Why some amps use two triode in parallel?

[Alan0354]

I have to read the article and go through the formulas, I only read two pages so far, so far so good, the model is exactly what I understand. Another thing, I never deal with low frequency and audio range low noise circuits. My designs is at least from 1MHz to 500MHz wide band and RF. I just never seen or even read anything on parallel.

But let me comment first. It is not straight forward to calculate noise, all different sources has to be root mean summed together.

1) Even the paper said in the first paragraph that the input resistance is very important and it is independent of parallel transistors/tubes.
2)For BJT and even tubes, grid current cause noise. For BJT, base current does not get reduced with more transistors in parallel. In fact it is likely to increase as BETA goes down when you reduce the collector current ( as you parallel more transistors).



I have not finished going through the calculation in the paper, this is from my knowledge:

As I said before, parallel more transistor is like making a bigger transistor. I designed IC before, a big transistor just simply has more emitter and base "fingers" and it's like a few transistors in parallel. The main noise contributing factor is the base spread resistance, so called rx in the paper. You reduce rx as much as possible for low noise transistors. This is where large geometry transistor comes into play. The more cross sectional area, the lower the resistance per unit length. This can be done by parallel transistor or bigger transistor.

Honestly, I have not seen any low noise design using many parallel transistor because of the increase in parasitic is very hard to deal with. I designed the front of ultra sound medical image scanner with color doppler in the 80s, we deal with signal in low uV range and in wide band. We never use parallel transistor because it was not a win after doing calculation. AND try to do parallel in RF circuits!!! You don't just stack transistor up!!! Every connection is a transmission line distribution elements and try to split and sum back all the transistors!!!!

There might be advantage of parallel devices, I need to look at the calculation. But I can tell you, there are a lot of other factors involve that might not make it practical, not to mention you can buy transistor that are designed specially for low noise.

[J M Fahey]

Yes.
What I said applies to Audio/HiFi/Mixers areas, what we usually talk about here , not RF or even Ultrasonics.
Different areas.

[Alan0354]

Guys, I am more correct than you think!!! I am using this article and went through the modelling of the noise source to make sure I agree. I skipped the derivation as it is long and look at the conclusion:http://users.ece.gatech.edu/mleach/papers/Parallel.pdf

1) For BJT, read 1.2 about Constant Stage Current which it is most likely the case we are talking. The ONLY difference is the rx as I talked about in post #16. Paragraph 1.3 in Constant Device Current that you consider -3db for every double the number of transistor is NOT USUABLE. The biggest fraud of this is the assumption of RSB= 0. You CANNOT DO THAT!!!! Everything has input impedance. Particular you put a 68K resistor in series. That is a big noise generator as shown in my calculation later. That's is no difference no matter how many tubes/transistors you parallel. ALSO, the base/grid current that creates the shot noise. See my formula below. This two could swamp the input noise. YOu cannot claim 3dB lower.

2) For MOSFET, read 2.1 and 2.2. Pretty much the conclusion is putting more devices in parallel is EXACTLY the same as making a single bigger transistor with wider gate width.

Read the conclusion in paragraph 3. For BJT, all you can count on is reducing base spread resistance rx......Which is exactly I said in post #16......which is like a bigger transistor!!!! You buy a bigger transistor that has low N.F!!!! For MOSFET, you buy a bigger transistor!!!

As I said over and over and proven if you go through the calculation in the article step by step, IT IS NOT THAT SIMPLE to say 3dB lower. You have to use Excel spread sheet to do and compare. YOu have to take into consideration of every single parameter to calculating the noise. Looking at a single unreal condition is a very simplistic point of view and it's inaccurate.

Actually looking back to my post #13, it is not complete, but what I included was spot on. You don't make simple assumption, you have to look at ALL contributing sources. All the advantage shown is #2 of the three components I named. And using the gain of 78 vs 54, you get very minor difference!!!

Below is the noise calculation that paragraph 1.3 of the article missed by assuming RSB= 0. These can be dominant contribution of noise.



You can see the 68K generates quite a bit of noise!!! I don't know why everyone put this in. I use a 20K and I am about to take it out. I don't need it to create a pole to lower the frequency. Put a cap to ground!!!!

Now that does not imply there is no advantage with two tubes in parallel. A tube is not BJT and MOSFET. I need to know the noise parameters of the tube, the grid current in order to calculate. As I stated repeatedly, you have to put all the parameters in first before you say anything. Just like some application choose BJT, other choose MOSFET to get the lowest noise because the surrounding circuit is everything. It is not that straight forward. Making a blanket statement of 3dB lower for parallel stages is misleading at best.

[Alan0354]

The three attached Excel files are the spread sheet that I created for noise calculations for BJT and Opamp in one of the old project. You can see I had to go through the steps like in the article. There is no obvious answer. More often you don't know the outcome until you put everything in.

You have to save and open with Excel, can't open with internet.

[Gingertube]

Alan,
There is quite a bit of merit in your arguement when we start to consider the relative contributions of all the noise sources. - in the typical guitar amp the largest contributor to noise is those grid stop resistors (usually 68K) on the input stage. If you read Merlin he advocates reducing those grid stops to 10K to reduce noise. Of course noise is dependednt upon bandwidth too, so he also advocates adding a 470pF cap from grid to 0V on the input stage grid to keep the high frequency roll off at the same frequency as you get with 68K and the tube Miller capacitance. With parallel triodes Miller Capacitance will be times 2 so adjustments need to be made for that as well. This is only the case because of the amount of gain following. If we talk about the input stage of a HIFI Power Amp where sensitivity for full output power is say 1V RMS then it no longer applies.

By paralleing triodes the noise introduced by the tube itself will be 1/root2 of the single triode case (flicker noise etc.).

Then there is also the noise from the statistical nature of the grid current. That will depend upon the impedance looking away from the input tube grid(s) that the grid current noise has to develop a noise voltage across. This is likely to depend upon the volume control setting on your guitar.

Cheers,
Ian

[Alan0354]

On yes, I never deny there is an advantage. It is very obvious that there is. I mainly question the 3dB over all. Like you said about the flicker noise which is the major problem for Audio frequency. All I said is you lump everything into calculation.

I got tricked before, that's why I said all these as I had seen one unobvious source can swamp all the good intentions.

Also, you might have thousands of low noise transistors to choose from. How many preamp tubes we have? Pretty much 12AX7!!!! So the only choice after getting rid of the stop resistor is paralleling the tube!!! I am absolutely not trying to discount the merit of paralleling the tube. Even in my original post #13, I clearly saw in point #2 that there is slight advantage already. It is my experience that when comes to noise, you chip away a little bit at a time. I have never seen a bit magic wan.

[Alan0354]

See this old thread on parallel triodes
Gain calcs I did are at post #16
http://music-electronics-forum.com/t29302/
Cheers,
Ian

BTW, this is a very good write up. We just had a thread that I had to reinvent a lot of what you said about the distortion, generating even harmonics by manipulating the plate current and plate voltage. I was pulling it out of my behind, you confirm my assumption. It would have save me a lot of work if I had read your article.

[Merlinb]

I mainly question the 3dB over all.

The '3dB rule' applies to a theroetically ideal case where adding a second device causes the gain to double. The two noise sources are uncorrelated and so increase by a factor of only (1^2 + 1^2)^0.5 = root 2, or 3dB. You therefore have 6dB extra gain but only 3dB extra noise, so the signal-noise ratio improves by 3dB.

In practice the figure acheived will depend on what other things are also changed (often incidentally).

It's a moot point in a guitar amp, however, as the grid stopper noise usually swamps any tube noise.

[Alan0354]

The '3dB rule' applies to a theroetically ideal case where adding a second device causes the gain to double. The two noise sources are uncorrelated and so increase by a factor of only (1^2 + 1^2)^0.5 = root 2, or 3dB. You therefore have 6dB extra gain but only 3dB extra noise, so the signal-noise ratio improves by 3dB.

In practice the figure acheived will depend on what other things are also changed (often incidentally).

It's a moot point in a guitar amp, however, as the grid stopper noise usually swamps any tube noise.

Gain does not double as the rp comes into play. It was calculated of gain of 78 vs 54 of a single tube with 100K plate resistor. This is hardly double. As I posted in #2 of post #13, it's about 3dB voltage gain increase to 3db increase in noise power, only slightly advantage, almost a push.

And when you say -3db, at best is ideal, but this is real life circuit that you have to consider all the contributing sources. Like you said the grid stopper swamps the tube noise, so people should not run out and double up. Get rid of the grid stopper, then look at the next dominant noise source and conquer it one by one. that's all I am saying. As shown in my pass calculations in post #19, noise calculation is never simple and straight forward.

Ultimately, the resistance of the guitar pup has about 10K resistance, that noise source is not going to go away. That is the bottom line noise floor.

[soundmasterg]

The stuff Merlin said was in an ideal gain and noise calculation with perfect devices, resistances, capacitances, etc. Real world calculations will obviously change.

Rather than theorize up and down about it, why not build it and try it for yourself. I've built amps with regular triode gain stages, and with parallel gain stages and the parallel setup does provide useful sounds that are different from what you can achieve otherwise.

Greg

[Alan0354]

The stuff Merlin said was in an ideal gain and noise calculation with perfect devices, resistances, capacitances, etc. Real world calculations will obviously change.

Rather than theorize up and down about it, why not build it and try it for yourself. I've built amps with regular triode gain stages, and with parallel gain stages and the parallel setup does provide useful sounds that are different from what you can achieve otherwise.

Greg

Agree, I notice the sound is a little more aggressive on top of more gain, I think it's because of lower output impedance.

[soundmasterg]

Agree, I notice the sound is a little more aggressive on top of more gain, I think it's because of lower output impedance.

Yeah it could be the lower output impedance....I suppose you could get a similar effect in a way by using a 12BZ7 instead of a 12AX7 in an otherwise identical circuit, since the BZ7 has the same 100 gain at half the output impedance, but if you have a parallel stage with different cathode setups (one cathode with a resistor and cap and the other with just a resistor, or one with an LED to bias etc) they have a way of interacting with each other that makes the parallel circuit kind of harmonically rich which is appealing. When I was working on the one I used in an amp for myself the thing I fought most was the loss of the higher end trebles that gave it a bit of edge....its sort of like some of the shimmer is lost....but it is still a cool circuit to use.

Greg

[Zozobra]

There is is also the argument that if you have a spare triode you may as well put it to use

[GainFreak]

Rather than theorize up and down about it, why not build it and try it for yourself.

+1. Rather than diving deep into an already much chewed theory just build it, tweak it and listen to the result!

[Alan0354]

There is a balance of theory and try it and see. You don't want to be totally one way or the other.

I know, I never have an EE degree, I started out as try and true type. I spent the first 10 years as an engineer on the bench as my office. But when you move up and get into more advanced design, this really falls short. I had to began studying more and more. I feel so strongly that after I retired, I spent the last 7 years studying 3 hours day, 5 days a week on math and theory to fulfill the holes I have. I can assure you it really makes a huge difference. I can see things in a totally different light because I can pull out my pass bench experience and unite the theory and practice together and see things differently.

Like the grounding and noise issues, you will not appreciate that until you studied electromagnetic theory and apply the EM theory on the components placement and grounding. It is all EM theory and people call it black magic!!! That it is not even straightly correct signal travels in form of electron or current movement!!! It is EM wave that propagates in all electronics. Until you understand that, you will not fully appreciate grounding and noise issue and has to resort to hear say and old regiments.

Believe me, I climbed very high in my career ladder with just hands on try and true experience to say this. That's when you see things that people don't see.