Icsaszar,
With a perfect CCS, and Zero grid currents, the Total current in the cathodes = CCS current.
And, the Total plate current = Total cathodes current, = CCS current.
This is not a nuclear reaction. None of the electrons convert from mass to energy. Conservation of electrons.
Therefore, every time one plate current Reduces by 100uA, the other plate current Increases by 100uA.
Given precision matched plate loads, the signal amplitudes of the plate voltages are exactly equal; but are exactly 180 degrees from each other.
This means the 2nd harmonic distortion of each triode is cancelled.
I hope that makes the operation of my favorite phase inverter understood by all readers of this thread.
The problem with phase splitters using Pentode wired Pentodes & a CCS, there is no guarantee that the Delta current of the screens is equal.
Can you understand, that disturbs the equality of the plate currents.
With a perfect CCS, and Zero grid currents, the Total current in the cathodes = CCS current.
And, the Total plate current = Total cathodes current, = CCS current.
This is not a nuclear reaction. None of the electrons convert from mass to energy. Conservation of electrons.
Therefore, every time one plate current Reduces by 100uA, the other plate current Increases by 100uA.
Given precision matched plate loads, the signal amplitudes of the plate voltages are exactly equal; but are exactly 180 degrees from each other.
This means the 2nd harmonic distortion of each triode is cancelled.
I hope that makes the operation of my favorite phase inverter understood by all readers of this thread.
The problem with phase splitters using Pentode wired Pentodes & a CCS, there is no guarantee that the Delta current of the screens is equal.
Can you understand, that disturbs the equality of the plate currents.
Last edited:
Agreed, with perfectly matched anode loads a perfect CCS forces perfect balance at the anode of a pair of triodes. We can regard grid current as negligible, so signal current just swings backwards and forwards without loss between the two anode loads. So far so good. Now allow the two triodes to be a bit different, perhaps slightly different r_a. The previous operation still occurs, but rejection of power supply interference is no longer perfect because the attenuation to each anode is different. Unfortunately, we're always stuck with that problem.
In my experience the better the tube match the better second harmonic cancellation is in the output voltage taken as a balanced signal from the two plates as a differential signal.
It is not only the DC drift that is a issue in CCS in a LTP.
It is also the AC impedance and linearity of the current source that is important.
The ac impedance of ALL current sources is limited.
The data sheet for the IXCP10M45S states 160K BUT it does not specify what frequency this is.
Is this for a DC step or a 200KHz signal? Who knows?
Has anyone done a network measurement on one of these?
If you have 47K plate loads in your LTP and a and effective 160K cathode resistor (due the the current source impedance) it will be very far from a perfect phase splitter.
The ac impedance in many (ALL electronic) current sources varies strongly with frequency.
Depending on the details of the current source design the AC impedance can drop to very low values at 20Khz and above.
Data sheets for things like the IXCP10M45 or the LM317 provide zero guidance on the impedance of the current source VS frequency.
If the data sheet does not mention a parameter do not expect it to be wonderful news.
Finally there is non-linearity. Electronic current sources are amplifiers setup as a special function.
All amplifiers have distortion, slew limits and noise issues. This becomes part of your signal path when you incorporate electronic current sources.
None of the electronic current sources I am familiar with specify performance data for AC signals.
You are on your own if you put this in the signal path. Does anyone have AC measurements on the IXCP10M45?
Ok you can design a good electronic current source for wide bandwidth high quality audio but it is non-trivial.
So there is after all something to be said for a CCS made with only a high value resistor feed from a large negative power supply voltage.
The CCS only needs to be very high impedance compared to 1/gm for good balance. Pentodes could be used if the Frank Blohbaum (mit der umlaut) cascoded G2 scheme is used.
All good fortune,
Chris
All good fortune,
Chris
Interesting read that suggests very high impedance in the test results.
The test setup is pretty impressive for being able to resolve such a high impedance.
Going to have to make one up.
The chart for IXCP 10M45S shows about 8meg ohms at 10Khz, way better than the data sheet suggests at 160K and a very low capacitance of 2pF.
Wonder if they are all that good? Perhaps my aversion to using them is not well placed.
I spiced one example current source listed as "IRF820 with 9 battery".
Spice suggests about 1Meg at 10Khz or about 1/4 what was measured.
Still this is in the ball park and tolerance and model accuracy could be the difference here.
Spice suggests about 5Meg at 1Khz and the chart indicates about 35meg ohms so the delta is getting a bit larger here to wave away the differences.
Some work is needed here to find that difference.
Still a single IRF820 works better than I would expect and better than a number of other MOSFETs I have tried.
I would like to see the actual circuits used for the current sources so I could replicate more of the tests.
If I get similar results then these are circuits I would want a adopt.
There is still the issue of distortion but if the CCS is really over 35Meg ohms will that matter much?
Also slew limiting at high frequency.
Heater to cathode capacitance will limit the CCS impedance at high frequency.
5.4pF for two cathodes of a 6DJ8 or about 2.9Meg ohms at 10Khz.
I like to consider CCS impedance out to the third harmonic from 20Khz to in the hope the circuit will still be well balanced for the feedback loop harmonic distortion cancellation.
In that case we are down to about 491K impedance @60Khz for a 6DJ8 triode for the two cathodes alone.
Give the cathode capacitance the CCS requirements become less important.
If you do not use feedback it gets easier for the front end circuits.
We still measure perfect balance of the output voltage swing on the anodes, if the LTP is not loaded. As soon as the next stage represents a finite AC load, the unequal r_a will result in unequal AC voltage on the anodes due to attanuation between r_a, Ra (anode resistor) and load impadance.
Hi Chris,
would you please explain? Silly me doesn't know...
Feeding both screens via a common resistor, or connecting them by a capacitor if individual resistors, doesn't suffice?
Best regards?
If I interpret this correctly, I must disagree. If load impedances are equal, output voltages will be equal and opposite polarity, independent of amplifier characteristics. This falls out from a CCS, which forces the sum of amplifier currents to be constant, so must divide into equal load impedances with a constant sum of voltages (zero if perfectly linear).
All good fortune,
Chris
All good fortune,
Chris
6A3sUMMER raised the issue above. If G2 currents were constant, we could do that, but the slight change in G2 current with signal corrupts balance. Blohbaum's (mit der Umlaut) cascode G2 restricts all cathode current into the anode current*, so maintains the "CC" part of the CCS.
All good fortune,
Chris
*He actually uses a BJT, so its base current is a leak. A JFET or vacuum valve instead would eliminate this small error.
The first CCS-tailed pentode LPT I built used type 6AU6. These have a lot of screen current vs plate current. I began building with an emphasis on finding tubes with minimum g2 current. The 6AU6 amplifiers worked quite well, and so did the ones with minimal screen current. I suspect the relationship between a dPlate current to a dScreen current is what lets these work acceptably. Building pentodes( triode as input, MOSFET as screen ) leaves near infinite output Z( nominally 'plate resistance'), and with no g2 current the amps also work quite well.
With the constructed pentode, the votlage divider reference for the gate can hang from a pot. This will allow an adjustment of what is effectively screen voltage, and the LTP can be balanced( voltage-wise ) very closely. Combined with an adjustable current( variable R-set ) the operating point can be set quite precisely.
Douglas
With the constructed pentode, the votlage divider reference for the gate can hang from a pot. This will allow an adjustment of what is effectively screen voltage, and the LTP can be balanced( voltage-wise ) very closely. Combined with an adjustable current( variable R-set ) the operating point can be set quite precisely.
Douglas
This is a very interesting discussion. I admit it makes my head hurt a bit.
Instinctively I feel that the match between the two triodes must make a difference in the output signal balance.
But spice says my instincts are wrong.
With a perfect current source even with grossly mismatched tubes spice says the output stays in perfect balance at lower frequencies.
Balance starts to decline at 10Khz as stray capacitance affects that so important current source.
Again with a perfect current source, with grossly mismatched tubes and a large common mode signal on the input grids the output still stays in perfect balance at lower frequencies.
My brain is objecting but spice is telling me I am just wrong, balance is still maintained.
Lower the quality of the current source to for example 1Meg ohms and the balance starts to degrade quickly with mismatched tubes.
So my take is...
1) The quality of the current source in the cathode circuit is where it's at.
-Given a CCS impedance of 10meg ohm or more the tube balance does not seem to matter in a meaningful way.
-With a CCS below 1 meg tube match begins to seem worthwhile.
2) Stray cathode capacitance
-The stray cathode capacitance becomes a key issue in maintaining balance at high frequencies in the presence of tube imbalance.
Keeping this as low as possible is important to a wide bandwidth amplifier's high frequency balance with tube mismatch.
3) Common mode input signal.
With a perfect CCS common mode input signals seems to have no effect on output balance even with grossly mismatched tubes.
However the quality of the CCS becomes very important in the presence of a common mode signal with mismatched tubes.
So the quality of the CCS seems the key.
Given a sufficiently high AC impedance across a wide enough bandwidth on the CCS with matched plate load impedance little else seems to matter in maintaining LTP balance.
My head is still objecting but that is part of the learning process.
My thank you to everyone for the opportunity to discard some dogma.
Instinctively I feel that the match between the two triodes must make a difference in the output signal balance.
But spice says my instincts are wrong.
With a perfect current source even with grossly mismatched tubes spice says the output stays in perfect balance at lower frequencies.
Balance starts to decline at 10Khz as stray capacitance affects that so important current source.
Again with a perfect current source, with grossly mismatched tubes and a large common mode signal on the input grids the output still stays in perfect balance at lower frequencies.
My brain is objecting but spice is telling me I am just wrong, balance is still maintained.
Lower the quality of the current source to for example 1Meg ohms and the balance starts to degrade quickly with mismatched tubes.
So my take is...
1) The quality of the current source in the cathode circuit is where it's at.
-Given a CCS impedance of 10meg ohm or more the tube balance does not seem to matter in a meaningful way.
-With a CCS below 1 meg tube match begins to seem worthwhile.
2) Stray cathode capacitance
-The stray cathode capacitance becomes a key issue in maintaining balance at high frequencies in the presence of tube imbalance.
Keeping this as low as possible is important to a wide bandwidth amplifier's high frequency balance with tube mismatch.
3) Common mode input signal.
With a perfect CCS common mode input signals seems to have no effect on output balance even with grossly mismatched tubes.
However the quality of the CCS becomes very important in the presence of a common mode signal with mismatched tubes.
So the quality of the CCS seems the key.
Given a sufficiently high AC impedance across a wide enough bandwidth on the CCS with matched plate load impedance little else seems to matter in maintaining LTP balance.
My head is still objecting but that is part of the learning process.
My thank you to everyone for the opportunity to discard some dogma.
Attachments
Yes, the quality of the CCS is key, but is always limited by Chk in the LTP. You should also consider common mode signals applied to the HT, not just the input grids or cathode.
I have seen instruments use a method called guarding to remove stray capacitance.
To guard the cathode from heater capacitance one takes the cathode signal runs it through a buffer and applies the buffered version to the LTP heater.
Now with the exact same signal on both cathode and heater the net capacitance will become zero, heater to cathode.
Not sure this is worth the complexity in the audio range.
Last edited:
Yes, guarding is a well known technique. But it does require being able to put a guard conductor between the two effected terminals. In this case, between the heater filament and its enclosing cathode. Alternatively, you can do as you suggest; the capacitance remains the same, but the voltage drop between the two is zero, so the current is also zero. But as you say, it's not worth the trouble.
Huh? The PSRR of the LTP is basically unity, whatever happens.
The balance of the LTP with a CCS tail is not affected by loading, provided the loading remains equal. r_a doesn't come into it.
Last edited:
Actually even if the tubes are perfectly matched they are still mismatched in-circuit, as the LH and RH tubes are obeying different gain laws, and as the RH is being fed from the LH cathode, so via a small loss.
One way to imagine the LTP with perfect current source and mismatched valves is to imagine "unraveling" the LTP by removing the load impedance of the undriven side from B+ and connecting it to ground (pretend that the valve could still work!). What we have left is a strange split-load inverter, which we've long accepted as output matched if the loads are matched.
In a split-load inverter the same current flows through both loads, and in this imaginary inverter the same current flows through both loads, plus a constant. Same current, same loads = same voltages.
All good fortune,
Chris
In a split-load inverter the same current flows through both loads, and in this imaginary inverter the same current flows through both loads, plus a constant. Same current, same loads = same voltages.
All good fortune,
Chris