I agree that CCS will improve the linearity of LTP, but since the phase splitter is inside the feedback loop, this improvement can not be realized in practice. The performance of the "simple" LTP is already exceeding the performance of the output stage and usually the voltage amplifying stage.
So improving the LTP from the existing performance (100Vpp out with 0,5 % THD) do not show in total performance of the amplifier.
Funny expression: "force AC balance with unequal resistors "
So we force the volume to lower or higher level when we use the volume pot ?
True, a potentiometer is not the most reliable component, but fortunately it can be replaced with fixed resistors when the required value is known.
I have observed several times that when using even matched output tubes and well adjusted bias, still the overall distortion minimum will be found when the drive voltages of output tubes are not equal.
To find this minimum, which seems to me as an optimum working condition of the output stage, we really need to "force" the amplifier into this condition.
With fixed anode resistors of LTP this condition will not (allways) be
achieved.
Which is more preferable, to optimize the phase splitter alone or phase splitter and output stage together ?
Well, why not both; LTP with variable anode resistor and CCS.
No, not linearity, but balance. The balance of the signals, balance over time and tube aging, balanced source impedances to the output stage.
Do you think I am wasting my time considering the 6M11? My thought was to use the pentode as the first stage and then feed either into another triode then into a phase inverter or pentode then a dual triode phase inverter. Sorry I am asking these questions, I have been out of this way too long. I went from being a EE to a Network guy.
Thanks,
Ray
Thanks,
Ray
Yes, also linearity, i.e. lower distortion. There are published test results showing this. I have not tested with CCS, but read an article.
I can find the link if required.
I would expect some comments about my previous point of view concerning certain required unbalance of the output of LTP in several (many) cases when the amplification of the output tubes differs from each other.
Or do I assume that most of you have not even noticed this kind of feature of typical output tubes.
The phase splitter is only one component in the well designed push pull amplifier.
The "over-optimization" of this single stage alone is not good engineering
if the designer do not see it's impact to total performance of the whole amplifier.
If what you're saying is "Output tubes don't match perfectly," I absolutely agree. If you're saying, "Because output tubes don't match perfectly, it's fine to drive them from unequal source impedances from a circuit that changes balance with different tubes and with tube aging," no, I can't agree.
Overall circuit linearity improves with a balanced phase splitter for the reasons I outlined. The phase splitter itself (i.e., distortion measured at each plate) does not significantly become more linear with a CCS in the tail.
Overall circuit linearity improves with a balanced phase splitter for the reasons I outlined. The phase splitter itself (i.e., distortion measured at each plate) does not significantly become more linear with a CCS in the tail.
I have found that the two sections of the OPT is where the disagreement lies. Sometimes the lowest distortion requires a slight AC and DC imbalance. I use a CCS in the tail of my LTP's because they are usually not inside a feedback loop. I assure equal and low impedance drive impedances by using mosfet followers, but there are trimmer pots to assure adjustment where needed. Lowest higher order harmonics don't usually coincide with lowest THD, since the second will null sharply but 3H and 5H drop off slowly.
Unless you have time and equipment for tweaking and a box full of tubes, just use a 6SN7, 6CG7 or a 12AT7 and be done with it! I have been experimenting with pentode LTP's (like the red board) and found some good results, but be prepared to tweak a lot.
There are things worth worrying about and things that are not. This is one of the things not worth worrying about. Sure, VT finals differ. So do transistors. And they will change with use and age. As for OPTs, these, too, are not exactly balanced either. The coefficient of coupling between turns is never 100%, and so the center tap is never precisely at the electrical neutral, though is pretty close if the OPT is half way decent.
As for mismatched finals, the Radiotron Designers Handbook describes an experiment where they stuck a 2A3 in one hole and a 45 in the other. The resulting distortion from even that severe mismatch was a lot less than what you'd expect.
About 15 years ago I was making "top secret" modifications to some local musicians guitar amps. One of my favorites was two dissimilar output tubes. Something like a 6L6GC and an EL34 (or even a stout 6V6GT) in a Bandmaster. The Fender amps have a tiny OPT so the imbalance leads to frequency dependent saturation. The same trick sounded bad (or at least different) on a Marshall.
The amp would sound reasonably normal at low volume but get a rather creamy distortion when cranked. A skilled player could adjust the amp to control the distortion with the guitar's volume control or even his playing style without big changes in the overall sound level.
Below is the link to earlier mentioned web-page.
Practical Phase Inverters.
There is a LTP built and measured without and with CCS (pentode).
First circuit has THD1 = 0,34 % (1. anode) and THD2 = 0,66 % at Vo = 70 Vpp.
Second circuit (with CCS) has THD = 0,11 % at both anodes.
So in this case the improvement in THD was from 10 to 16 dB due to CCS.
I agree that CCS is an impovement to LTP phase splitter.
However, a good construction needs a point to adjust AC-balance as well.
The perfect AC balance at output tube anodes is more valuable than at the anodes of LTP.
Practical Phase Inverters.
There is a LTP built and measured without and with CCS (pentode).
First circuit has THD1 = 0,34 % (1. anode) and THD2 = 0,66 % at Vo = 70 Vpp.
Second circuit (with CCS) has THD = 0,11 % at both anodes.
So in this case the improvement in THD was from 10 to 16 dB due to CCS.
I agree that CCS is an impovement to LTP phase splitter.
However, a good construction needs a point to adjust AC-balance as well.
The perfect AC balance at output tube anodes is more valuable than at the anodes of LTP.
An LTP phase splitter has two input signals: a differential mode signal equal to the input, and a common mode signal equal to half the input. We only want the differential mode signal at the output. Roughly speaking, the differential gain is mu (i.e. mu/2 to each anode). The common mode gain is Rload/(2xRtail), assuming Rtail is much larger than 1/gm. So a CCS will force the common mode gain down to nearly zero. However, the output stage will largely only respond to the differential mode signal (due to the OPT) so eliminating the common mode signal might not be as important as people think.
This means that when measuring phase splitter distortion you should really measure the differential output, not the two separately. The importance of the common mode signal depends on the degree of balance in the output stage, and how non-linear the output valves are. The common-mode signal can create distortion before it is balanced out.
My guess is that a CCS is much more important when using low mu valves in the LTP, as the differential gain will be smaller. So use a CCS with 6SN7, but a resistor is fine for ECC83. Does that settle the CCS vs. non-CCS argument?
This means that when measuring phase splitter distortion you should really measure the differential output, not the two separately. The importance of the common mode signal depends on the degree of balance in the output stage, and how non-linear the output valves are. The common-mode signal can create distortion before it is balanced out.
My guess is that a CCS is much more important when using low mu valves in the LTP, as the differential gain will be smaller. So use a CCS with 6SN7, but a resistor is fine for ECC83. Does that settle the CCS vs. non-CCS argument?
If you're comparing Figure 4 to Figure 6 (which correspond to the numbers you're citing), note that they are using different tubes and different operating points. Attributing that difference to the CCS is, I believe, inapt.
The other thing one must take care about when measuring these circuits is that the measurement itself will upset balance and give misleading readings unless the load is maintained equal at both outputs.
Dave, I agree that high mu tubes are less critical as regards the length of the tail. But even with a 12AX7, you get a built in imbalance of at least 1%, generally several times more in practical circuits with finite plate loads.
I would not worry too much about a few % imbalance. Some floating paraphase PS in popular circuits have 5-10% imbalance, because low mu valves are used, but people still seem to like them.
Well, some people seem to like SET circuits with cringe-worthy linearity and pitiful bandwidth! For those of us (and I know you're one) who want the amplifier's output to replicate the input without editorializing, having a 5% imbalance and the consequent distortion compromises this goal.
Yes, of course. I would want somewhat less than 5% imbalance. However, let's do a rough calculation - I am thinking out loud here so as I start I don't actually know where I will finish.
Assume a low distortion PS, but with a common mode output equal to 5% of its differential mode output. I'm not sure whether people would call that 5% imbalance or 2.5% imbalance. Assume a balanced output stage - any imbalance just lets the common-mode signal straight through. Now assume that the output stage is biased so that over the full signal swing the gain of one half varies from zero to twice normal - this is not as bad as it seems as any second-order distortion from this will be cancelled in the OPT (that is how PP works). However, second-order products from the combination of the differential mode and common mode signals will not cancel but add.
If I have my factors of 2 in the right place, then this will result in 2.5% second-harmonic distortion and 2.5% DC shift (the two always occur together - trigonometry ensures this). A real output stage will not normally be as bad as assumed above, unless it suffers from significant bias shift with signal e.g. cathode bias or CCS bias.
So let's assume a ball-park figure of 1-2% 2nd arising from a 5% PS imbalance, assuming distortionless PS, perfectly balanced output, and typical output valve non-linearity. I think that means that, unless you are building an effects box rather than a reproducer, you want PS imbalance to be down at the 1-2% level or better. That means ECC83 with a resistor tail is just about good enough (but only just), any other valve (with smaller mu) needs a CCS.
Note that you can good automatic balance from a low mu valve using a CCS in the LTP. There is no corresponding trick for the floating paraphase - there you need high mu or fiddle with resistor values to correct the inherent circuit imbalance. So its LTP or cathodyne!
Assume a low distortion PS, but with a common mode output equal to 5% of its differential mode output. I'm not sure whether people would call that 5% imbalance or 2.5% imbalance. Assume a balanced output stage - any imbalance just lets the common-mode signal straight through. Now assume that the output stage is biased so that over the full signal swing the gain of one half varies from zero to twice normal - this is not as bad as it seems as any second-order distortion from this will be cancelled in the OPT (that is how PP works). However, second-order products from the combination of the differential mode and common mode signals will not cancel but add.
If I have my factors of 2 in the right place, then this will result in 2.5% second-harmonic distortion and 2.5% DC shift (the two always occur together - trigonometry ensures this). A real output stage will not normally be as bad as assumed above, unless it suffers from significant bias shift with signal e.g. cathode bias or CCS bias.
So let's assume a ball-park figure of 1-2% 2nd arising from a 5% PS imbalance, assuming distortionless PS, perfectly balanced output, and typical output valve non-linearity. I think that means that, unless you are building an effects box rather than a reproducer, you want PS imbalance to be down at the 1-2% level or better. That means ECC83 with a resistor tail is just about good enough (but only just), any other valve (with smaller mu) needs a CCS.
Note that you can good automatic balance from a low mu valve using a CCS in the LTP. There is no corresponding trick for the floating paraphase - there you need high mu or fiddle with resistor values to correct the inherent circuit imbalance. So its LTP or cathodyne!
And you guessed wrong. There's no need to guess at all since the matter's long been settled. Cathode Phase Inversion Pt. I and Cathode Phase Inversion Pt II. Schmidt already did all the math and what results he got is:
Vgk1/Vgk2= 1 + 1 / (Gm X Rtail)
If it wasn't for that extra term, then the LTP would have perfect phase-to-phase balance. Therefore, you need to make that Gm X Rtail term as large as possible. Look at the nominal values:
6SN7: Gm= 3.0mA/V
12AX7: Gm= 1.6mA/V
The 12AX7 is almost twice as horrible as the 6SN7 (since the 12AX7 gets its high-u by driving up the rp).
It's also the reason why you can get away with short tails when implementing LTPs with BJTs. Even at a very modest Ic= 1.0mA, a BJT will have a Gm= 38.46mA/V -- not many VTs that can touch that. Since you can virtually drive up the Gm at will, you're a lot less concerned with using long tails for BJT LTPs (although these too will show improved performance with active tail loads).
With VTs, the only thing you have any real control over is Rtail, and you can maximize that by using active tail loads. As I said before, I prefer cascoded BJTs for this purpose.
This is clear.
This is not that clear.
With 12AX7 you can use much bigger Rtail than with 6SN7, since 6SN7 requires essentially higher working current.
AFAIC, it's totally irrelevant. I'd stick a cascoded BJT CCS under the cathodes regardless if I was using 6SN7s or 12AX7s. Much less voltage margin required, and that'll give a higher effective Rtail than passive tail loading.
It would only be a concern if you have some psychological aversion to sand, but I don't since I started out designing solid state amps anyway and got into hollow state later.
Since the term
In the case of typical resistive tail, with 12AX7 you can build better balanced LTP phase splitter than with 6SN7.
The reason is the term (Gm X Rtail), which in case of 12AX7 is bigger than in case of 6SN7 in practice. (say, 1,6 mA/V * 47 kohms vs. 3,0 mA/V * 15 kohms)
So, 12AX7 "wins" by 75 vs. 45
I see that the only (but good) reason to use CCS in LTP is the lower distortion it gives.
The balance can allways be adjusted to meet the requirement of the rest of the power amplifier, i.e. the matching of output tubes and OPT.
is essential for the balance of LTP, then it is wrong to emphasize the value of Gm only.
In the case of typical resistive tail, with 12AX7 you can build better balanced LTP phase splitter than with 6SN7.
The reason is the term (Gm X Rtail), which in case of 12AX7 is bigger than in case of 6SN7 in practice. (say, 1,6 mA/V * 47 kohms vs. 3,0 mA/V * 15 kohms)
So, 12AX7 "wins" by 75 vs. 45
I see that the only (but good) reason to use CCS in LTP is the lower distortion it gives.
The balance can allways be adjusted to meet the requirement of the rest of the power amplifier, i.e. the matching of output tubes and OPT.
Schmitt is wrong. In his first paper he confuses input voltage balance with output voltage balance. He calculates the former, then treats it as the latter. In his second paper he misquotes from his first paper (this is then quoted above by Miles Prower). Just because a paper has appeared in a peer-reviewed journal it doesn't mean it is necessarily correct. Schmitt takes account of mu in calculating the grid-cathode voltage of each valve, then forgets that mu also has a role in calculating the anode voltage too. This means that his formula, when correctly applied, only gives the input balance not the output balance.
If you want the truth go to Langford-Smith (ch 7 sect 2). He says "The two voltages differ by a small amount, which can be reduced by using a large value of Rk and/or by using high mu valves." L-S's formula includes mu, as it should. He therefore reaches the same conclusion as me: high Rk or high mu.
I hope no commercial amps have been designed using Schmitt's incorrect results!
If you want the truth go to Langford-Smith (ch 7 sect 2). He says "The two voltages differ by a small amount, which can be reduced by using a large value of Rk and/or by using high mu valves." L-S's formula includes mu, as it should. He therefore reaches the same conclusion as me: high Rk or high mu.
I hope no commercial amps have been designed using Schmitt's incorrect results!