course of the experimentation, the parts for the V3 (2016) board became hard-to-impossible to buy, so the design is not usable, currently. The new V5 will be a great improvement, but I don't have a date yet for getting it fully ready.
But I imagine you could rig up a circuit to try it out.
Yes, the anode voltage for the 6E5P, and the stress it places on the PNP cascode transistor need careful consideration. A p-channel FET will probably work, and allow higher voltage, but the distortion increases from higher capacitance and lower gm. Still, you can get plenty of gain, and with loop feedback, should be able to iron it out.
The other way to get more swing is to return the load resistor to a negative voltage. The voltage must be stable and low noise though.
This is exactly the approach I have tried, but for lack of time, I have not yet perfected it. With loop feedback, the harmonic cadences did not look so favourable, and the sound not so good as when the feedback arises from the tube's cathode resistor, so I am planning to look harder at it when I get a moment.
I found that the 6E5P needs simple regulation of the heater to maintain stable anode current; and the current sources should also be as stable as you can get them, for a stable dc output. But other than that there is no good reason why it will not be able to drive 75 Vrms.
The great advantage of Shunt Cascode is that you get the high gain accompanied by low open-loop distortion; Also, with the designs I have used the bandwidth runs to 100-800kHz (depending on capacitive load) with minimal phase shift - so closed-loop-feedback is easy to apply.
In the course of the experimentation, the parts for the V3 (2016) board became hard-to-impossible to buy, so the design is not usable, currently. The new V5 will be a great improvement, but I don't have a date yet for getting it fully ready.
But I imagine you could rig up a circuit to try it out.
Hi Rod,
Thanks for taking a look at this and responding. I spent the morning sifting through PNP transistor datasheets. I picked a few out to try but honestly I didn't see anything better than the ZTX558. Luckily, that doesn't seem to be one of the obsolete parts from the previous design! I picked a couple of others out that had higher power or voltage handling capability but there are other parameters that are compromised. I've got some good 500V p-channel FETs around. I'll have plenty of parts to test.
I hadn't thought of returning to a negative voltage. I'd prefer not to but that may be where I end up.
Bandwidth was not very impressive with the pentode input stage so I'm hoping this will do much better, and I'm hoping it will do a little bit better than the 6BN11 in terms of distortion as well. Obviously, I want it all.
It seems like in order to get lots of gain I have to have less current going through the pnp device. That seems fine for me since I am just driving a mosfet follower gate and less current will help me stay in the pnp device SOA.
Well, I guess it is time to order some parts and start playing with them. 🙂
I'll report back with results eventually.
Thanks for taking a look at this and responding. I spent the morning sifting through PNP transistor datasheets. I picked a few out to try but honestly I didn't see anything better than the ZTX558. Luckily, that doesn't seem to be one of the obsolete parts from the previous design! I picked a couple of others out that had higher power or voltage handling capability but there are other parameters that are compromised. I've got some good 500V p-channel FETs around. I'll have plenty of parts to test.
I hadn't thought of returning to a negative voltage. I'd prefer not to but that may be where I end up.
Bandwidth was not very impressive with the pentode input stage so I'm hoping this will do much better, and I'm hoping it will do a little bit better than the 6BN11 in terms of distortion as well. Obviously, I want it all.
It seems like in order to get lots of gain I have to have less current going through the pnp device. That seems fine for me since I am just driving a mosfet follower gate and less current will help me stay in the pnp device SOA.
Well, I guess it is time to order some parts and start playing with them. 🙂
I'll report back with results eventually.
Here's my plan for an initial experiment. I thought more an more about it and started to like the negative rail idea more and more. I know that I will have to construct some very high quality supplies to make this work but that seems doable.
For the initial experiment I'm going to stack some 48V supplies to generate the rails. I just need to answer the question of whether this is a better approach than the CCS-loaded pentode that I am using, so this will get me a simple distortion measurement even if the output is noisy from poor PSRR.
The negative supply has one tremendous advantage: you can use a low voltage driver tube - even with a 50V supply - and still drive an output swing of 220V pp or more.
I have been working on my next Shunt Cascode kit solution. Finally I have a positive regulator that is low-noise, stable and very tough; a new, tough current source, and a design ready to go out to fab for a negative regulator, somewhat like the positive.
I'm looking forward to measuring it up, because the (open-loop) distortion was very low before, and the negative voltage supply promises to improve on it.
I have been working on my next Shunt Cascode kit solution. Finally I have a positive regulator that is low-noise, stable and very tough; a new, tough current source, and a design ready to go out to fab for a negative regulator, somewhat like the positive.
I'm looking forward to measuring it up, because the (open-loop) distortion was very low before, and the negative voltage supply promises to improve on it.
Just a comment: If you use CCS loaded tube, with diode/LED biasing in cathode possible eliminate input capacitor.
Let me know when you are finished with the kit design.
6e5p looked very linear in the lower voltage, higher current region. I hope it is as good as I think it is going to be.
I'm also considering another benefit to the negative rail. I may be able to direct couple this input stage to the output stage mosfet grid driver. This will require some fairly extraordinary care in biasing the system. I'm looking into doing this with a microcontroller.
I think that the extreme gain I'm going for here will make this particular configuration prone to lots of drift over time, so I'll probably keep the input cap and have some sort of active circuit that adjusts bias over time. Plus, I plan on driving the cathode of this stage in the final configuration, like I show in the first post in this thread. That further complicates things.
@Rod Coleman: Looking for your new version as well for the other projects (SE and PP).
@SpreadSpectrum: great job as usual, I'm curious about possible further improvements on your original design.
@SpreadSpectrum: great job as usual, I'm curious about possible further improvements on your original design.
In my experience LED cathode bias and high Gm tubes don't play well together. There is too much tube to tube variation to get good results unless you socket the LED and roll tubes and LED's together. In this case you can spend a lot of time with a box full of tubes and a box full of LED's before finding a few combinations with a suitable plate voltage. Some tube types are worse than others. The WE417 / 5842 is all over the place, so the TSE and TSE-II have a pot across the cathode resistor.
I use this method for decades (for example with D3a/E280F/E180F/E810F, C3g/C3m/C3o) without any disadvantage.
After more than a decade continuos (4-6 hour/day) work doesn't move the operating point in my 300B VAS stage
I think a big issue here is that this is not exactly a CCS-loaded tube. The load is a CCS in parallel with the emitter of the pnp transistor. The CCS is high impedance and the pnp emitter is very low impedance. Therefore, the tube will have almost no say in what its idle plate voltage will be. It will be set by the voltage supplied to the base of the transistor.
Output at the collector of the pnp will be very touchy indeed once the idle point is established. The very high gm tube will be working into a nearly vertical load line. The stage is acting like a V-to-I converter and an I-to-V converter all in one. That's how I think of it at least.
I'm trying to get something close to the 65dB of gain that I get out of the CCS-loaded pentode stage that I'm trying to replace.
I'm prepared to build an active bias circuit. In fact, I've ordered a little Arduino board with some DACs that I think will do the job nicely. In the past, I have done this job with op-amps but I think it can be done much more simply and better with a microprocessor with DACs and ADCs.
Like I stated above, I'm getting the crazy idea that I may be able to direct couple this stage to the mosfet grid driver of the transmitter tube if my bias controller is good enough. Of course, once I have a bunch of ADC channels in the game, I can do a lot more fault detection and use the controller to shut things down if things go wrong in the amp.
Output at the collector of the pnp will be very touchy indeed once the idle point is established. The very high gm tube will be working into a nearly vertical load line. The stage is acting like a V-to-I converter and an I-to-V converter all in one. That's how I think of it at least.
I'm trying to get something close to the 65dB of gain that I get out of the CCS-loaded pentode stage that I'm trying to replace.
I'm prepared to build an active bias circuit. In fact, I've ordered a little Arduino board with some DACs that I think will do the job nicely. In the past, I have done this job with op-amps but I think it can be done much more simply and better with a microprocessor with DACs and ADCs.
Like I stated above, I'm getting the crazy idea that I may be able to direct couple this stage to the mosfet grid driver of the transmitter tube if my bias controller is good enough. Of course, once I have a bunch of ADC channels in the game, I can do a lot more fault detection and use the controller to shut things down if things go wrong in the amp.
I remember Rod suggesting me in one of my threads a gyrator to dc-couple his own driver to the grid of the output tube(s).
I must state that much of my work was done about 15 years ago with CCS loaded triodes, CCS loaded triode wired pentodes, and resistor loaded pentodes where the plate load was rather high in order to extract a lot of gain for use in a 2 stage DHT amp. I have not tried any of the tubes listed as I do not have any.
In the case of the TSE design there is a cathode biased 5842 with a CCS load. The sweet spot for plate voltage in order to get low and primarily 2nd order THD is around 175 volts for 5842's and 165 volts for WE471A's. After selecting an LED for about 175 volts with my best performing 5842, I see huge variations from tube to tube without changing the LED. Some tubes run well under 100 volts resulting in high distortion, while others run well over 200 volts resulting in higher 3H and excessive plate dissipation. Experiments with the usual collection of TV IF amp pentodes (6EJ7 / EF184 and anything pin compatible including the 12BY7) in both triode connection and pentode connection show a similar, though not as severe spread in plate voltages. A trimpot on the screen voltage can work in pentode wired tubes.
Well constructed frame grid tubes may work fine in some applications, but I have done most of my testing with low cost tubes that are readily available everywhere. These often come from several manufacturers with a wide spread in real world Gm. I try to make my designs capable of dealing with wide tolerance parts since they are often duplicated by others of differing skill levels with a different palate of parts to choose from.
Ok, so this is the first time I have used a BJT at tube voltages. The 1N4148 will protect the transistor from violating max emitter-base voltage (-5V). It looks like the emitter-base voltage will be like 48V before the tube warms up. Is there any problem with that? Is there any sort of clamping that needs to take place for positive voltages? The datasheet doesn't even address it.
I've started construction on the breadboard circuit so hopefully I will have some distortion measurements before too long.
I've started construction on the breadboard circuit so hopefully I will have some distortion measurements before too long.
If you mean base-emitter forward bias, that's ok so long as the base current does not exceed the Maximum Rating.
Yes, I see no max base current rating or characteristics that would show how much base current would flow under these conditions. The datasheet doesn't give much.
Before the tube heats up, it is essentially out of the circuit. So the emitter will be connected to a 50mA CCS, the base will be anchored at a voltage 48V below the CCS supply, and the transistor will turn on completely. It will draw 4.5mA at most in collector current. The base is going to be driven pretty hard at startup, before the tube starts drawing current from the CCS.
So is there typically something added to limit base current or clamp the voltage in the other direction? I didn't see anything in any of the examples I've seen around the internet, but I wanted to ask the question to prevent letting the smoke out of these little guys.
The max Ic rating is 200mA, if you use the ZTX558.
50mA (For Ic=500mA, Ic/Ib = 10) of Ib is given in the Vbe (sat) curves as producing a Vbe of ca. 800mV.
So, 50mA will dissipate <50mW in the ZTX.
Does not look like a problem to me.
In practice, I have had no problems, even with lower-power PNPs and similar arrangements...
50mA (For Ic=500mA, Ic/Ib = 10) of Ib is given in the Vbe (sat) curves as producing a Vbe of ca. 800mV.
So, 50mA will dissipate <50mW in the ZTX.
Does not look like a problem to me.
In practice, I have had no problems, even with lower-power PNPs and similar arrangements...
Ok, so it sounds like the base can handle 50mA without too much stress. Thanks for the help on this.
As a rule of thumb, where do you draw the line on dissipation on these? The datasheet says 1W, but do you try to keep it under 500mW under all conditions or something like that?
My method:
Look up the Theta-JA for the package (E-line in this case).
Then decide a high ambient temperature (say 50 Deg. C for inside the amp). Then, apply no more power than would raise the temperature to 100 Deg. C
So, if 200K per W, and 50 Deg. C start, allow 0.25W.
You can go higher, but it decreases long term reliability, and eats into margins.
Look up the Theta-JA for the package (E-line in this case).
Then decide a high ambient temperature (say 50 Deg. C for inside the amp). Then, apply no more power than would raise the temperature to 100 Deg. C
So, if 200K per W, and 50 Deg. C start, allow 0.25W.
You can go higher, but it decreases long term reliability, and eats into margins.