Pros and Cons of power tubes used in parallel per half-wave in a push pull circuit.

[KT88]

Is it good to connect tubes in parallel? I still have a MC2102 with 2 KT88 per phase, so 8 tubes for two channels in total.. Ok, the output impedance is halved in comparison to use just one tube per half wave. The MC2102 uses the same OTs as the MC275 but with the impedance halved, 2, 4 and 8 ohms instead of 4, 8 and 16 ohms outputs.

With the MC2103 I think the driver tube has to "slave" more, which is a disadvantage, I can't detect any overall sonic advantages. On the contrary, (I know that the circuit of the MC2102 is also different, a cascade circuit without coupling capacitors in this circuit area) can it be that one makes a compromise with parallel tubes? 

I'm a technical layman but wouldn't these tubes have to be perfectly equal which doesn't exist in practice? Is this how you get an "average curve" of the characteristic curve of each of the tubes?
Is the zero crossing of the phase badly affected if both tubes are not exactly the same? All questions assuming that the tubes are best matched and biased as possible.

On the other side I have read about OTL amps, never heard one, which are said to have a very good sound using much more tubes in parallel…

[captainbeefheart]

2 hours ago, KT88 said:

Is it good to connect tubes in parallel? I still have a MC2102 with 2 KT88 per phase, so 8 tubes for two channels in total.. Ok, the output impedance is halved in comparison to use just one tube per half wave. The MC2102 uses the same OTs as the MC275 but with the impedance halved, 2, 4 and 8 ohms instead of 4, 8 and 16 ohms outputs.

With the MC2103 I think the driver tube has to "slave" more, which is a disadvantage, I can't detect any overall sonic advantages. On the contrary, (I know that the circuit of the MC2102 is also different, a cascade circuit without coupling capacitors in this circuit area) can it be that one makes a compromise with parallel tubes? 

I'm a technical layman but wouldn't these tubes have to be perfectly equal which doesn't exist in practice? Is this how you get an "average curve" of the characteristic curve of each of the tubes?
Is the zero crossing of the phase badly affected if both tubes are not exactly the same? All questions assuming that the tubes are best matched and biased as possible.

On the other side I have read about OTL amps, never heard one, which are said to have a very good sound using much more tubes in parallel…

 

Great questions but not easy to answer in way many will understand, I will try and explain as best as I can.

 

With power tubes in parallel, specifically a Class AB push pull amplifier like you are discussing you want them to be matched for both bias current and transconductance. For bias current, if they are not matched the issues will be most apparent at higher powers and transients. Let's say you have two KT88's in parallel and they should be biased at 60mA, if one biases up at -45v grid to cathode while the other needs -40v, when signal is applied the one that is biased at -45 it can reach cut-off (no current) sooner and the one biased at -40 can reach grid current sooner. The latter being the most likely, this means the one tube that reaches 0v grid to cathode sooner will draw grid current clamping off the signal at both grid and plate output while the other is still in circuit. As for the one more negative, -45 if it reaches cut-off sooner (no current class B ) it will also drop out of circuit. Both conditions will cause the amp to wimp out at higher powers as now one of the parallel paired tubes is no longer in circuit and you just have one tube driving the load and not two. Transconductance will effect all performance, the tube with a higher transconductance will pull more current through the tube for the same swing grid to cathode. This issue has a whole host of problems too complex to explain here. 

 

If you want to parallel tubes lets view them as a one tube model because that is what is happening. The parallel tube's mu stays the same so you do not have to worry about the drive signal working any harder with two tubes in parallel vs one single tube. Input impedance is set by the grid leak resistor and since mu is the same the two parallel tubes will require the same amount of swing to bring to full power as the single tube. So what happens when you parallel two tubes? The model would be one tube except the transconductance is doubled and the internal impedance is halved.

 

Since;

 

mu=gm*rp

 

gm=transconductance

rp=internal plate impedance

 

when you double gm and halve rp you can see how the math proves mu is exactly the same. The benefit is you now have a "tube" that has double the gm, which means it pulls twice the current for the same deltaV grid to cathode as the single tube.

 

Example, gm=10mA/V, for a deltaV of 1v grid to cathode you get 10mA delta plate current. With them in parallel the gm doubles to 20mA/V so for the same 1v delta grid to cathode you get 20mA delta plate current. So for the same signal swing we are moving double the current through the load. It's as simple as that.

[KT88]

captainbeefheart, thank you for the very well worded explanation. 
First on the simple aspects. That which you describe in the last third of your text is what I understood the general benefit to be... "a" "new" tube that has twice the current delivery capability by connecting two tubes in parallel. 

 

To the first part of your explanations. I understood that both - grit current - and - cut off - are respective corner points so that one of the two tubes is "saturated" at one or the other corner point sooner or later than the other tube so that working together beyond these two respective points is not possible. One tube leaves its sister in the lurch...This may not be a correct rendering by me but in the end it could mean that the possible bandwidth of the amplification factor offered by a single tube is limited at least at larger amplitudes when two parallel tubes are not exactly equal.
But I understood your explanation to mean at this point no difference in sound...within the range where both tubes work together. So it's more of a shared power range issue than a sonic limitation.

 

Even if the problem of different transconductance is technically too complex at this point (I'll google it) can this problem cause a sound change?

To sum it up, possible problems of parallel connection are more the "quantitative" ones described above (e.g. narrowing the common working range) but can there also be sonic changes or degradation unlike single tubes per half-wave? in the sense of  "Qualitative" issues.
Without having to describe it technically if it overtaxes our general knowledge, but if so, what kind of sonic changes can be related to different transconductance? How can these sonic changes be described? Is it e.g. less fine timing, coloration, floppier impulse, etc.? 

 

Thanks for your great help!

 

[captainbeefheart]

40 minutes ago, KT88 said:

Even if the problem of different transconductance is technically too complex at this point (I'll google it) can this problem cause a sound change?

To sum it up, possible problems of parallel connection are more the "quantitative" ones described above (e.g. narrowing the common working range) but can there also be sonic changes or degradation unlike single tubes per half-wave? in the sense of  "Qualitative" issues.
Without having to describe it technically if it overtaxes our general knowledge, but if so, what kind of sonic changes can be related to different transconductance? How can these sonic changes be described? Is it e.g. less fine timing, coloration, floppier impulse, etc.? 

 

The easiest way I can explain it is first in the two tubes rolled up into one model, since the transconductance of each tube sums to the total transconductance the net "new single" tube model can possibly have less total transcondcutance so less current, less power and gain. Example both tubes should have a gm of 10mA/V but one is on the low side of 8mA/V, total gm is now 18mA/V instead of 20mA/V. One tube is working harder passing more current vs the other. Or thinking of it reciprocally the weak tube is holding the rolled up lump sum "singe tube" back, they are not working together well to form the optimal single tube model. When speaking in terms of how things sound, well to my ears when I take the time to match tubes very precisely for parallel use the total sound is less veiled, with better sound stage and imaging between instruments and vocals. For example I purchased a box of 50 power tubes for my amp which gives me the means to get really good matched pairs for parallel use.

 

On a side note, the 2A3 tube is actually two 45 tubes in parallel inside one bottle. Take a look at the datasheets, the 45 has rp=1600 and gm=2500uMhos (2.5mA/V). The 2A3 has rp=800 and gm=5000uMhos (5mA/V). Neato

[KT88]

15 minutes ago, captainbeefheart said:

When speaking in terms of how things sound, well to my ears when I take the time to match tubes very precisely for parallel use the total sound is veiled, better sound stage and imaging between instruments and vocals. For example I purchased a box of 50 power tubes for my amp which gives me the means to get really good matched pairs for parallel use.

You say "the total sound is veiled, better sound stage and imaging between instruments and vocals". I would understand "veiled" negatively but the other descriptions are very positive?

Veiled also means obscured or covered up in German. So that it loses transparency. How does that go together with better stage imaging etc.?

 

[captainbeefheart]

3 minutes ago, KT88 said:

You say "the total sound is veiled, better sound stage and imaging between instruments and vocals". I would understand "veiled" negatively but the other descriptions are very positive?

Veiled also means obscured or covered up in German. So that it loses transparency. How does that go together with better stage imaging etc.?

 

Sorry it was a typo, I went and edited it to say "less veiled" 🙃

[KT88]

36 minutes ago, captainbeefheart said:

Sorry it was a typo, I went and edited it to say "less veiled" 🙃

So, if the tubes really have the same properties then the parallel connection is even a sonic advantage. I can understand it in the way that what the two tubes achieve together is similar to more torque in a gasoline engine. So everything is less strained. Interesting that a 2A3 consists of two 45's connected in parallel.

 

 Maybe it also has to do with the fact that I need less turns in the output transformer for the same output impedance, and thus the signal could be additionally less negatively influenced by hysteresis or ESR?

 

Another point related to the topic. To drive two parallel power tubes you need more driver current?

The Mc2102 has to my humble knowledge only one 12AT7 for both parallel tubes. With the MC275, one 12AT7 has to drive only one tube at a time.  To put it psychologically (which is my profession) we have a stress shift from the power amp tube section to the driver?

So do I need to look out for the best 12AT7 to drive two KT88 in parallel or am I worrying too much?

 

Before the MC2102 came out circa the year of 2001 Mcintosh had released an anniversary amplifier. The MC2000. There the drivers were realized by transistors with high current capability (but I don't want transistors in a tube amplifier signal path). Can a 12AT7 do this job as well as sand?

 

I feel bad for having so many questions. But maybe other forum members will find it interesting so they can participate.

 

[captainbeefheart]

16 minutes ago, KT88 said:

Another point related to the topic. To drive two parallel power tubes you need more driver current?

The Mc2102 has to my humble knowledge only one 12AT7 for both parallel tubes. With the MC275, one 12AT7 has to drive only one tube at a time.  To put it psychologically (which is my profession) we have a stress shift from the power amp tube section to the driver?

So do I need to look out for the best 12AT7 to drive two KT88 in parallel or am I worrying too much?

 

The input impedance of the power tubes is set by a resistor value not the actual tubes, this is for Class 1 operation like the majority of amps. Class 2 operation can operate the power tubes into grid conduction. So the drive requirements for two tubes in parallel is the same as a single tube. Tubes are voltage controlled devices unlike BJT's which are current controlled devices, beta multiplies base current into collector current. A tube changes plate current proportional to it's grid to cathode voltage. This is what we are changing really when driving a tube is the grid to cathode voltage. Yes a 12AT7 is plenty fine driver for a single KT88 or two in parallel.

 

22 minutes ago, KT88 said:

So, if the tubes really have the same properties then the parallel connection is even a sonic advantage.

 

Yes, for the most part a tube with low plate impedance and high transconductance is going to perform better into tougher loads vs a high plate impedance low transconductance tube. You can see from the very first triodes made up too the sweep tubes for TV's that they originally had very low transconductance and as time and technology advances you see the height of tube technology with amazing characteristics. A large majority of us tube nerds actually build amps with TV sweep tubes because they are vastly superior to standard audio receiving tubes. This is mainly for pentode type performance where they really are incredible. Of course they can be wired like triodes also, for example if you wire a 6AV5GA as a triode they have practically identical properties as a 2A3, some actually came wired as triodes for direct replacements for 6B4G.This is well known in underground tube engineer networks where you can make an amazing push pull triode power amp that sounds like an expensive 2A3 push pull amp but instead using $3 6AV5GA sweep tubes.

 

The reason many use transistors for driving tubes is because if you want to have the best performance you want direct coupling into Class 2 operation. Transistors are superior at delivering a low impedance source to drive the tube grids into conduction without wimping out.

[captainbeefheart]

A little history for anyone not in the know.

 

The 6A3 is a 6 volt filament version of the famous 2A3 DHT.

 

The 6B4G is an octal base version of the 6A3.

 

Early 6B4G replacements were 6AV5GA tubes internally connected screen and plate relabeled as 6B4G because they knew the triode strapped 6AV5GA will perform identical to the 6B4G. So basically the cheap triode wired 6AV5GA is a great alternative to using 2A3 DHT tubes.

 

I actually have one old 6B4G marked tube that is really an internally triode wired 6AV5GA sweep tube.