They're a mixed bag of Russian manufacture. They don't really have 'brands' like the West does, so I don't really know. I don't think brand makes a hill o' beans difference to be honest, you pretty much always get a spread like this. I've resigned myself to the fact that, whatever tube type you choose, you can usually count on an EIN of about 1uV in the audio band. If you want significantly less then you usually have to trade it for something (Miller capacitance/microphonics/cherry picking) 🙄
Last edited:
Yes, it is clearly not 20dB/decade. From 2KHz down to 20Hz it changes by about 20dB so it is more like 10dB/decade. Like you I am not sure why. I'll check out one of my books on the subject.
Cheers
Ian
1/f noise (where the exponent of f is 1) should have a 3dB/octave or 10dB/decade characteristic.
Yes, thanks Brad, I was being dumb. And when I smooth my data and re-plot on log-log, that's about what I see. No mystery.
Scott
Scott
Here's a tutorial on the spectrum of 1/f noise I found:
http://www.analog.com/media/en/training-seminars/tutorials/MT-048.pdf
Looks like the readings I made have a turnover frequency of about 2KHz.
Cheers
Ian
http://www.analog.com/media/en/training-seminars/tutorials/MT-048.pdf
Looks like the readings I made have a turnover frequency of about 2KHz.
Cheers
Ian
My comment in the EDN set is perhaps the biggest harumph. The mysterious thing about 1/f noise is we don't understand it.
And both articles suggest that chopper stabilization can remove it. Nonsense! The components of the chopper are not immune either, although you can do quite well.
But ensemble averaging works to get lower noise.
And both articles suggest that chopper stabilization can remove it. Nonsense! The components of the chopper are not immune either, although you can do quite well.
But ensemble averaging works to get lower noise.
There are several theories about what might cause it in tubes but none of them is wholly satisfactory. What is certain is that shot noise proportional to 1/gm at the operating point but since tubes were first made, they had to be selected for low 1/f noise.
Cheers
Ian
Thanks for sharing these tests with us. So the humble 12AX7 type is leading the race, interesting. Was the 12AY7 tested? It was marketed as the low noise/low microphonics/input stage specialist member of the 12A_7 family. I don't know if the marketing was true, but the datasheets described it that way anyhow. The people who buy the current production versions are using them to tailor overdriven guitar circuits so I don't know if modern production 12AY7s would bother striving for the low noise thing anymore. Would be interesting to see though.
Last edited:
I haven't tested the 12AY7. A quick look through the tube vendor webpages shows that there are currently-produced Shuguang, Reflektor, and JJ versions of it, so it might be interesting to try. Given everything I've learned in this study, I suspect the intrinsic noise might not be much different (and certainly no lower) than a 12AX7.
Claims of low noise from vendors or manufacturers seem to be meaningless. But where a tube is manufactured does seem to make a difference: Reflektor-manufactured tubes measure slightly noisier in my study. Given modern manufacturing techniques, I suspect manufacturers could produce lower noise tubes, if there were some motivation to do it........ With statistical process control, you don't really need to fully understand 1/f noise to control and reduce it.
Scott
Claims of low noise from vendors or manufacturers seem to be meaningless. But where a tube is manufactured does seem to make a difference: Reflektor-manufactured tubes measure slightly noisier in my study. Given modern manufacturing techniques, I suspect manufacturers could produce lower noise tubes, if there were some motivation to do it........ With statistical process control, you don't really need to fully understand 1/f noise to control and reduce it.
Scott
testing the likes of 6AG5’s
Scott,
Will you consider testing the likes of 6AG5’s and 6BC5’s? (triode connected)
They are plentiful and cheap?
DT
Scott,
Will you consider testing the likes of 6AG5’s and 6BC5’s? (triode connected)
They are plentiful and cheap?
DT
Hi DT,
I've been quite busy and it's been a few months since I updated the noise measurements online. But I'll put those on the to-do list.
When I began the noise measurement project, I had high expectations for RF tubes - I expected that high Gm would result in low audio noise. One of the first tubes I measured was the 6GK5. But noise in the audio band seems to be dominated by other things, like 1/f noise. So based on my other measurements, I wouldn't expect the 6AG5 or 6BG5 to be as low noise as a 12AX7 for audio. But expectations don't always = reality, so I'll try them. Some RF tubes (like the 6C45PI) do have low audio noise.
Scott
I've been quite busy and it's been a few months since I updated the noise measurements online. But I'll put those on the to-do list.
When I began the noise measurement project, I had high expectations for RF tubes - I expected that high Gm would result in low audio noise. One of the first tubes I measured was the 6GK5. But noise in the audio band seems to be dominated by other things, like 1/f noise. So based on my other measurements, I wouldn't expect the 6AG5 or 6BG5 to be as low noise as a 12AX7 for audio. But expectations don't always = reality, so I'll try them. Some RF tubes (like the 6C45PI) do have low audio noise.
Scott
for lowest noise, build your preamp using the 2SK209 Jfet or other low noise silicon transistor as the front end feeding the tube. jfet plus 6922 for 2 stage RIAA preamp.
The 2SK209 is actually in my list of tube noise measurements, and it's an interesting comparison to the tubes. I haven't measured many 2SK209, because they're surface mount, but I've noticed that their noise level also varies a lot from sample-to-sample. I wish I knew of any inexpensive socket I could use to screen them.
The JFET / tube cascode works nicely, I agree. But even as the upper device, the tube does contribute significant noise to the cascode circuit. So the cascode noise level is not as low as the JFET alone. For a MC preamp, the JFET / tube cascode is a bit noisy (hiss), at least compared to a step-up transformer.
Scott
The JFET / tube cascode works nicely, I agree. But even as the upper device, the tube does contribute significant noise to the cascode circuit. So the cascode noise level is not as low as the JFET alone. For a MC preamp, the JFET / tube cascode is a bit noisy (hiss), at least compared to a step-up transformer.
Scott
