Why some amps use two triode in parallel?

[Tom Phillips]

...I use typical 68K grid stop resistor...

Most (just about all) people plug into the #1 input jack and most guitar amps use the classic Fender input circuit in front of the first stage triode. If you analyze the circuit then you will find that the actual value of the input grid stopper, under the conditions I described above, is 34k. All those boutique "designers" who use a 68k resistor on their single jack input amps just copied the part value they saw on an old schematic without understanding what is really going on. See attached.
Cheers,
Tom

[Alan0354]

34K will lower the noise contribution from the grid current by half, but the thermal noise is lower only by 0.707. Still larger than the gain dependent parts.

People keep copying the 68K grid stop.......matter of fact, a lot of people copy the 100K plate resistor!!!! That's the problem, most people in this business have a notion that "It was invented and perfected long time ago".........That the old people know what they are doing. That there is no point of people now to challenge the well established design.....Then accusing people that dare to question and "Trying to re-inventing the wheel"!!!!

In my book, like President Reagan said: "Trust but verify"!!! It's the questioning of the established "truth" that electronics are moving in light speed in the last 30 or 40 years. Just look at the low tech video recording. VCR came in with a bang, just to be replaced in 15 years by Lazar Disc.....to DVD recorder..........to DVDR..........., From a cell phone the size of a brick to Star Tec, to camera phone to a whole damn computer inside the phone now!!!! AND here we are, still talking about 12AX7!!!!

I designed the complete RF front end and analog signal processor of an ultra sound medical scanner with color doppler for Seimens in 1986 to 1988. The unit was a floor standing rack size. Now they have the whole machine in a laptop doing better job than what we had and can be bought for a few thousand dollars!!! The way it is going, I don't think I can find a job even if I want to after 8 years away from the field. I think I am being left behind, way behind.

Electronics theory has evolve so much in the last 30 years and is still constantly being challenged and rewritten. Theories that was not even known 30 years ago become basic foundation in introduction of electronics text books!!! I spent 6 years after retirement, studying 3 hours a day 5 to 6 days a week on math and electronics, and I am only scratching the surface.

[Merlinb]

Blanket saying parallel two tubes lower noise by 3 dB not only is misleading, it is irresponsible and WRONG. Anyone saying that NEVER work a day in low noise design. AND you can take this to the bank.

No. Saying that two amplifiers* in parallel will improve the signal/noise ratio by 3dB is a fundamental engineering rule of thumb, and it is true, and every low-noise engineer know this rule by heart. Except the maths applies to the theoretically ideal situation where you get twice the gain. (Remember, we're talking about improving the SNR, not the absoluate value of the noise, which is of course higher as you add more devices).

But I think you are confusing the idea of a theroetical maxim with a guarentee. Every engineer knows perfectly well that 'ideal' situations do not arise in real life: rules of thumb are not gurantees. In practice you may get a larger or smaller total improvement depending on how 'ideal' your sitation is. But as a rule of thumb, it stands. If you optimise a design for a given noise level, then adding a second amplifier* in parallel with the first will improve the SNR by 3dB.

*Notice I said "amplifier", not individual transistor or valve. It is a simplification to say that an individual device is responsible for the total gain, but in many cases, it is a good approximation (especially with transistor circuits).

Incidentally, your calcuations in post #56 are unrealistic (rp does not generate noise since it is not a physical resistor, and the shot noise of anode current is not equal to 2*q*i*BW, due to the space charge smoothing effect), but I know the point you were trying to make, so I'll let you off! Ultimately you are correct that paralleling two triodes in a guitar amp may not give you any useful improvement in noise, since the grid stopper usually produces a lot more noise than the triodes. Paralleling triodes in a guitar amp is mainly a tonal consideration.

[Alan0354]

No. Saying that two amplifiers* in parallel will improve the signal/noise ratio by 3dB is a fundamental engineering rule of thumb, and it is true, and every low-noise engineer know this rule by heart. Except the maths applies to the theoretically ideal situation where you get twice the gain. (Remember, we're talking about improving the SNR, not the absoluate value of the noise, which is of course higher as you add more devices).

But I think you are confusing the idea of a theroetical maxim with a guarentee. Every engineer knows perfectly well that 'ideal' situations do not arise in real life: rules of thumb are not gurantees. In practice you may get a larger or smaller total improvement depending on how 'ideal' your sitation is. But as a rule of thumb, it stands. If you optimise a design for a given noise level, then adding a second amplifier* in parallel with the first will improve the SNR by 3dB.

*Notice I said "amplifier", not individual transistor or valve. It is a simplification to say that an individual device is responsible for the total gain, but in many cases, it is a good approximation (especially with transistor circuits).

Incidentally, your calcuations in post #56 are unrealistic (rp does not generate noise since it is not a physical resistor, and the shot noise of anode current is not equal to 2*q*i*BW, due to the space charge smoothing effect), but I know the point you were trying to make, so I'll let you off! Ultimately you are correct that paralleling two triodes in a guitar amp may not give you any useful improvement in noise, since the grid stopper usually produces a lot more noise than the triodes. Paralleling triodes in a guitar amp is mainly a tonal consideration.

I absolutely disagree paralleling two "amplifiers" in parallel lower the noise by 3dB. Many amplifiers have a set gain, you don't increase gain by parallel two of them. I worked with low noise designs, I guess I can be called a low noise engineer. I designed all kinds of photodiode detectors, lazar detectors, wide band amps, 50ohm input impedance amps for years, I absolutely DO NOT KNOW IT BY HEART. They all have to be closed loop feedback to get a stable gain. You don't.......SAY NEVER parallel two of them together. This is ABSOLUTE WRONG.

You only MAYBE help some if you parallel the input transistor or valve to get lower noise design. Then you have to take into consideration of the environment inside the "amplifier" it is in. That's the reason they have hundreds of different low noise transistors, each for specific application adapting to the input requirement of the environment.

Look at my calculation again in using the rp. For one, the gain dependent part is not even significant AND I never use rp as a noise source. I use rp in parallel with the 100K that reduce the resistor value down to 35K in single and 23K in parallel. this is fundamental argument by others on lowering the thermal noise of the plate resistor. I use the resulting Rout to calculate the noise voltage due to the shot noise of 1mA of plate current

Read this, http://users.ece.gatech.edu/mleach/papers/Parallel.pdf posted by JMF. I did not go through the entire calculations. BUT I checked the first two pages how the author model the noise sources. I agree with the author. Look at the conclusion, I agree with the finding.

[J M Fahey]

Well, I didn't want to have to write this, ever, but can't avoid it any more.

MEF now has been blessed with "a second soundguruman"

Trying to solve real problems or discussions with some friendly disagreement thrown in have been replaced, once more, with unmitigated fundamentalism and discussion just for the sake of discussion.

By the way, boring to death, not forgetting *absolutely* unrelated.

Anybody unduly worrying about 50 micro seconds delay at 20 KHz in a **GUITAR** amp or constantly deviating discussions towards ultrasonic medical detectors, LNA for infrared photodiodes , etc. should have his own head examined .... with an ultrasonic infrared LNA photodiode powered brain scanner

[Merlinb]

I absolutely disagree paralleling two "amplifiers" in parallel lower the noise by 3dB.

And yet the maths disagrees with you. I trust the maths.

I guess I can be called a low noise engineer. I absolutely DO NOT KNOW IT BY HEART.

Then you really ought to; it is as fundamental as V=IR.

They all have to be closed loop feedback to get a stable gain. You don't.......SAY NEVER parallel two of them together. This is ABSOLUTE WRONG.

Whatever the amplifier, you can ultimately model it like an opamp. You can parallel two of them by putting low-value resistors in series with the outputs, for example, to ensure equal current sharing. In that case you're doubling the power gain rather than the voltage gain.

That's the reason they have hundreds of different low noise transistors, each for specific application adapting to the input requirement of the environment.

And yet, if you choose one, and design a circuit with it, then two such circuits in parallel will still have a 3dB better SNR.

Look at my calculation again in using the rp. ... I never use rp as a noise source. I use rp in parallel with the 100K that reduce the resistor value down to 35K in single and 23K in parallel.

No, you calculated the Johnson noise of a 37.5k resistance (1.75uV) which is not the same thing as the Johnson noise of a 100k resistor (2.8uV) attenuated by a factor of rp/(Ra+rp), = 1.05uV. I have to say, for a self professed expert, I would not expect you to make such an elementary error...

Shot noise in anode current is more complicated than your post implies, but I wouldn't expect you to know this, as valve noise is a pretty specialist subject to say the least!

[Alan0354]

Not to make the last post too long. A lot of low noise amplifiers are used for photodiode and Lazar diode. Those are not voltage device. They a current device. It detect the signal and generate a current. the output impedance are very high. Using a voltage amp with low input impedance does not even work as you need a virtue ground to get a linear signal out of those diode.. The amplifier used are TRANSIMPEDANCE amplifiers that has virtue ground input. There is only fixed current generated by the diodes, parallel amplifiers DO NOT DETECT ANY EXTRA CURRENT!!!! You absolutely do not get any more signal by parallel two amps. Being closed loop feedback in all cases of these transimpedance amps. Putting them in parallel will only cause oscillation!!!

Not only input impedance govern the topology of the amp, the frequency of operation absolutely govern the type of front end device used. For low input impedance amps, grid/base current is not important, you can use BJT or tubes for the first stage. For detecting source with high output impedance, any grid/base current will cause noise voltage. You cannot use BJT/tube as the first stage. You have to use MOSFET frontend even though the flicker noise is very high at low frequency.

GaAs type of exotic material transistor has high flicker to 1MHz, but they are very popular in RF application as you use them in GHz area and band limit in the lower frequency to get rid of the noise.

Noise design is APPLICATION SPECIFIC. There is NO BLANKET statement for it.

[Alan0354]

And yet the maths disagrees with you. I trust the maths.


Then you really ought to; it is as fundamental as V=IR.


Whatever the amplifier, you can ultimately model it like an opamp. You can parallel two of them by putting low-value resistors in series with the outputs, for example, to ensure equal current sharing. In that case you're doubling the power gain rather than the voltage gain.


And yet, if you choose one, and design a circuit with it, then two such circuits in parallel will still have a 3dB better SNR.


No, you calculated the Johnson noise of a 37.5k resistance (1.75uV) which is not the same thing as the Johnson noise of a 100k resistor (2.8uV) attenuated by a factor of rp/(Ra+rp), = 1.05uV. I have to say, for a self professed expert, I would not expect you to make such an elementary error...

Shot noise in anode current is more complicated than your post implies, but I wouldn't expect you to know this, as valve noise is a pretty specialist subject to say the least!

Show me the math.

WRONG, read my post again, you cannot parallel most of the amplifier. What are you talking about?

AGAIN, read my post. I never claimed it is a complete picture. In fact I repeat so many time that I only use one source at a time for demo. I left out the Flicker noise as I have no info on that and I said that might again change the whole picture. You wrote the book on tubes, I expect you to know tubes.

SAYING parallel two amps is JUST WRONG. You cannot make general statement like this where you cannot even parallel two amps together. I gave examples already.

I don't even think you can make a statement on this for tube amps.

[Leo_Gnardo]

Well, I didn't want to have to write this, ever, but can't avoid it any more.

MEF now has been blessed with "a second soundguruman"

In a vague attempt to leaven the conversation - - - If I play thru two amps in parallel, a couple of Twins or Marshall half-stacks for instance, it seems my threshold of hearing has been affected afterward to the point of pushing the S/N ratio far beyond 3 dB improvement. IOW I've driven myself deaf. But - - - that's a whole 'nother thing.

I've got to give Alan some points for being tenacious. And making a living designing all those hi tech thingies.

But someone ought to tell Paul Rivera he's been wasting that extra parallel input triode all these years. I'm sure he'll be disappointed.

3 dB. Hardly seems worth fighting for. If the result was 20 dB, now you have my attention.

[Merlinb]

You absolutely do not get any more signal by parallel two amps. Being closed loop feedback in all cases of these transimpedance amps. Putting them in parallel will only cause oscillation!!!

You can always connect two amplifiers in parallel in theory, though it may not be practical to do so in real life. Nevertheless, if you could, you could contrive to get twice the gain (whether voltage gain, current gain, or power gain) and since the noise sources are uncorrelated, you always get your 3dB improvement. I don't know how else to explain it to you. You seem to be willfully misinterpreting what is being said and simply focusing on very specific circuit examples to try to defy some very simple maths. It may not be practical to parallel two amplifiers, but if you can, then you must get a 3dB improvement. You can find this in any noise textbook.

Noise design is APPLICATION SPECIFIC.

Yes, noise design IS application specific, but the mathematical principles of noise are universal.

[Alan0354]

You can always connect two amplifiers in parallel in theory, though it may not be practical to do so in real life. Nevertheless, if you could, you could contrive to get twice the gain (whether voltage gain, current gain, or power gain) and since the noise sources are uncorrelated, you always get your 3dB improvement. I don't know how else to explain it to you. You seem to be willfully misinterpreting what is being said and simply focusing on very specific circuit examples to try to defy some very simple maths. It may not be practical to parallel two amplifiers, but if you can, then you must get a 3dB improvement. You can find this in any noise textbook.


Yes, noise design IS application specific, but the mathematical principles of noise are universal.

I don't have to prove everything I said is right, I just have to to make one single case that the theory is wrong, then the whole theory is wrong. If you are a scientist, you should know better. This is how science and theory work. You can be right 999 times, all it takes is one wrong, the whole theory is wrong. this is how the scientific community works, that's why before publishing any invention and finding, every paper has to go through peer review and it only take one decent, it's over, back to the drawing board.

And if you read my post, you will see I am very very careful to specify I only take specific components and use as an example. In the input, I only use thermal noise of the grid stop and the shot noise of the grid current. I did specified time after time that there are other components, I even talk about flicker noise is the biggest unknown that might overwhelm everything I said. Those are just examples how to calculate noise, that people cannot have a simplistic view and theory in noise calculation.

There is no one lowest noise amplifier, it all depends on application. In the paper JMF posted that I used, it really gives a good idea what to look for. Like for MOSFET, doubling up is like making a bigger MOSFET with wider channel. But that will increase input capacitance and hurt the speed and also in closed loop feed back, it create a pole that hurt stability. Then you have to compensate and increase noise!!!

It is not that I refuse to look at your point of view. You can't just ignore the rest and just talk about the amp alone. If you can make the theory stand, that theory is absolutely useless because there is no practical use as in the real world, the environment is going to kill the nice little amp. Just like the partial calculation so far on the parallel triode, the parallel is loosing big already. Maybe you can provide more info to make better calculation. Info like flicker noise at the input that is not affected by the gain, flicker noise at the plate that is affected by the gain. Also any other noise that I don't know about the tubes. Then we can have the whole picture and calculate the noise contribution of every single one. I believe what I have so far stands, just need to add the rest.

BTW, if you have a mathematical presentation or a link to your assertion, I would like to read it. I am willing to be wrong. I made my case already, I like to see the other side.

[Merlinb]

And if you read my post, you will see I am very very careful to specify I only take specific components and use as an example. In the input, I only use thermal noise of the grid stop and the shot noise of the grid current. I did specified time after time that there are other components.

In your example you only connected two triodes in parallel, so your results did not meet the theoretical ideal since the gain was not doubled. The principle holds if you connect to whole amplifiers in parallel.

Twice the gain, twice the uncorrelated noise sources, 3dB SNR improvement.

[Alan0354]

In your example you only connected two triodes in parallel, so your results did not meet the theoretical ideal since the gain was not doubled. The principle holds if you connect to whole amplifiers in parallel.

Twice the gain, twice the uncorrelated noise sources, 3dB SNR improvement.
[ATTACH=CONFIG]27967[/ATTACH]

I thought I made it very clear that parallel the triode does not double the gain. All my calculations were based on gain=54 for single triode and 78 for parallel. Again, read my posts.

For voltage amp like you show, you can gain S/N like that. but the tricky part is the summing that you just said it happens. If it is tube amps, the summing part become tricky. You cannot sum two voltage and double the voltage!!! You need to use a summing amp, which again back into transimpedance amp or something else. You have to use resistor to transform the two voltages into current, then sum the current and produce the result voltage. Now you have a new set of noise parameters to deal with.

BTW, the one case I posted about transimpedance amp. It is one case, but it just happens that this is the kind of amps that form the backbone of the whole telecommunication industry. Billions and billions of dollars poured into this to invent new material like GaAs, AlGaAs, Indium Asinide etal to create a low noise low capacitance, fast transit time material to create super transistors. All these are lazar diode, transimpedance amps. I worked in SONET equipment 12 years ago and we were doing 10GHz lazar. I don't know whether this is a trillion dollars industry, but it is way into multi billions industry.......all started with that little transimpedance amp I talked about. Want to bet how many of those multi giga hertz detector modules they sell per guitar amp sold?!!!

We even did nitrogen cooled amps for low noise astronomy application. When it gets down to the mud, cooling to close to absolute zero is about the only way!!!!

[Dave H]

Electrometer amplifiers are ‘Transimpedance’ amplifiers with a high value feedback resistor which is the dominant noise source. If you double the value of the feedback resistor the output voltage doubles but the noise only goes up by root 2 so you’ve increased the signal to noise ratio by 3dB. It is just another example of Merlin’s rule of thumb.

[Alan0354]

I have to agree that for certain cases you do gain quite a bit parallel:



In this opamp case where it is a lot more clean cut than tubes, you gain quite a bit if the source impedance is low. But it gets complicated with tubes as you have to go to common grid stage for the summing amp.