That's a good observation. I updated the project webpage and pdf summary to include actual measurement histograms for the JJ 5751, JJ E88CC, and Tung Sol Reissue (Reflector) 6SL7GT (the three tubes for which I have the most measurements). I'll try to attach the JJ 5751 histogram to this post. They are all beginning to take on Poisson characteristics.
It is interesting to compare the three distributions. The JJ 5751 shows a relatively tight clustering between 0.5 - 0.6 uV RMS, while the JJ E88CC is spread from 0.5 - >1.5 uV RMS. It seems JJ may be having trouble manufacturing the E88CC, from a noise perspective. The Tung Sol Reissue 6SL7GT is clustered from 0.7 - 0.9 uV RMS, showing more consistent manufacturing but relatively high noise.
I'd like to find a more consistent 6DJ8 / E88CC type - does anyone have experience with similar Chinese types (from a noise perspective)? 6N1P?
I haven't forgotten about the 6CW4, I'm curious too.
Scott
Attachments
DF96, petertub,
Thanks for mentioning that. I knew that the ECC99 was a JJ design, but I assumed it was an older design that they reintroduced and "re-purposed" for audio. I didn't know it was a new design specifically for audio. I guess I assumed it was an RF tube because it is similar to the US type 5687.
At any rate, it seems like a decent performer from a noise perspective, based on the few samples I've measured. I certainly intend to measure a few more.
Scott
In all the applications in with i used the JJ ECC99 the russian 6n6p had better performance noise wise.
Also have russian 6n16 and 6n28 very low noise, i use them in my phonostages.
Also have russian 6n16 and 6n28 very low noise, i use them in my phonostages.
Last edited:
Hi,
JJ also call it an RF valve themselves even though they say it was designed specifically to drive DH valves a la 300B and consorts.
Either way, one shouldn't expect too much technically correct definitions from JJ.
Cheers, 😉
JJ also call it an RF valve themselves even though they say it was designed specifically to drive DH valves a la 300B and consorts.
Either way, one shouldn't expect too much technically correct definitions from JJ.
Cheers, 😉
Thanks for the recommendations!
Do you know where those tubes are manufactured? Are they currently produced? As you can see from my list, all the Reflektor-manufactured tubes I've measured have shown fairly consistent performance, but also relatively high noise. I'd be happy to find some Russian tubes that had lower inherent noise.
Which would you try first, from a noise perspective?
Scott
6N16B (6Н16Б-В) dual and 6S6B-V (6С6Б-В) single triodes, tiny wire-ended and extremely nice valves.
--They have characteristics not unlike to 6SN7/6J5 types. 20 gain, but lower voltage/wattage devices.
The High Mu equivalent, the 6N17B and 6S7B could be thought of like a 6SL7 are also available on ebay, tend to be more expensive, but I have no experience with them myself....
I wire these wire-ended into either std. DIL or SIL sockets according to type, and these then can be plugged into another socket mounted on whatever...
Not in current production as far as I know--but there's loads on ebay at reasonable price, I can definitely recommend experiments with 'em....
Makers,-- Orel and Melz ive seen, and seem to have a few 'HQ' type versions around with lesser known suffixes like 'OC' which allegedly means 'high-stability'.
--They have characteristics not unlike to 6SN7/6J5 types. 20 gain, but lower voltage/wattage devices.
The High Mu equivalent, the 6N17B and 6S7B could be thought of like a 6SL7 are also available on ebay, tend to be more expensive, but I have no experience with them myself....
I wire these wire-ended into either std. DIL or SIL sockets according to type, and these then can be plugged into another socket mounted on whatever...
Not in current production as far as I know--but there's loads on ebay at reasonable price, I can definitely recommend experiments with 'em....
Makers,-- Orel and Melz ive seen, and seem to have a few 'HQ' type versions around with lesser known suffixes like 'OC' which allegedly means 'high-stability'.
Last edited:
Interesting. These are NOS Soviet-era tubes, similar to the US/European subminiature types 6021, 6111, 6112. Different pinouts, however. I have some subminiature tubes on hand that I purchased for a microphone project years ago, 5840W, I believe. Remarkably, all seem to still be available from tube vendors.
I'll put these on my to-do list for noise measurement. Thanks.
Scott
I am curious for the noise of the 6n16. I can send you some tubes free of charge. Just send me a PM with your adress.
Ronny
Ronny
6C45 noise
This is a great thread.
As mostly a favor to a hollow-state-or-bust friend, who is pursuing an MC stepup using paralleled triodes and having some difficulties, I did a fairly crude air circuit to test some 6C45s. He had been following a design based on an old post somewhere with 2 6922s and the four triodes in parallel, and using a transformer in the plate circuit and small unbypassed cathode resistors.
In order to extract data about the three samples I had of the 45s I went from a 1.5k resistor in the plate operating from an 80V bench supply and with R-C filtering, to a simple ~10mA current generator using an MPSW92 and biased with an HLMP-6000 LED. The measuring instrument is an Audio Precision set for the 100k input resistance.
Even with the Ap highpass set for its highest frequency of 400Hz, there is still an appreciable amount of low-frequency noise, based on the rise in noise at the plate versus the lowpass settings, but little hum and harmonics to 180Hz at least from the partially-shielded crude breadboard. A d.c. bench supply was used for the heater.
The gains are measured for each tube versus frequency. The grids were driven from a voltage divider of 20k/200 ohms, among other things so that noise from the Ap generator could be neglected when the generator was switched off for measuring noise. For an averaged input-referred noise the 200 ohm thermal noise was backed out root-difference-of-squares. As an additional reality check, and for easier assessment of grid current, I changed the divider to 100k/1k, and got close to the anticipated increase in noise due to the 1k thermal noise.
The resulting gains were a bit less than the datasheet mu for a considerably larger current and plate voltage, but still (times) -45 to -48 (the Sovtek datasheet says ~52). The attenuation of the Ap loading is fairly small. The cathodes were variously grounded and then at the guy's request small unbypassed resistors were inserted (10 and 20 ohms).
The lowest noise referred to the input was around 2.3nV/sq rt Hz. The contribution of the current source was not subtracted but a rough analysis based on the noise sources suggested that it would be fairly small. A much better current source could be constructed.
The distortion versus level for the tubes was exemplary, indicating over the range a nearly constant mu until rather high input voltages (for any reasonable MC cartridge). If the curve of growth of the predominant 2nd is directly proportional to level, typical input voltages would produce silly-low distortions.
Microphonics though. I'd noticed these when using the C45s in a much-less demanding application and seen a ringing that implied a very high mechanical Q. Looking at the Ap residual the decay after excitation into the noise floor takes several seconds, depending on the initial ping. The flipside of such behavior is that it will take a pretty impulsive excitation to excite these resonances to begin with. One of the guy's friends said that C45s "sounded metallic", and perhaps this is associated with the microphonics.
As I have other things on my plate I am going to have to set this investigation aside, but I thought it might be helpful to mention these somewhat tentative results (initially posted in another thread).
Brad
This is a great thread.
As mostly a favor to a hollow-state-or-bust friend, who is pursuing an MC stepup using paralleled triodes and having some difficulties, I did a fairly crude air circuit to test some 6C45s. He had been following a design based on an old post somewhere with 2 6922s and the four triodes in parallel, and using a transformer in the plate circuit and small unbypassed cathode resistors.
In order to extract data about the three samples I had of the 45s I went from a 1.5k resistor in the plate operating from an 80V bench supply and with R-C filtering, to a simple ~10mA current generator using an MPSW92 and biased with an HLMP-6000 LED. The measuring instrument is an Audio Precision set for the 100k input resistance.
Even with the Ap highpass set for its highest frequency of 400Hz, there is still an appreciable amount of low-frequency noise, based on the rise in noise at the plate versus the lowpass settings, but little hum and harmonics to 180Hz at least from the partially-shielded crude breadboard. A d.c. bench supply was used for the heater.
The gains are measured for each tube versus frequency. The grids were driven from a voltage divider of 20k/200 ohms, among other things so that noise from the Ap generator could be neglected when the generator was switched off for measuring noise. For an averaged input-referred noise the 200 ohm thermal noise was backed out root-difference-of-squares. As an additional reality check, and for easier assessment of grid current, I changed the divider to 100k/1k, and got close to the anticipated increase in noise due to the 1k thermal noise.
The resulting gains were a bit less than the datasheet mu for a considerably larger current and plate voltage, but still (times) -45 to -48 (the Sovtek datasheet says ~52). The attenuation of the Ap loading is fairly small. The cathodes were variously grounded and then at the guy's request small unbypassed resistors were inserted (10 and 20 ohms).
The lowest noise referred to the input was around 2.3nV/sq rt Hz. The contribution of the current source was not subtracted but a rough analysis based on the noise sources suggested that it would be fairly small. A much better current source could be constructed.
The distortion versus level for the tubes was exemplary, indicating over the range a nearly constant mu until rather high input voltages (for any reasonable MC cartridge). If the curve of growth of the predominant 2nd is directly proportional to level, typical input voltages would produce silly-low distortions.
Microphonics though. I'd noticed these when using the C45s in a much-less demanding application and seen a ringing that implied a very high mechanical Q. Looking at the Ap residual the decay after excitation into the noise floor takes several seconds, depending on the initial ping. The flipside of such behavior is that it will take a pretty impulsive excitation to excite these resonances to begin with. One of the guy's friends said that C45s "sounded metallic", and perhaps this is associated with the microphonics.
As I have other things on my plate I am going to have to set this investigation aside, but I thought it might be helpful to mention these somewhat tentative results (initially posted in another thread).
Brad
@Koifarm: For phono stages wouldn't the 6n17 be a better candidate. Only just starting to research these tubes so may have missed something fundamental.
And apologies for going somewhat off topic.
And apologies for going somewhat off topic.
Last edited:
Thus altering the terms of the experiment.
I built a HiZ, low noise front end, and lifted a page from the Tektronix Diff Amp (5a22/7a22) to put less stress on the DUT, and a lot less on the AP.
Whose experiment? Whose terms?
As I say this was a favor to a friend, not an attempt to replicate the conditions of the thread starter's tests.
But a nanofarad or so and 100k is hardly a great deal of stress. The low distortion is indicative of that, although the primary interest was a decent noise estimate. And I got that.
You are welcome to do your own measurements.
The choice of with tubes to use is inherrent to the design you like to use.
I use 3 stage with passive RIAA and no feedback. So i need tubes with low amplication factor. The first tube needs as quit as possible. 6n17 has much more noise.
Ronny
Wow! I can't refuse that offer, and I'm very curious as well. I'll send you a PM. Thanks, Scott
Hi Brad,
It definitely sounds like I should measure the 6C45, thanks sharing for the measurements!
I've been measuring integrated, C-weighted (30Hz - 8 kHz) noise. If I take your 2.3nV/rt-Hz spot noise figure and integrate it over 8 kHz, I get ~0.2 uV RMS, which would be the lowest noise in the list, even better than the D3a. What frequency does the 2.3 nV/rt-Hz refer to? Was it flat over freq?
And to confirm, you measured this at 10mA for a single tube, correct?
Retail price is the US seems to be $30 each, but I'll buy or find a few......
Thanks,
Scott
I don't think U47's were underheated to reduce noise, it was to reduce the
microphone temperature. Further to this, I believe only very few tubes can
be run in this fashion without other detrimental effects.
http://www.wagner-microphones.com/compare inside3.htm
FWIW, I think the whole test procedure here is somewhat flawed. If you are
trying to measure very low levels of noise, especially in very high gm frame
grid tubes, their self noise can easily be well below that of a 390 ohm
input resistor.
Noise figure is not a very useful specification when it comes to noise.
Input referred noise in nV/rt Hz versus frequency (with a shorted input) is
far more useful. If I see nV/rt Hz and in the case of a BJT current noise in
pA/rt Hz I immediately know what I am dealing with and how it will work in a
given situation.
cheers
Terry
Setup was one tube at a time, for the plate a simple PNP 10mA current source with LED bias of an MPSW92, 88.7 ohms in the emitter, that ran from a filtered bench supply of about 90V. The input divider was 200 ohms to ground, 20k in series from the generator (when measuring gain), so an attenuation of times 101. Zero ohms (or once, at the request of the friend, 10 ohms) in the cathode to ground. DC heater supply from a bench supply. The Ap bandwidth was typically 400Hz to 22kHz, so 21.6kHz. Of course the low frequency noise makes lower highpass cutoffs worse. I really need to get a spectrum with a cleaner setup, as the onset of 1/f is indeed probably near the cited 4kHz.
So for determining input noise one has to subtract the thermal noise of the 200 ohms, which is significant, by root difference of squares.
I have better numbers for one tube by accounting for the slight loading by the Ap 100k input R, and I also changed the input divider to 1.00k and 10 ohms. So the new attenuation with the Ap gen output set to 50 ohms is times 106 (1 + [1.05k/10]). I don't have other than a number from simulations for the I source output resistance, which sims say is 776k (flat over most of the audio range)---so that's another small uncertainty.
For determination of the plate resistance and transconductance, I added, for one measurement of gain, a 1.50k R with a 10uF electrolytic to common, for a predictable a.c. load on the plate. Solving simultaneous equations for Rp and gm I got Rp = 1.80k, gm = 28.3mA/V. The plate voltage for this was 64.8V. Note that the mu is now 50.9, compared to the Sovtek datasheet nominal of 52. Not bad for running things at such lower currents/voltages.
A really quick look at noise as a function of Ip showed that 10mA maybe was a lucky first guess, as noise seemed a bit worse on both 5mA and 15mA.
Again, this is an aggregate number for noise density, although things didn't go south too badly at lower HP cutoffs. RIAA will make the low frequencies a lot worse.
With the close proximity of grid to cathode there is appreciable negative grid current---even with zero plate current! This will complicate matters if the tube were to be used with MM/MI cartridges, but the friend's application is as an MC stepup. He was worried about positive grid current, but it remains net negative for rather large input signals. The negative current could be offset I suppose by a large resistor from a quiet voltage source, but I can't see it doing much harm to a low resistance MC cartridge.
As far as the C weighting filter, I'm not sure what the ENB (equivalent noise bandwidth) is. The Ap 22kHz filter is fairly sharp cutoff, so NBW is close to -3dB BW. I'm not sure about the highpass ones.
low freq noise C45 sample
Some rough numbers for noise at lower highpass frequencies: an increase (compared to the 400Hz - 22kHz numbers) of 13% for 100Hz, an increase of 33% for 22Hz. Again these are somewhat eyeballed off the Ap "meters". But they are better than I expected. Some low frequency noise in the I source may be contributing.
Some rough numbers for noise at lower highpass frequencies: an increase (compared to the 400Hz - 22kHz numbers) of 13% for 100Hz, an increase of 33% for 22Hz. Again these are somewhat eyeballed off the Ap "meters". But they are better than I expected. Some low frequency noise in the I source may be contributing.
Last edited: