Scott Wurcer just solders 'em down and measures -- this recounted in the last issue of Linear Audio in the article describing the bridge method of measuring capacitor distortion.
There are SMT test adapters sold on EBay --don't see any at present.
Ever try some 3M copper tape on phenolic substrate and a clothes pin?
Big problems are microphony; many in the frame group 12BY's GN's which were mass produced for the TV market, now pressed into low level audio as I do; the slightest chassis vibration on the grids can sound like hundred pianos going off, which masks any attempt for real electron and/or thd+noise ;measurements. That's the reason I use them for moderate gain triodes.
Out of a 200 batch of high gm 12BY7's; only a handful would be on.
richy
Scott,
It is interesting that you are testing these triodes with a C weight filter and merlin tests 200 Hz to 20K Hz. There is a AP 2522 analyzer on my bench I can test with a 100 Hz or 400 highpass filter and 20K low pass filter. There seems to be a fairly large delta in measured noise between the 100 Hz and 400 Hz filters. I do not recall the weighting used by Vogal.
It is of interest to me that when tube noise is measured we speak of EIN. When we speak of the riaa preamplifier we speak of Signal to Noise Raito.
DT
It is interesting that you are testing these triodes with a C weight filter and merlin tests 200 Hz to 20K Hz. There is a AP 2522 analyzer on my bench I can test with a 100 Hz or 400 highpass filter and 20K low pass filter. There seems to be a fairly large delta in measured noise between the 100 Hz and 400 Hz filters. I do not recall the weighting used by Vogal.
It is of interest to me that when tube noise is measured we speak of EIN. When we speak of the riaa preamplifier we speak of Signal to Noise Raito.
DT
Sorry for the delayed response. The choice of flat C weighting (30Hz-8kHz) was deliberate. Mostly, I thought the results would be more generally useful with flat weighting. And the 8kHz BW corresponds to the point where the sensitivity of human hearing begins to fall off sharply.
I was tempted to use a 400 Hz highpass filter to eliminate powerline harmonics. But there is a lot of real tube 1/f noise at those low frequencies that I wanted to capture. The measurements were made with a PicoScope 4262, which is battery operated (from a PC). I used regulated power supplies for the tube supplies, which were line operated, but I placed them more than 6 feet away from my test setup, so I was able to avoid any line AC powerline noise on my benchtop. As a result, I was able to keep powerline harmonics below the level of tube noise on a spectrum analyzer.
In the case of the JFET measurements, I had to use batteries, since the JFET noise was so low.
I was tempted to use a 400 Hz highpass filter to eliminate powerline harmonics. But there is a lot of real tube 1/f noise at those low frequencies that I wanted to capture. The measurements were made with a PicoScope 4262, which is battery operated (from a PC). I used regulated power supplies for the tube supplies, which were line operated, but I placed them more than 6 feet away from my test setup, so I was able to avoid any line AC powerline noise on my benchtop. As a result, I was able to keep powerline harmonics below the level of tube noise on a spectrum analyzer.
In the case of the JFET measurements, I had to use batteries, since the JFET noise was so low.
Lower noise depends on what you want to use it for, and also how much trouble you are willing to go to to get the lowest noise.
The Lighthouse Planar Triode 416 is about as quiet as it gets. But it is usually used in Grounded Grid RF amps. And you need one heck of a special socket!
Try this: A 416 moving coil cartridge preamp: Use a shorting relay at the input (always shorted until the tube and input cap stabilize, and always shorted before the DC filament power and B+ are turned off and collapse). Then use a very large low leakage (Non Electrolytic) coupling cap from the input to the cathode, and a very good quality cathode bias resistor to ground. Use a very good quality plate resistor to B+, and cap couple the signal to the next stage.
Caution: Be sure to regulate the DC filaments, and to regulate the B+. Otherwise, the line voltage will change, and you have to have the moving coil on the cartridge re-wound! (the re-wound one is a new one, I do not think they re-wind the blown one).
And be sure the shorting relay on the input is timed correctly, or you will also need a new cartridge.
The Lighthouse Planar Triode 416 is about as quiet as it gets. But it is usually used in Grounded Grid RF amps. And you need one heck of a special socket!
Try this: A 416 moving coil cartridge preamp: Use a shorting relay at the input (always shorted until the tube and input cap stabilize, and always shorted before the DC filament power and B+ are turned off and collapse). Then use a very large low leakage (Non Electrolytic) coupling cap from the input to the cathode, and a very good quality cathode bias resistor to ground. Use a very good quality plate resistor to B+, and cap couple the signal to the next stage.
Caution: Be sure to regulate the DC filaments, and to regulate the B+. Otherwise, the line voltage will change, and you have to have the moving coil on the cartridge re-wound! (the re-wound one is a new one, I do not think they re-wind the blown one).
And be sure the shorting relay on the input is timed correctly, or you will also need a new cartridge.
Thanks for the suggestions! The C3G, 416, and 6C17K would all be interesting additions to the list, from a standpoint of seeing what is possible with tubes. Unfortunately, they all appear to be difficult-to-get and expensive, and unlikely candidates for any new preamp design. And each would require some time investment in a custom test fixture. So, alas, they probably won't make it onto my list.
I have been thinking about testing the Sovtek 6H30Pi, to see if its microphonics level is any better than the Sovtek 6C45Pi I tested earlier.
The next tube likely to make it onto the list is the 12AY7, since I'm considering it for a new design. It will be interesting to see if its noise level is any different from a 12AX7. Does anyone have experience with either the Reflektor or JJ 12AY7s, from a noise standpoint?
Scott
I have been thinking about testing the Sovtek 6H30Pi, to see if its microphonics level is any better than the Sovtek 6C45Pi I tested earlier.
The next tube likely to make it onto the list is the 12AY7, since I'm considering it for a new design. It will be interesting to see if its noise level is any different from a 12AX7. Does anyone have experience with either the Reflektor or JJ 12AY7s, from a noise standpoint?
Scott
Will start another thread about protecting your low noise amplifier for measuring tube noise -- just want to post 2 different methods. The dotted cyan line is the Supertex protection scheme using two back to back LND150 depletion mosfets. It allows a maximum of 1mA to flow into the protection diodes. Totally useless for tubes.
The red line is a 4.7uF 630V polypropylene capacitor in series with a small incandescent lamp. The cap is biased with a 9V battery. The lead length picks up significant mains energy, but otherwise the scheme adds just under 1nV/Rt Hz from 200Hz to 100kHz.
Baseline for the amplifier, and "reality check" for 100R and 1kR included.
The red line is a 4.7uF 630V polypropylene capacitor in series with a small incandescent lamp. The cap is biased with a 9V battery. The lead length picks up significant mains energy, but otherwise the scheme adds just under 1nV/Rt Hz from 200Hz to 100kHz.
Baseline for the amplifier, and "reality check" for 100R and 1kR included.
Attachments
$2 for 6S17K on eBay is expensive and difficult to get? Plenty!
6N30P is overpriced 6N6P (almost), and I would not expect low microphonics from it.
I solder it. I did not measure, but it is awfully quiet, and zero microphonics. Please don't think that I am pushing to sell. 
However, with 8 micron grid wires, 18 micron between them, so close to cathode, grid current is pretty high, compared to 12AX7. And cathode is directly connected to the filament.
However, with 8 micron grid wires, 18 micron between them, so close to cathode, grid current is pretty high, compared to 12AX7. And cathode is directly connected to the filament.
Use an European ECC83 or a 7025 instead. The difference from your 12AX7 are their bifilar heaters.
Why? Remember these many SS designs that make use of either NPN and PNP devices with only one rail, or with them on the same side at a dual rail supply. It's just a matter of design. With a low impedance supply, it doesn't matter if the emitter shows to ground or to Eb.
Best regards!
I solder it. I did not measure, but it is awfully quiet, and zero microphonics. Please don't think that I am pushing to sell.
However, with 8 micron grid wires, 18 micron between them, so close to cathode, grid current is pretty high, compared to 12AX7. And cathode is directly connected to the filament.
Ok, I'll try a few and see how they work. Thanks. The goal is a MC pre-preamp with a single tube, or maybe two paralleled tubes. But the noise level has to be equivalent to the 2SK209 JFET in my list, which is REALLY difficult for a tube.
Scott
PNP and NPN transistors are different beasts, since electrons are always negative. Different constructions, different technology, different results. Particularly, long time ago were made in Tomsk Institute of Semiconductor Devices medium-power transistors KT626 that had very low noises due to emitter structure. The guy who designed them said, "Because emitter looked nice such a way!"
Probably some modern Zetex devices can be even better.
If you read my list and accompanying text, you'll see I've tried several, including the ubiquitous 6DJ8 / 6922 / ECC88 family. Also a few NOS types. I, too, had originally believed that these would provide the lowest noise. But, on average, current production 12AX7 types have lower EIN than current production 6DJ8 / 6922 / ECC88 types. High-Gm doesn't necessarily mean low noise in the audio range.
Do you have specific suggestions?
Scott