Hello,
We are possibly coming around to some common ground.
There are at least two sources of noise; shot noise and flicker noise. Perhaps we can come to some agreement that flicker or 1/f is pink noise (greater levels at lower octaves) and has a corner frequency where it is greater than shot noise and dominates a say below 4K Hz. Above 4K Hz shot noise dominates. (Also in the audio range.) 1/f will vary from brand to brand and from lot to lot depending on materials used and how clean the clean room was at the time of manufacture.
Not to forget 2.5/gm can be a good representation of shot noise.
Yes, selection of different operating points will result in differing Valve noise and a single reported point is not sufficient to represent how the valve is used in different circuits. This would be good stuff to report on a “data sheet”.
DT
We are possibly coming around to some common ground.
There are at least two sources of noise; shot noise and flicker noise. Perhaps we can come to some agreement that flicker or 1/f is pink noise (greater levels at lower octaves) and has a corner frequency where it is greater than shot noise and dominates a say below 4K Hz. Above 4K Hz shot noise dominates. (Also in the audio range.) 1/f will vary from brand to brand and from lot to lot depending on materials used and how clean the clean room was at the time of manufacture.
Not to forget 2.5/gm can be a good representation of shot noise.
Yes, selection of different operating points will result in differing Valve noise and a single reported point is not sufficient to represent how the valve is used in different circuits. This would be good stuff to report on a “data sheet”.
DT
Hi,
In my own designs I often use current starving as a technique to a) makes the valves used last a lifetime and b) reduces noise to the minimum.
I apply this particularly in error amps as used in series regs and phono stages (both MM an MC).
As always designing is a matter of balancing a set of compromises
At no volume setting a valved stage should produce level of hiss that is interfering even when using high efficiency ribbon tweeters for instance. If it does your design is faulty.
Hiss is what you would hear if you stick your "oar" to the tweeters.....
Cheers, 😉
In my own designs I often use current starving as a technique to a) makes the valves used last a lifetime and b) reduces noise to the minimum.
I apply this particularly in error amps as used in series regs and phono stages (both MM an MC).
As always designing is a matter of balancing a set of compromises
At no volume setting a valved stage should produce level of hiss that is interfering even when using high efficiency ribbon tweeters for instance. If it does your design is faulty.
Hiss is what you would hear if you stick your "oar" to the tweeters.....
Cheers, 😉
Frank,
Current starving in what sense, low plate current or low heater current/voltage ? Can you give some specifics, perhaps I can attempt to measure this?
Scott
Current starving in what sense, low plate current or low heater current/voltage ? Can you give some specifics, perhaps I can attempt to measure this?
Scott
2.5/gm is only a rough estimate of shot noise, but not all the valve textbooks make this clear. Shot noise matters for triodes in the range from tens of kHz up to tens of MHz; below this flicker noise dominates, and above this grid noise dominates.DualTriode said:
For tetrodes and pentodes partition noise matters too; unlike shot noise you can accurately calculate partition noise.
Hi,
@Scott:
Often less than 1/3 of what the valve can be run at. Lower heater voltage sometimes helps as well, heater current should remain the same though.
Cheers, 😉
@Scott:
Often less than 1/3 of what the valve can be run at. Lower heater voltage sometimes helps as well, heater current should remain the same though.
Cheers, 😉
Hello DF96 and All,
Yes 2.5/gm is a RF designer’s rule of thumb. Shot noise is white noise, where 1/f noise is pink noise. 1/f noise has a corner frequency where it increases rapidly below about 4K Hz (another rule of thumb).
Concerning the 2.5/gm rule, some valves follow the rule better than others. The 6AK5 follows the rule well where the 6J4 is a notable exception. The 6DJ8 family does a good job of following the rule as well. As you would say it has a lot to with the valve’s internal geometry.
Partition noise does not contribute to triode noise.
DT
Yes 2.5/gm is a RF designer’s rule of thumb. Shot noise is white noise, where 1/f noise is pink noise. 1/f noise has a corner frequency where it increases rapidly below about 4K Hz (another rule of thumb).
Concerning the 2.5/gm rule, some valves follow the rule better than others. The 6AK5 follows the rule well where the 6J4 is a notable exception. The 6DJ8 family does a good job of following the rule as well. As you would say it has a lot to with the valve’s internal geometry.
Partition noise does not contribute to triode noise.
DT
Hi Everyone,
I updated the measurement list on my website to include the D3a. I had four NOS Siemens D3a shipped to me from Europe. This is a remarkable tube and it does display very low noise, almost 3dB better than other tubes in my study.
I measured the tube as a triode, following the connections in a 1968 Philips data sheet. I tried a variety of bias currents and found lowest noise in the 5 – 8mA range, although it was a very broad optimum. I made measurements at 7.0 mA.
This tube has a fascinating internal construction; I’ve never seen anything like it.
No current production tube that I’ve measured can match the input-referred noise of this tube, individually. The only way to match it would be to operate multiple tubes in parallel.
Unfortunately, these tubes are extremely microphonic. Clapping my hands across the room showed up wildly on the noise output. To be useful in a low-noise audio circuit, the tubes would have to be shock mounted, at a minimum. Even then, I suspect the microphonics would be big problem.
The next measurements in queue are more samples of the JJ E88CC. I also plan to measure some other high-gm RF tubes to see if any have the low audio noise of the D3a (hopefully without the microphonics).
Downloads – Tavish Design
Scott
I updated the measurement list on my website to include the D3a. I had four NOS Siemens D3a shipped to me from Europe. This is a remarkable tube and it does display very low noise, almost 3dB better than other tubes in my study.
I measured the tube as a triode, following the connections in a 1968 Philips data sheet. I tried a variety of bias currents and found lowest noise in the 5 – 8mA range, although it was a very broad optimum. I made measurements at 7.0 mA.
This tube has a fascinating internal construction; I’ve never seen anything like it.
No current production tube that I’ve measured can match the input-referred noise of this tube, individually. The only way to match it would be to operate multiple tubes in parallel.
Unfortunately, these tubes are extremely microphonic. Clapping my hands across the room showed up wildly on the noise output. To be useful in a low-noise audio circuit, the tubes would have to be shock mounted, at a minimum. Even then, I suspect the microphonics would be big problem.
The next measurements in queue are more samples of the JJ E88CC. I also plan to measure some other high-gm RF tubes to see if any have the low audio noise of the D3a (hopefully without the microphonics).
Downloads – Tavish Design
Scott
Hi,
Thanks for doing all this work, Scott.
I can't say I'm surprised by the measurement results for the D3a but I'm wondering how a Russian 6C45 would do by comparison.
Should you be able to lay your hands on some.....?
As for microphony, that may vary from sample to sample with the D3a. The better ones may be long gone though.
Cheers, 😉
Thanks for doing all this work, Scott.
I can't say I'm surprised by the measurement results for the D3a but I'm wondering how a Russian 6C45 would do by comparison.
Should you be able to lay your hands on some.....?
As for microphony, that may vary from sample to sample with the D3a. The better ones may be long gone though.
Cheers, 😉
Other thing to check is oscillation- D3a are particularly prone to that and high microphonics are often a symptom. In my MC phono stage, I don't get much in the way of microphonics, but I had to work a bit on layout and grid-stopping to achieve this.
Hi,
Thanks for your input, Stuart.
That's the problem, isn't it. On one hand you have the high gm with all its advantages, on the other you need to keep the animal in check.
That said, your preamp used a xformer to lift the signal of the MC out of the mud already....
Another question I wonder about is how much grid stoppers contribute to overall noise and, and this is something to ponder about, is there a way to (somewhat or somehow) block the other input nodes of a triode so that that can't contribute to noise either.
Sounds far fetched and I admit I have not thought this through, the Sauternes is just too good, sigh......
Cheers, 😉
Thanks for your input, Stuart.
That's the problem, isn't it. On one hand you have the high gm with all its advantages, on the other you need to keep the animal in check.
That said, your preamp used a xformer to lift the signal of the MC out of the mud already....
Another question I wonder about is how much grid stoppers contribute to overall noise and, and this is something to ponder about, is there a way to (somewhat or somehow) block the other input nodes of a triode so that that can't contribute to noise either.
Sounds far fetched and I admit I have not thought this through, the Sauternes is just too good, sigh......
Cheers, 😉
Hello Scott,
nice to hear that your measurments proof that the D3a is one of the best tubes ever made.
Regarding the microphonie. This can be a sign of oszillation. As well as hum.
The usage of gridstoppers is wise. Not only g1. g2 and the anode as well. You should also use ferrit beards. 100nF ceramic caps from both heaters to the chassis. Sockets should be isolated from the chassis with some kind of suspension.
The datasheet highly recommends the usage of a positive g1 voltage together with a high value cathode resistor for a stable operating point. Though, the D3a is not that problematic as the E810F for example.
I´ve been using this tube for years in my phonostage and I don´t have any microphonie. I can knock on the case with my fist and don´t hear anything.
Best regards,
Martin
nice to hear that your measurments proof that the D3a is one of the best tubes ever made.
Regarding the microphonie. This can be a sign of oszillation. As well as hum.
The usage of gridstoppers is wise. Not only g1. g2 and the anode as well. You should also use ferrit beards. 100nF ceramic caps from both heaters to the chassis. Sockets should be isolated from the chassis with some kind of suspension.
The datasheet highly recommends the usage of a positive g1 voltage together with a high value cathode resistor for a stable operating point. Though, the D3a is not that problematic as the E810F for example.
I´ve been using this tube for years in my phonostage and I don´t have any microphonie. I can knock on the case with my fist and don´t hear anything.
Best regards,
Martin
The grid stopper generates Johnson noise, but but provided it is significantly less than the EIN of the valve, then it won't affect the overall noise figure. A value of 100 ohms generates less than 0.2uV (20-20kHz) which is much less than any valve noise, so it is safe to use.
Not intimating, saying it outright. 😀 I've experienced this with high gm tubes; the first 6KN8 preamp I built back in 1980 was a terrific vibration sensor. It was a slow learning process for me to diagnose why and figure out the fixes. If memory serves, Allen Wright wrote about experiencing the same thing with his ECC88 designs.
I'm a little uncertain about whether the microphonic output is due to physical vibration of the tube, or electrical stresses resulting from an unstable amplifier stage? Ie. would there be a signature microphonic spectrum output if you had a dummy load on the output of the amp, and hence no external vibration sources to excite physical vibration of the grid structure?
If there's no physical vibration source, how do you know that it's microphonic?
If I were to guess (and I stress that word), the sensitivity comes from the modulation of interelectrode capacitances. No data, just empirical observation.
No probs, sorry to prolong the questioning - I just had the initial impression from your first comments that the microphonic output was just due to the electrical stage instability, with no physical influence such as from speaker vibration.
No apologies needed, I love real technical discussion and sorting out what we actually know from what we believe. 😀 In this case, if my hypothesis is correct, there's still a mechanical effect, but it is strongly exacerbated by the sensitivity of oscillators to tiny changes in capacitance.
It’s quite possible that RF instability does contribute to the highly microphonic behavior I’m seeing with the D3a, especially given my “haywired” test circuit. This is a hot tube, with strong gain out to several hundred MHz. I can improve the test fixture – maybe even design a simple PCB using proper RF design techniques. Small PCBs are quite inexpensive through oshpark, so it’s not outrageous to do.
The internal structure of this tube, with long, unsupported, high-tension grid wires makes me suspect that it will always be microphonic, but I’m often wrong when I speculate…….. (That’s why it’s so important to measure things). So if other folks have gotten good results, I’ll have another try.
The low noise I measured with the D3a makes me wonder about other RF tubes. Tubes with low RF noise don’t necessarily have low audio noise, but obviously some do. I had almost given up on NOS tubes, but I may try the 6FQ5A / 6GK5. The main advantage of this tube is that I have a carton of them, rescued from a trash dumpster about 25 years ago. But these were mass production VHF tubes, not “special quality” like the D3a.
Scott
The internal structure of this tube, with long, unsupported, high-tension grid wires makes me suspect that it will always be microphonic, but I’m often wrong when I speculate…….. (That’s why it’s so important to measure things). So if other folks have gotten good results, I’ll have another try.
The low noise I measured with the D3a makes me wonder about other RF tubes. Tubes with low RF noise don’t necessarily have low audio noise, but obviously some do. I had almost given up on NOS tubes, but I may try the 6FQ5A / 6GK5. The main advantage of this tube is that I have a carton of them, rescued from a trash dumpster about 25 years ago. But these were mass production VHF tubes, not “special quality” like the D3a.
Scott
Pete Millett sells tube proto-boards on PCB via EBay. I have several and find them very useful. Here's the octal version:
8 Pin Octal Breadboard Prototype Tube Socket for DIY Experimenting | eBay