Any reason not to use smaller grid stopper at input?

[Wombaticus]

PS: if you spot any error in my Math and offer your own calculations, fine with me

Er, one quibble is that you can't just add resistance and inductive reactance in the usual way. What you're looking for is the complex impedance, so technically, you need to use Z = sqrt(R^2 + X^2). Doesn't mostly make a huge difference to the numbers in your table, I suppose.

[Alan0354]

We're not talking hifi here. It's not about SNR, it's about the *noise floor.*

Using your example of a 100k resistor generating 4uV of noise (you forgot to include bandwidth in your calculation!), that noise will be amplified by the gain of the preamp.

If the preamp is high-gain, say 1 million times, this will produce 4V of noise at the power valve, causing audible hiss in the speaker, which nobody wants.

Now, your SNR at the *input* of the amp may be great, but if the maximum headroom of the power valve is, say 30V, your SNR at the *output* is a shocking -17.5dB at best! And you have to sit and listen to lots of hiss during 'quiet' passages in your playing.

No, it's 40nV, the output would be 0.04V from you example.

[Alan0354]

As I see it, all these considerations only matter when the guitar volume is turned down close to zero. As soon as you turn it up a bit, the resistance of the volume pot becomes the biggest noise source. (the DC resistance of the pickups isn't relevant because of their inductance as JM explains)

I suppose it is still a valid issue if you're trying to design a high-gain amp that can be cleaned up with the guitar volume ( and stays quiet when muted with the guitar volume)

JM: the pickups are Duncan Alnico 2 Pros.

After I post last night before going to bad, I thought about this. I was going to come here a post as soon as I get up......which is late, you beat me to it.

The pickup is say 4H inductance. At 1KHz is 25K, at 5KHz is 125K ohm. So at higher freq, the resistance of the volume pot come into play. That would be in series with the input resistance and that is much prominent to 20K or even 68K.

[Alan0354]

No, you didn't understand JM's point. When the guitar is at full volume its output impedance is more than 100k in the middle of the audio band, so its thermal noise contribution swamps that of the first tube stage. The high inductance of the pickups means that their DC resistance can't shunt the volume pot at audio frequencies.

I recently designed a preamp for ultrasonic piezo transducers where I had to deal with the exact opposite effect, the transducer being capacitive rather than inductive. It shunted the first stage's noise current at high frequencies, so I got the best results with a low-impedance preamp designed for low noise voltage at the expense of noise current. I tried all sorts of expensive op-amps but in the end a 2N4401 outperformed them all.

You are absolutely correct. I might be a novice in tubes, but I work with a lot of low noise designs. It is not obvious where the dominant noise source. You need to really establish the noise model, take into consideration of the source impedance and input impedance. Then you balance the shot noise induced noise voltage vs thermal noise voltage and design the optimal input impedance to optimize noise. It is not as simple as people thing. That's the reason you have so many opamps out there, there have bipolar input that has lower noise voltage but higher noise current refer to input. They are only good if you have a low impedance source. Then the JFET or MOSFET input where there's almost no input current that work exactly the opposite.

I wrote my own excel program and fill in the source impedance, input noise voltage and noise current of the amp etc. Then input different value to optimize the noise. It is very involve subject, not just a resistor or anything particular.

Actually there is one more noise that nobody here talk about. the 1/f ( pop corn noise ) noise that those carbon composite resistors people love to use. All the pop pop pop!!!! that's where BJT shine.

Anyone has noise model for 12AX7?

[Austin]

I have a 1992 thinline tele that does not lose treble when I roll back the volume to clean up the sound. Every other guitar I have does lose treble when I reduce the guitar volume knob though. I always wondered why the tele didn't and the others did.

[Alan0354]

There is no thermal noise associated with a reactance in the first place.

The pickup is in parallel with the volume pot, when the impedance of the reactance of the pickup goes up, the resistance of the volume pot become dominant. And particularly if you are at half volume, you mainly see the resistance of the volume pot ( of cause you have to do the calculation of the two resistors formed by the volume pot, but you get the drift!!).

[Merlinb]

[b]Frequency------inductive----resistive----total
80Hz-----------2K2------------7K5----------7.8k
160Hz----------4K4------------7K5---------8.6k
320Hz----------8K8------------7K5---------11.6k

4H is a relatively high inductance pickup, but even so, a 10k resistor is still greater than the pickup impedance up to around 260Hz, which still qualifies as a 'wide range' of the guitar fretboard, which is what I said in the first place.
Ok, so you can't make the grid shunt capacitance enormous, but you could probably make it bigger than the cable capacitance without gross side effects, and that's bigger than any sane person would go.

No, it's 40nV, the output would be 0.04V from you example.

The world would be a lot quieter if a 100k resistor only generated 40nV in a 10kHz bandwidth
It's 4uV- you forgot to include the bandwidth in your calculation. Try it on this calculator: Noise calculation calculator calculate Thermal noise Johnson noise white voltage level Nyquist dBu dBV signal-to-noise ratio S/N temperature bandwidth noise figure - sengpielaudio

(You number is correct if you use nano volts per root hertz)

Anyone has noise model for 12AX7?

Some measurement here. The equivalent input noise resistance of most triodes is around 5-6k, mostly pink.
http://www.jogis-roehrenbude.de/Lese...ch/rauschc.htm

[Alan0354]

4H is a relatively high inductance pickup, but even so, a 10k resistor is still greater than the pickup impedance up to around 260Hz, which still qualifies as a 'wide range' of the guitar fretboard, which is what I said in the first place.
Ok, so you can't make the grid shunt capacitance enormous, but you could probably make it bigger than the cable capacitance without gross side effects, and that's bigger than any sane person would go.



The world would be a lot quieter if a 100k resistor only generated 40nV in a 10kHz bandwidth
It's 4uV- you forgot to include the bandwidth in your calculation. Try it on this calculator: Noise calculation calculator calculate Thermal noise Johnson noise white voltage level Nyquist dBu dBV signal-to-noise ratio S/N temperature bandwidth noise figure - sengpielaudio

(You number is correct if you use nano volts per root hertz)


Some measurement here. The equivalent input noise resistance of most triodes is around 5-6k, mostly pink.
c) Rauschmessungen an Röhren

You are right, I went back and check, I got the right formula, but when I did the calculation, I forgot to put in the BW term which is 10KHz. I corrected the original post using red already.

BTW 4H is not high inductance, average strat pup is about 4H. HB is higher.

You have the article in English?

[Alan0354]

I just plug my strat with hb into my high gain channel and crank it up. I have only a 20K ohm resistor in series to the input of the 12AX7.

1) I can hear the hiss even when I turn the volume all the way off.( shorted output). It was the same intensity as if I unplug the jack from the amp( verify as it is shorted without the jack).

2) I slowly crank up the volume, the hiss slowly increase a little, not anything alarming.

3) As I get pass 6 or 7, the hum took over. I had the place as quiet as possible, lights off, walk in all different position to reduce hum. Still the hum took over.

As for crank up the amp and roll back the volume to play clean, I tried that also. I don't think that is an issue, the guitar is still much more dominate than the noise. I just think this is a non issue.

My question is, do you really care? Also, I am not convinced that the input resistance is the only source of noise yet. I yet to find the model of the 12AX7 to confirm. As Steve and I both agree, noise calculation is a lot more complicated than just the input Johnson noise. You really need to have the whole picture to do the calculation.

Guys, any RF is at least 20 times higher than the frequency of interest. Any simple filter after the first stage will filter out the RF if even that is the case. I for one never in my life see or hear any of it. But just say it's there, use a multiple pole filter, with 20:1 frequency ratio, you are not going to affect the audio sound. There will be minimal phase shift or anything at the audio frequency. Consider a series resonance tank circuit to snub out the RF or something. I have not gone into calculation, just talking out loud.

[J M Fahey]

@ wombaticus.
Yes, it's true. I agree with you. Impedance is a complex, vectorial parameter.
I was just trying to keep it simple to non-engineers.
Maybe I should have added a small footnote saying so.
As you noticed, the actual difference is small; specially above , say, 300 Hz where inductive impedance starts to dominate the equation, and becomes real small, I'd say irrelevant, at the frequencies where hiss starts to become a nuisance.
I'm happy you simulated using a pickup model as a generator and not got into the usual mistake of considering *just* the grid stop resistor driven by a "perfect" generator, which is what simulation software provides if you do not tell it otherwise.
It's also quite clear that noise increases a lot, depending very little on the resistive grid stop chosen, at the pickup resonant frequency , which by the way lies *well* in the audio band under real (stage/studio) conditions.
To put it in simpler words: whether you have 10K, 34K or even 68K does not matter that much, compared to the huge generator impedance shown by any real world pickup.

In fact, I must be grateful for showing how accurate my statements were
In your first curve, with traditional 34K grid stop, you show a 46nV/Hz^1/2 noise peak .
On the second one, with a *much* smaller 10K grid stop resistor, *aided* by extra 220pF added, you show a 41 nV/Hz^1/2 noise peak.
Noise reduction is 41/46 , which means an impressive 1dB reduction. Or maybe it's not so impressive?
For those who want to check: Decibel Calculator
and 41/46 is a 0.89 gain in noise.

Wait !!!! maybe all this happens at some irrelevant frequency !!!!
Good question, let's check.
Graph 1 shows the peak at around 2500Hz. Wow!!
Graph 2= *slightly* lower ... could we agree on 2400 Hz?

Not only smack in the guitar band, but even worse, roughly matching a very common peak on guitar speaker response, precisely what separates them from regular Audio speakers.
So any noise there gets much exacerbated by the speaker itself:
Here's the curve of a classic , typical guitar speaker, the Legend 125

It shows a *huge* peak around ... oooops !!!! .... 2500 Hz.

[Merlinb]

@ wombaticus.
In your first curve, with traditional 34K grid stop, you show a 46nV/Hz noise peak .
On the second one, with a *much* smaller 10K grid stop resistor, *aided* by extra 220pF added, you show a 41 nV/Hz noise peak.
Noise reduction is 41/46 , which means an impressive 1dB reduction. Or maybe it's not so impressive?

You're being misleading with your choice of numbers. That 1dB improvement is only at the resonant frequency. (And I think 1dB is worth having anyway)

But what really matters is the total noise integral. Notice that below the resonant frequency the noise is much less with the 10k resistor- only about 16nV/^/Hz versus 26nV/^/Hz for the 34k stopper. So between 20-1kHz you're getting about 500nVrms noise versus 810nVrms, which is a 4.2dB improvement. If you include the resonant peak as well then we're probably getting close to a 6dB reduction, which is certainly not to be sniffed at.

[J M Fahey]

Well, the resonant frequency matches a strong typical speaker peak , around 8dB or more.
I bet *that's* annoying.
There is a reason why "hiss" is vocalized as, precisely, "hisssssssssssssssssssss".
Sorry for adding emphasis on the part of the word that imitates the sound.
At lower frequencies, although it doubtlessly exists, it's not that annoying.

[Alan0354]

We're not talking hifi here. It's not about SNR, it's about the *noise floor.*

Using your example of a 100k resistor generating 4uV of noise (you forgot to include bandwidth in your calculation!), that noise will be amplified by the gain of the preamp.

If the preamp is high-gain, say 1 million times, this will produce 4V of noise at the power valve, causing audible hiss in the speaker, which nobody wants.

Now, your SNR at the *input* of the amp may be great, but if the maximum headroom of the power valve is, say 30V, your SNR at the *output* is a shocking -17.5dB at best! And you have to sit and listen to lots of hiss during 'quiet' passages in your playing.

I was thinking about this. I think 1 million, which is 1EE6 is high. Say you get gain of 40 per stage, and you have 4 stages for high gain. But remember, all the high gain schematic I saw, they have voltage divider at least divided by 2 on every stage for obvious reason. You can't over drive the next stage. So you get only gain of 20 max out per stage. Then you have the tone network that attenuate the signal. All tone network attenuate the signal big time, like divided by 10. Using Fender differential stage and power amp that has feedback with gain of only about 8 ( 820/100 as you see in the resistor divider). so you really has 20X20X20X20X0.1X8=1.28EE5, not 10EE6. That's only 1/8 so instead of 4V, you get 4/8=0.5V as in your calculation. So output power is (V^2)/R= (0.5^2)/8ohm= 0.03125W of noise at full crank for say a 30W amp. That's not a whole lot of noise. Also, to stop squealing with such high gain, you have to have cap to ground in the stages, I doubt if the BW is even beyond 2KHz. That needs to be factored into the noise calculation. Lastly, you ever seen people crank the drive, in between gain and master all the the max.....in any situation? It'll go wild!!!!

As for why we have divider, it is obvious for people in this forum that you need that to prevent DC pump up of the next stage. If you don't do that, the sound is ugly. If you look a Mesa, it is easily divided beyond 2!!!.

As I posted from my real life observation of cranking my own designed amp all the way up and turn off the volume of the guitar, it is not that bad. And when I turn the guitar volume beyond 6, the hum swamp the noise no matter how I try to lower the hum. This is on a HB pickup.


Also the real high gain amp like Mesa sounded so bad. In order for them to prevent feedback from preamp tube to the guitar direct, they have to put capacitance to ground in the stages. If you put them in high gain and roll back your guitar's volume pot, it sounds terrible. It's like you put the tone pot to 1 in your guitar. There is no spark or soul anymore. They have to do that to prevent squealing from guitar to tube feedback. Having even 10EE5 gain of the amp is not a good formular to produce good sound. I personally dispice Mesa, I tried so hard those days to buy a Mesa. I tried Dual Caliber, Mark IV, Blue Angel, Mervrick, Dual Caliber in the late 90s. They just sounded so awful. That Randy Smith just put too much stages and just lost the quality sound of the guitar. Without sounding bias, from my memory and from my tape recording, my first design from 1978 sounded so much better than all the Mesa I tested in the late 90s. That's nothing. You just can't get the sweet cleans like the vintage Fender, Vox AC30 or the Matchless with simple minimal stages amplifiers.

The reason I kept testing Mesa was because the guitarist I looked up to in my days was Carlos Santana and he used a Mesa. I gave so so many chance to give a reason to buy a Mesa, and I just can't. I think if I were to buy a reference amp to judge my build, it would likely be the Baseman 59 or a Matchless.

[Merlinb]

Well, the resonant frequency matches a strong typical speaker peak , around 8dB or more.

That's the frequency response of the driver, but once you put it in a cabinet everything can change. For example, here's the response of a 4x12; the strongest response is now below 1kHz.


And what about the impedance of the driver? The Legend 125 may show a sensitivity peak at 2.5kHz, but the impedance has doubled so you're only getting a quarter of the power into it (assuming a valve amp)...

You can't use the speaker as an excuse to concentrate on noise at just one spot frequency. Where should we stop? What about the frequency response of the amp itself? The room acoustics? The equal loudness curves of your ears? Too many variables- let's stick to a sensible bandwidth considered at the input only, say 5kHz.

I think 1 million, which is 1EE6 is high. Say you get gain of 40 per stage, and you have 4 stages for high gain.

A gain of 40 per stage is abnormally low- typical 12AX7 will give you a gain of closer to 60.
The 4th root of 1 million is 31.6, so four 12AX7s with about 0.5 attenuation between stages will give you a gain of 1 million, which is entirely possible, especially if you include the gain of the phase inverter on top of that.

I only chose 1 million as a convenient number- do the same calculation with a gain of 250000 if you prefer; you've still got a 1V noise signal at the power valve, which is still a lot! You can argue about the minutiae of the amp as much as you like, it doesn't change my point: the noise floor is what we care about, not the SNR, and reducing the grid stopper can help with that. Plenty of people have actually observed the improvement in real life, so I'm not sure why you're determined to oppose me on this one..?

[Merlinb]

Dear Merlin, please let me repeat that your notion of:

10k is much greater than most pickup impedances over a wide range, so the input capacitance can be as big as you like without greatly loading the pickup.

is wrong and can lead to much confusion.

Here is a simulation using a pickup model based on a real unbranded humbucker I measured. The guitar has a 500k pot turned up full, and a 10k grid stopper is used with 100pF Miller capacitance. Now, let's suppose someone increased the grid capacitance to a full 1nF, which is far more than anyone would choose in reality.

Here's the response with zero length of cable- there is clearly a big difference between the 100pF and 1nF case.



But now look at the case with a mere 4m of guitar cable added. In creasing the grid capacitance by a full factor of ten only moves the resonant peak by a factor of about 1.8. So I think I am fairly justified in saying that with a 10k stopper you can add a lot of grid capacitance but the response is still dominated by the cable.