tube noise?

[jhaible]

Excuse my ignorance, but while planning to build a rather simple circuit next (a phono preamp), I noticed that I don't know a thing about noise in tubes. And it's not so much the physical background I'm after - just plain practical noise levels for tubes.

In the transistor and opamp world, I think I can say 5nV/SQRT(Hz) is decent, 1nV/SQRT(Hz) is good, and everything below would be superb.

What would I expect in a "good" triode, and what type would be a good one in terms of noise?

I see that often selected tubes are sold, so maybe selection within a certain type is more important than choosing a certain type in the first place? Or rather not? Is it save to say that, for instance, a 12AX7 is less noisy than a 12AT7? What would be a good choice for a first stage in a phono amp anyway?

Pointers to past threads are as welcome as direct answers, if this question has already been beaten to death.

JH.

 

[bcarso]

The 1/f noise corner is usually pretty nasty, coming on pretty early as the frequency descends, but 2kHz-10kHz and up you are usually in competitive territory compared to garden-variety op amps. The supertriodes can be a lot better, in the couple nV/root Hz range.

Sometimes the "low-noise" moniker has as much or more to do with the low microphonics though.

 

[CJ]

We compared 12AX7 with EF804 when Ollie was guest mod at PSW.
Might be somne interesting reading there.
I would link it, but Netscape nuked all my bookmarks for the third time. I have to re-plant the backup file.

 

[PRR]

Figure 2 microvolts across the audio band. That's even in some of the tube spec-sheets, such as 7025.

The high-Gm tuner tubes, working at high current (so you really get high Gm) can get below 1uV over the audio band.

An approximation that often works for triodes is Req = 2.5/Gm. This assumes most of the noise is cathode impedance, 1/Gm, except any noise source is a function of temperature, and a cathode runs very hot, 2 to 3 times warmer than room temperature.

Pentodes have excess noise due to partition effects, which hurt a lot more than you would guess from the small screen current.

Any given tube can have noise far in excess of the theoretical or typical. If you need low noise, be prepared to sort-out 10%-50% of tubes.

Grid current noise is very low, can usually be neglected (i.e. the optimum transformer ratio is infinity), but can be significant on capacitor transducers.

 

[CJ]

This guy did a DIY noise tester.
Don't know, never built one:

http://members.aol.com/sbench102/RAT_Noise/noisep1.gif

PARENT:

http://members.aol.com/sbench101/#Noise

 

[Learner]

Hi JH,
From my limited experience I have found that it is possible to minimise the noise down to 50mV, however this is accompanied by an input and output coupling resistor in series with the coupling capacitor.

The noise can be minimised by the input and output resistor however at the cost of input level to the tube and the output level of the tube, the resistor value varies depending on the circuit/gain stage layout.

For biasing pentodes I have found a method which has worked quite well for me that is, I would use the gm curve and work out the plate and cathode resistor then wire it. Next, I would use a variable resistor to tweak the screen voltage and listen to the noise on the speaker. You may or may not require an input resistor(try a 100k pot) but I think you would definitely need an output resistor(try a 500k pot), connect the output to a power amp listen to the noise and tweak for optimum result.

Just out of curiosity what tube are you trying to use, I have found that for pentodes there is only 1 optimum biasing level with minimum noise(That's not including FB). It is faily close to the data sheet recommanded value but not exactly the same, even if you decrease the plate resistor value aiming to allow higher current peak the bias point does not change or you will get more noise. Connect an amp meter to check the bias current from the power supply, to check the biasing point and see if by lowering the plate resistor you will be able to bias at a higher current point without being plagued by higher level of noise. The ultimate no noise current bias is when the screen voltage is close or equal to the plate voltage, however you'll get no gain which is useless.....

I have found that when try to increase the biasing current beyond the optimum point, the noise sounds like a high pitched metallic tingling noise whereas when you try to lower the current bias beyond the optimum level you get more of a hum and buzzing going on. However, at just slightly lower than the optimum current point eg. <0.2-0.3mA from the optimum point you get pops and crackles like old vinyls.....you still get that noise at the optimum current point however it is a little bit less, plus it will get masked/attenuated by the output coupling resistor.

For triodes, it is much simpler it works pretty well just by plotting the load on the gm curve to work out the plate and cathode resistor value. Otherwise you can try tweaking the cathode resistor with a fixed plate resistor, in order to find the optimum bias point for noise.


Have fun tweakin!!! :grin:

 

[jhaible]

Thanks for all the answers !

JH.

 

[jhaible]

[quote author="PRR"]The high-Gm tuner tubes, working at high current (so you really get high Gm) can get below 1uV over the audio band.

An approximation that often works for triodes is Req = 2.5/Gm. This assumes most of the noise is cathode impedance, 1/Gm, except any noise source is a function of temperature, and a cathode runs very hot, 2 to 3 times warmer than room temperature. [/quote]

Ahh - I understand. Thank you!

So a PCC88 would be a good choice?

JH.

 

[gyraf]

Yes, the ECC88/PCC88 aka 6DJ8 and 7DJ8 are good candidates for low-noise preamplification. They're commonly used for this in HiFi. Even paralleled pairs and quads of these has been seen in very-high-gain very-low-noise MC-RIAA preamps..

Jakob E.

 

[jhaible]

[quote author="gyraf"]Yes, the ECC88/PCC88 aka 6DJ8 and 7DJ8 are good candidates for low-noise preamplification. They're commonly used for this in HiFi. Even paralleled pairs and quads of these has been seen in very-high-gain very-low-noise MC-RIAA preamps..

Jakob E.[/quote]

I think I'll go for the PCC88 then. I have a handful of these, so I can select for lowest noise in the first stage. And current-hungry as they are, and working at a rather low voltage, they may even be good for a transformerless headphone output. So I can build the whole thing with just one tube type.
I'll either try to get 7V DC from a 6.3V winding (but that would be a very tight fit!), or I'll use a step-down / step-up dual transformer approach with a higher intermediate voltage, and go for serial heating.

JH.

 

[gyraf]

..6.3V is within the needed 10% of the 7V..

 

[jhaible]

[quote author="gyraf"]..6.3V is within the needed 10% of the 7V..[/quote]

I know, but I really want DC heating here. And stabilized DC that would be.

The idea goes like this: Using a MOSFET with very low Rdson as a voltage regulator. Schottky rectifier and big caps for low ripple. Then, instead of using a fixed reference voltage, I'd use a heavily lowpass filtered fraction of the rectified signal for reference. Probably use a pot to set the output voltage to 7V (or to 6.9V - whatever is possible). Then, in case of the mains voltage going down for some reason, I'll get an accordingly lower heater voltage, but still without ripple. (If I'd use a fixed voltage regulator just working at the edge of its possible voltage drop, I'd get a nasty ripple when the mains voltage goes slightly down.)

JH.

 

[analag]

http://www.glass-ware.com/tubecircuits/page3.html
http://www.borbelyaudio.com/adobe/phono.pdf

Follow the links, it could provide you with useful info from those in the know.

Analag

 

[gyraf]

Yup, that one looks pretty healthy..!

Thanks, Analag.. :thumb:

Jakob E.

 

[chrissugar]

Mister Borbely's circuits are top. Good links Rowan. :thumb:

chrissugar

 

[jhaible]

[quote author="analag"]

http://www.glass-ware.com/tubecircuits/page3.html
http://www.borbelyaudio.com/adobe/phono.pdf

Follow the links, it could provide you with useful info from those in the know.

Analag[/quote]

Interesting circuits - thanks!

I wonder why the glass ware circuit has its 1st stage buffered with a cathode follower, though. The passive RIAA network wouldn't load the 1st stage much, even without the extra buffering, would it?
I read about the constant current idea. But does it really make a difference here?

JH.

 

[NewYorkDave]

Probably for the convenience of the designer... It makes the RIAA network easier to calculate. Plate-driven networks are a bit more work to design :wink:

 

[jhaible]

[quote author="NewYorkDave"]Probably for the convenience of the designer... It makes the RIAA network easier to calculate. Plate-driven networks are a bit more work to design :wink:[/quote]

I see. Otherwise the plate resistance // external plate resistor would be in series with the filter network, so the 76.8k must be reduced.

Which raises two questions:

1) In simulation, it's easy to recalculate that value. At least it's *temping* to do this. No idea how accurate the tube model is.

and

2) Even if the model is correct (at that bias point), how much variation
of rp is there in a real tube?

Sure, if the filter network input impedance is an order of magnitude higher than rp, 10% error of rp will roughly give 1% error in the frequency response. But I have no idea how big the variation can be in real life.

JH.

 

[jhaible]

And there's another question about the glass ware circuit.

If I'm not mistaken, the signal level at high frequencies is *lower* after the first stage + passive filter than on the input!

That means while the 1st tube is mostly responsible for low and mid freq. noise, the high freq. noise is dominated by the 2nd tube, which is a 12AT7.

So, is that 12AT7 really a good choice here?

JH.

 

[PRR]

> If I'm not mistaken, the signal level at high frequencies is *lower* after the first stage + passive filter than on the input!

Probably not a real problem. The output from a dynamic phono needle is VERY hot in the highs. You are really more likely to run into overload than tube noise, even if the first stage and EQ shows a "loss" and the electronic noise level is set in the second stage.

People whine about "low noise" phono stages, but it is mostly bogus. You can use quite rude amplifiers and still recover the surface noise of the vinyl. Not a lot of point in reducing tube noise lower than that. Not a lot of fun listening to a needle or dummy-source when it isn't in a groove.