I have a Lindos audio test set that I use a lot to make noise measurements. I often use it to measure noise on my tube mic ore design which can have up to 7dB of gain. The Lindos adds a 50dB amplifier in circuit when it measures noise so there can be 120dB of gain from input to measurement. The Lindos also has a small built in speaker so you can listen to the noise - very handy if there is some small residual hum. But what is most interesting is listening when you turn off the power to a mic pre. The HT stays up for quite a long time so you can actually hear the pings and tings as the metal internal parts of the tube cool down.
Cheers
Ian
The four NOS D3a I measured all came from the same source. I may just have gotten microphonic tubes. Maybe not so many good ones left, I don't know.......
But the 6C45 is a current production tube, and an expensive one. So it is reasonable to expect that the tubes I tested are representative of current Sovtek production.
Some more noise data on the 6C45: With the 390 ohm grid resistor shorted out, I see input-referred noise of 0.29 uV RMS integrated over the 8kHz BW. That's equivalent to a noise spectral density of 3.2 nV/rt-Hz, if we assume the noise is flat over the 8kHz. But the spectrum shows it rises at lower frequencies, so my data may not be far off from Brad's.
That is really low noise for a single tube at audio frequencies. Too bad about the microphonics.
Scott
The nice thing about the high Q resonances is that they are fairly hard to excite unless you are spot on the center frequency. So they may not be as deleterious as first appears.
I did find the second batch of the Sovteks I'd has in storage to be somewhat higher noise than the first. All in all I think there was about a factor of two in noise density between the best and the worst. I don't think it was a change in the measurement setup, although I did somewhat improvise with the changes I'd made to permit two tubes to be operated in parallel. I had to tear things down for other projects before testing by returning to the first batch.
I really like the 6C45 and would like to make it work in a MC pre-preamp. But I am concerned about acoustic excitation of those microphonic modes.
I did a simple test with the 6C45 in my noise measurement setup. I took the test circuit out of the aluminum box (in case the box is making the microphonics worse) and put it on a foam pad on my test bench, so that the bench vibrations would not affect the tube. Then I stood 2 meters away, clapped my hands, and recorded the time domain output on the picoscope.
I did the same test with a Shuguang 12AX7B, and I’ve attached the two scope plots.
The 12AX7 scope plot has a higher background noise level because the 12AX7 test circuit has higher gain, and the 12AX7 itself has about 3dB higher input-referred noise.
The 6C45 scope plot has at least 5 times greater microphonic response, even more when you consider that the 6C45 test circuit has lower gain. And there is a much longer decay time, extending off the right end of the plot. This long decay is the broad 12 – 17 kHz microphonic spectrum I noted in an earlier post.
It is difficult to do this test quantitatively, because there is no such thing as a “standard clap”. But, if anything, I clapped louder for the 12AX7, if only because it was necessary to get a big enough response to trigger the scope.
Scott
Attachments
Yes, it would be most interesting to know the peak level of the claps. But I agree that there needs to be some good isolation and acoustical insulation. The simple tube dampers are probably far from adequate.
Well I might get involved in making tubes, so something comparable to the grid-cathode spacing of the 45 would be something to attempt.
Hi,
Surely there must be dozens of ways to acoustically isolate the circuit ?
Wow, lucky you. 😎
Cheers, 😉
Surely there must be dozens of ways to acoustically isolate the circuit ?
Wow, lucky you. 😎
Cheers, 😉
It would be exciting to actually do tubes with application to audio being the primary driver rather than an "accident".
And yes, acoustical isolation is not far-fetched. Although it is nice to see the glow of the heaters. But something like double-pane glass, and something to act as a heat exchanger. It could work.
Scott, you will need to get IEC approval to change CNR(dB) from clang to noise ratio, to clap to noise ratio when doing microphonic standardised tests ;-)
To get an idea of the peak level of the claps, I put my sound level meter on the work table, on the foam pad, and set it for fastest response time (100ms) and “peak hold”. Then I stood 2 meters away and clapped. I got readings in the range of 92 – 94 dB SPL, C-weighted. Claps have a fast rise time, so it’s hard to know if this is really the peak or not, but it gives an idea.
I’m still trying to work out how big a problem this microphonics would be in practice. I may try the 6C45 test circuit connected to my audio system, to see if I can hear the slowly decaying microphonics.
On a separate matter, I updated the list of tube noise measurements on our “Downloads” page to include the RU-6H9C. These are often sold re-labeled as Sovtek 6SL7GT.
I was very pleasantly surprised by this 6SL7!! The three samples I measured had an average input-referred noise around 0.56 uV RMS, which ranks them with the best 12AX7 tubes, and among the lowest noise tubes in the study.
These tubes are around $10 each ($14 if you want them re-labeled as Sovtek 6SL7GT), compared to almost $40 for the Tung Sol Reissue 6SL7GT. And the 3 samples I tested had 3dB lower inherent noise than the Tung Sol Reissue. Go figure. And note that you could wire two of these Sovtek 6H9C/6SL7 sections in parallel and get about the same noise as with a 6C45.........
The usual caveat applies: I’ve only measured 3 tubes, but I intend to measure more. Scott
More good work, thanks! It would also be interesting to bombard the device under test with a sinusoid from a tweeter and tune the frequency for peak responses. My impression watching the decaying residuals on the scope from the Ap is that they were fairly monochromatic. This might even bode well for tuned acoustical filtering.
In other news on tube noise
The man may love tubes, and I applaud this interest, but the lax editorial scrutiny in audioXpress has never been more evident than in the November 2015 issue, which features an article by Dr. Honeycutt on noise. To my slack-jawed amazement he asserts that 1/f noise is only found in resistors! Moreover, since it is difficult to calculate, he advises that we just use quieter resistors.
As Anna Russell used to say about Wagner opera plots, "I am not making this up".
Honeycutt has a PhD in acoustics.
Now it's instructive to realize that this stuff is fairly new --- even the glimmerings of understanding of random electrical noise date back to the late 1920's, and in an IRE review paper on the subject from 1930, there is a remark about tube noise around 1 kilocycle/second being something not exactly due to shot noise or thermal noise. And it is also true that 1/f noise is still not well-understood and is an active area of investigation. But doesn't anyone read these submissions? I have abandoned my practice of writing LTEs (they don't print them anymore I understand --- perhaps it has a negative impact on authorial morale and discourages contributions). The last article I read where I was maddened by aspects of the text was someone who described a single JFET stage, and began the piece by stating that we all like distortion, and that it was just a matter of what kind it was. He also described a unity-gain buffer amplifier as a big transconductance, or words to that effect. But I later heard, at least, that the article evoked a large response from others, none of which I saw. And I'd suppose that within those there were many who took issue with the pronouncements like the above.
Although reading Stereophile can be exceedingly frustrating, JA has been good to state his own degree of being baffled by the discrepancy between his measurements and the accounts of the subjective reviewers. Remember the Croft integrated amp several issues back? How about a 6dB error in the RIAA response? Or the recent modular amplifier, that had a couple of modules with decent performance but a bunch more with significant difficulties. But I rant and veer off topic.
Back to noise.
The man may love tubes, and I applaud this interest, but the lax editorial scrutiny in audioXpress has never been more evident than in the November 2015 issue, which features an article by Dr. Honeycutt on noise. To my slack-jawed amazement he asserts that 1/f noise is only found in resistors! Moreover, since it is difficult to calculate, he advises that we just use quieter resistors.
As Anna Russell used to say about Wagner opera plots, "I am not making this up".
Honeycutt has a PhD in acoustics.
Now it's instructive to realize that this stuff is fairly new --- even the glimmerings of understanding of random electrical noise date back to the late 1920's, and in an IRE review paper on the subject from 1930, there is a remark about tube noise around 1 kilocycle/second being something not exactly due to shot noise or thermal noise. And it is also true that 1/f noise is still not well-understood and is an active area of investigation. But doesn't anyone read these submissions? I have abandoned my practice of writing LTEs (they don't print them anymore I understand --- perhaps it has a negative impact on authorial morale and discourages contributions). The last article I read where I was maddened by aspects of the text was someone who described a single JFET stage, and began the piece by stating that we all like distortion, and that it was just a matter of what kind it was. He also described a unity-gain buffer amplifier as a big transconductance, or words to that effect. But I later heard, at least, that the article evoked a large response from others, none of which I saw. And I'd suppose that within those there were many who took issue with the pronouncements like the above.
Although reading Stereophile can be exceedingly frustrating, JA has been good to state his own degree of being baffled by the discrepancy between his measurements and the accounts of the subjective reviewers. Remember the Croft integrated amp several issues back? How about a 6dB error in the RIAA response? Or the recent modular amplifier, that had a couple of modules with decent performance but a bunch more with significant difficulties. But I rant and veer off topic.
Back to noise.
I subscribed to AudioXpress for about three years but found it was so full of technical inaccuracies as to be all but useless.
Cheers
Ian
Cheers
Ian
Sorry to indulge further in off-topic stuff. Ed Dell did yeoman's service as the editor of a few magazines for many years, but was clearly in personal decline toward the end. Not knowing better, and before I bought a complete set of back issues and better appreciated the history, I sent an LTE a good many years ago which took one of his editorial pieces to task, and criticized two articles with either errors or simply misleading bits. Ed printed the letter, but re-ran his column almost verbatim with a repeat of the errors shortly thereafter. A bad sign. Meanwhile I criticized a mistake by Norm Thagard and expressed disappointment about an Erno Borbely article. Norm's response was to become a friend and sometimes pass things by me privately for constructive criticism and proofing before publication.
Erno was in a strop and said I obviously didn't read the magazine much --- his response prima facie was in its way a vindication of my criticisms, namely that there was little new in his article, despite having "The New Frontier" in the title. I had looked forward to it eagerly as I hoped I'd learn about new FETs, and that new circuits would be presented (the article might well have been called Good JFETs: The Soon-To-Disappear Parts). As many may know his contributions to diyaudio ceased roughly around that time as well. I suspect it was just unrewarding and tough on the blood pressure. This is a shame, although he has found an acceptable forum in Didden's journal, evidently.
No one is perfect, but we are witnessing and sometimes participating in a culture with an increasingly prevalent notion that virtually any individual's opinion is just as good as someone else's. There is a quote from Asimov that is apt.
I have no objection to the expression of opinions, and I'm sure plenty of mine would be found dubious by others. But there is a consensus reality, and within it, usually science and the scientific method still "work". And attempts to countermand this on the basis of mere whims, while presenting such as just as good as anything else, one ought to at least notice.*
But that's just my opinion.
The alternatives for more reliable information include stuffy refereed academic journals, often quite expensive, although at least some of which are open access. And with the famous publish-or-perish traditions, there is a lot of redundancy. Books as well continue to flourish in numbers at least, but are very uneven. I mentioned one to a friend and he said Oh I sent that one back, it was so bad!
*The passage by one legislature to make Pi equal to 3 because it made the calculation of the circumferences and areas of circles easier to compute comes to mind (but I veer into politics).
Honeycutt once asked me for permission to use some images from an article I published on my website. I gave him permission but was then diasappointed to find that he then plagiarised my entire article. Badly. To add insult to injury, my original article had been accepted by Ed Dell just before Elektor bought out the magazine, so it never hapened. I havent bought it since.
My only interaction with him (despite us both writing for the same publication) was to correct some errors in his recent pentode article. He was quite gracious about it, but I don't know if a correction has appeared.
The actual story is a bit different than that. Nothing was ever passed, that's an urban legend.
https://en.wikipedia.org/wiki/Indiana_Pi_Bill
Here in yoorp the Carrot has been redefined as a fruit (I kid you not). This is because Jam has been defined as only containing fruit and sugar. manufacturers of cheap jam bulk it out with carrots.
I did also read (but have not managed to confirm) that edible snails were reclassified as fish so french snail farmers could get better subsidies.
I did also read (but have not managed to confirm) that edible snails were reclassified as fish so french snail farmers could get better subsidies.
I’ve been wondering how much of the 6C45’s low noise performance is actually due to the close grid-cathode spacing, and how much is due to other features of its design and construction. The relationship:
Close grid-cathode spacing -> High Gm -> Low Shot Noise -> Low overall noise
is well established at RF, but other noise sources become important at audio frequencies, and the importance of high Gm is less certain. Many high Gm tubes I’ve measured don’t show low noise. I’m wondering if the cathode materials, cathode temperature, and cathode geometry might be more important.
The planar (as opposed to radial) geometry of the 6C45 and D3a results in large cathode and anode area relative to the current flow. Many tubes with high emission cathodes don’t have such high cathode area. Another consequence of the planar geometry is two independent anodes that are wired in parallel inside the tube, doubling the effective tube area.
As a comparison to the 6C45, the vintage (~1940) design of the 6H9C / 6SL7GT has very loose grid-cathode spacing (and low Gm) but still provides remarkably low noise.
I updated my list on our “Downloads” page to include measurements on two 6H9C / 6SL7GT wired in parallel. In this parallel configuration, the 6H9C NF improves to 5.8 dB, which is within a dB of the 6C45 performance. I also posted a close-up photo of the 6C45, attempting to show its planar geometry, for those who aren’t familiar with it.
I’ll see if the 6H9C low noise holds up when I measure more samples.
Scott
