Tweet
I’m not sure where I’m going to go with this. Maybe it’ll be a brief flame that will be immediately extinguished in a torrent of criticism or perhaps it will become some kind of community effort. I hope the latter. I’m willing to develop it, if there’s an interest. For now, let’s just try to have some fun.
Inspired by recent thread “Do tubes have different tones”, or something like that, I decided to embark on my latest folly. Get a bunch of tubes, do some sound tests and capture selected electrical data with the fond, but probably unrealistic, hope that the sounds can be correlated with the measurements and that we can all be enlightened by the experience.
So, without as much as a fanfare, let’s meet the players...
I picked a collection of new production 12AX7 type, but threw in a couple of other 12A*7 types just for interest. The reason for this choice is that I wanted to have information that would actually be useful to a wide population and the tubes they are likely to have access to. I certainly don’t have access to a bunch of NOS tubes and in any case it would blow away the tiny budget for this project. Needless to say, donations of NOS tubes for the project, especially your most precious, unusual and rare ones would be welcomed
For now, it’s just these preamp tubes. Power tubes may come later but will certainly require a lot more thought and effort.
Next the test amp, irreverently but affectionately named after its predecessor, the Fender Champ, meet “The Chump”.
Sorry about the badly printed stick-on label, it was kind of a joke. Look, this a test vehicle so you can just leave your snide comments at the door. Here is the schematic.
As you can see it’s a bare bones 6F6 based class A amp with a single preamp tube, no controls and no tone shaping to speak of. There didn’t seem be any benefit in making it any more complicated than necessary. It’s capable of about 2 watts clean although for this series of tests it’s nowhere near that as I’m trying to avoid too much coloration by the output stage. I chose to take the measurements from the main output as it seems that this is how anyone would use an amp in practice. The input comes from one channel of a 24 bit 48KHz sound card via an attenuator. The output goes to a dummy load, second attenuator and thence to the line in of the same sound card. I’ll ask that we don’t go into any great debate, at least at this juncture, on the topology of the test amp, as I don’t believe it’s germane to the discussion. It is what it is so let’s move on.
A note on noise and hum: I did what I felt was reasonable to minimize hum. I tried a hum bucking pot on the heaters but I wasn’t happy with the result. I could see 60Hz at about -48dB with pot at optimum so I changed to use a DC heater for the preamp tube. The arrangement was something of a kludge but effective as it took that hum down to -60dB, comparable with the noise floor. Don’t try to copy this – it will only work in this design. I also changed the grounding to break a loop between formed by the safety ground in both the test amp and the desktop PC.
I used the software tools at my disposal, Adobe Audition, Audacity, and Right Mark Audio Analyzer. I used Audition to create a test sequence consisting of a 10 second 440hz tone, 10 seconds of silence to check hum and noise, 10 seconds of ‘A” major tones to listen for intermodulation products, 10 second of silence for microphonics and finally a couple of different styles of guitar-heavy music clips. The microphonics test was done by stroking each tube with a short length of a twisted pair of PVC insulated wire.
Here is an mp3 version of this sample result just so you can hear it. This is not suitable for any testing due to the extreme compression loss.
JJ ECC803S test.mp3
This is what the envelope looks like:
What I Did
The test sequence was replayed through ‘The Chump’ and the results saved. Microsoft format 32 bit floating point PCM ‘wav’ format files were used for all the tests. The input level was changed in each test so each had the same output level as closely as I could, within <0.5dB. This was to make it easier to compare spectra and to keep the distortion introduced by the output stage constant. Using Audacity I selected the portion of the clip of interest to zoom in and measure amplitude and spectra. Audacity has the ability to export the spectrum as text which I could then import into Excel. The Audio Analyzer generates its own test signals and again input levels were adjusted for each test to give a constant output level.
Copies of results were saved in ‘flac’ format. Each test is about 5MB long. I want to give you access to all these but I haven’t figured out the best way. I could attach them but I don’t want to be the reason for a slow response from this site or the one clogging up the nice-people-who-make-this-sites-possible's server with gunk. For now, pm me if you want them so I have some control over the bandwidth. The zip of all the files in ‘flac’ format is 112MB and includes all the audio analyzer results with their charts you can look at if you’ve a mind to do that.
Test 1 - Audio Measurements
I took DC volt measurements at the supply to the first stage and to both the first stage plates. The purpose was to check the basic operation of the tube to DC specs and to illustrate what happens if you replace a 12AX7 with say a 12AT7 without changing anything else. Using the Audio Analyzer I measured Frequency Response, Noise, THD and IMD.
Test 2 - Microphonics
I replayed the microphonics section and noted the peak value. I ignored any obviously anomalous peaks.
Test 3 Spectral Comparison
I plotted the spectra up to 5kHz on chart for all samples you The difference between tubes could be seen. Harmonics over 5Khz were lost in noise so there was no value in plotting them.
Test 4 Listening
The difference in noise level, hum characteristic and microphonics were obvious but the tone and music tests were much tougher. I’m only one set of ears so this is where you can help. If you’re interested pm me and I’ll give you the files. All I ask in return is that you post what you think of the differences.
Results
1) Spectra – 440Hz tone
The highest peak is the fundamental at 440Hz. You can see the higher harmonics to the right. The peaks on the left are hum at 60Hz and its harmonics. The frequency labels were removed as it was too cluttered.
2) Data & Scoring
Note: the gain is really level – an indirect measurement of relative gain.
More Notes
The DC operating point was very similar for all tubes, despite the obvious difference in construction. Noise had a greater spread of about 6dB between tubes of the same type. Gain was better, with about a 3dB spread, or 1.5dB per gain stage, a very small number. There is almost no difference in frequency response. Microphonics is very different between tubes, both in level and character – some tubes sounded quite bassy and others jangley or brittle.
With the caveat that more tubes should be measured and by this scoring, the ranking is:
Thoughts
I couldn’t reliably tell any difference between the music samples from different tubes. It seemed a little easier with the 440Hz tone but I wasn't able to translate that into better or worse sounding. You can see from the spectral plots that the biggest difference in harmonics is at the high end. It is known that THD is not a good indicator of pleasant sounding and that it’s better to use a square weighting of harmonics. Perhaps we could go further and assign a greater weight to the dissonant harmonics. It’s easy to do the calculation but difficult to get the listening tests to correlate them. In the end the differences are small. For certain, I didn’t hear anything like the differences that are ascribed to these brands by the distributors or in folk lore on Internet sites. I saw terms like ‘rich’, ‘lush’, ‘musical tone’ and ‘detailed’. But maybe that's just me. One site claimed ‘superb linearity’ for the Tung Sol when in fact it’s one of the worst. Again, I caution, most of the differences are tiny. Anyway, you don’t have to take my word for it and I may well be applying my own biases. You can listen for yourself.
More Things I'd Like to do
1) Effect of Output Stage
It would be helpful to remove the effects of the output stage, just because the artifacts are so small. To do this would require a high impedance buffer e.g. a cathode follower. Maybe use more cascaded stages.
2) Like Tube Spreads
The data doesn’t tell us if the variations are due to tube brand or manufacturing spreads. Many more tubes would need to be tested to do this. I did happen to have a total of six Tung Sol 12AX7’s. I found they were all very similar as far as THD, Noise and IMD but I didn’t run the full range of tests on them
3) Microphonics
A more repeatable method needs to be devised and an RMS value computed.
4) Hum on Heater supply
The heater supply needs to be improved. The current one had about 300mV of ripple on it. I’d like to be able to switch between AC or low ripple DC so the sensitivity of a tube to heater hum would be quantified.
5) Listening Tests
Listening tests need to be done and added to the scoring table. It seems that this could be very effectively accomplished by an Internet community encompassing a wide variety of individuals.
6) Reliability
There are no tests here for reliability; I cannot see how this could be done without considerable expense. Anecdotal evidence is not likely to be useful. I also think that reliability is terribly important thing to know and there is a dearth of data on this for new production tubes.
7) Clipping
It seems that behavior while clipping is crucial to a complete investigation. This needs careful thought on how such a test might be contrived as circuit topology has an effect and it might be unwise to try to generalize.
8) Power Tube Tests
Again, these would greatly affected by circuit topology and need careful thought. It seems certain at the very least that the operation in single ended and push-pull modes would have to be tested. There are also many variables, drive level, operating point, plate load and supply voltage to name a few.
I hope you've found this interesting.
Nick
Inspired by recent thread “Do tubes have different tones”, or something like that, I decided to embark on my latest folly. Get a bunch of tubes, do some sound tests and capture selected electrical data with the fond, but probably unrealistic, hope that the sounds can be correlated with the measurements and that we can all be enlightened by the experience.
So, without as much as a fanfare, let’s meet the players...
I picked a collection of new production 12AX7 type, but threw in a couple of other 12A*7 types just for interest. The reason for this choice is that I wanted to have information that would actually be useful to a wide population and the tubes they are likely to have access to. I certainly don’t have access to a bunch of NOS tubes and in any case it would blow away the tiny budget for this project. Needless to say, donations of NOS tubes for the project, especially your most precious, unusual and rare ones would be welcomed
For now, it’s just these preamp tubes. Power tubes may come later but will certainly require a lot more thought and effort.Next the test amp, irreverently but affectionately named after its predecessor, the Fender Champ, meet “The Chump”.
Sorry about the badly printed stick-on label, it was kind of a joke. Look, this a test vehicle so you can just leave your snide comments at the door. Here is the schematic.
As you can see it’s a bare bones 6F6 based class A amp with a single preamp tube, no controls and no tone shaping to speak of. There didn’t seem be any benefit in making it any more complicated than necessary. It’s capable of about 2 watts clean although for this series of tests it’s nowhere near that as I’m trying to avoid too much coloration by the output stage. I chose to take the measurements from the main output as it seems that this is how anyone would use an amp in practice. The input comes from one channel of a 24 bit 48KHz sound card via an attenuator. The output goes to a dummy load, second attenuator and thence to the line in of the same sound card. I’ll ask that we don’t go into any great debate, at least at this juncture, on the topology of the test amp, as I don’t believe it’s germane to the discussion. It is what it is so let’s move on.
A note on noise and hum: I did what I felt was reasonable to minimize hum. I tried a hum bucking pot on the heaters but I wasn’t happy with the result. I could see 60Hz at about -48dB with pot at optimum so I changed to use a DC heater for the preamp tube. The arrangement was something of a kludge but effective as it took that hum down to -60dB, comparable with the noise floor. Don’t try to copy this – it will only work in this design. I also changed the grounding to break a loop between formed by the safety ground in both the test amp and the desktop PC.
I used the software tools at my disposal, Adobe Audition, Audacity, and Right Mark Audio Analyzer. I used Audition to create a test sequence consisting of a 10 second 440hz tone, 10 seconds of silence to check hum and noise, 10 seconds of ‘A” major tones to listen for intermodulation products, 10 second of silence for microphonics and finally a couple of different styles of guitar-heavy music clips. The microphonics test was done by stroking each tube with a short length of a twisted pair of PVC insulated wire.
Here is an mp3 version of this sample result just so you can hear it. This is not suitable for any testing due to the extreme compression loss.
JJ ECC803S test.mp3
This is what the envelope looks like:
What I Did
The test sequence was replayed through ‘The Chump’ and the results saved. Microsoft format 32 bit floating point PCM ‘wav’ format files were used for all the tests. The input level was changed in each test so each had the same output level as closely as I could, within <0.5dB. This was to make it easier to compare spectra and to keep the distortion introduced by the output stage constant. Using Audacity I selected the portion of the clip of interest to zoom in and measure amplitude and spectra. Audacity has the ability to export the spectrum as text which I could then import into Excel. The Audio Analyzer generates its own test signals and again input levels were adjusted for each test to give a constant output level.
Copies of results were saved in ‘flac’ format. Each test is about 5MB long. I want to give you access to all these but I haven’t figured out the best way. I could attach them but I don’t want to be the reason for a slow response from this site or the one clogging up the nice-people-who-make-this-sites-possible's server with gunk. For now, pm me if you want them so I have some control over the bandwidth. The zip of all the files in ‘flac’ format is 112MB and includes all the audio analyzer results with their charts you can look at if you’ve a mind to do that.
Test 1 - Audio Measurements
I took DC volt measurements at the supply to the first stage and to both the first stage plates. The purpose was to check the basic operation of the tube to DC specs and to illustrate what happens if you replace a 12AX7 with say a 12AT7 without changing anything else. Using the Audio Analyzer I measured Frequency Response, Noise, THD and IMD.
Test 2 - Microphonics
I replayed the microphonics section and noted the peak value. I ignored any obviously anomalous peaks.
Test 3 Spectral Comparison
I plotted the spectra up to 5kHz on chart for all samples you The difference between tubes could be seen. Harmonics over 5Khz were lost in noise so there was no value in plotting them.
Test 4 Listening
The difference in noise level, hum characteristic and microphonics were obvious but the tone and music tests were much tougher. I’m only one set of ears so this is where you can help. If you’re interested pm me and I’ll give you the files. All I ask in return is that you post what you think of the differences.
Results
1) Spectra – 440Hz tone
The highest peak is the fundamental at 440Hz. You can see the higher harmonics to the right. The peaks on the left are hum at 60Hz and its harmonics. The frequency labels were removed as it was too cluttered.
2) Data & Scoring
Note: the gain is really level – an indirect measurement of relative gain.
More Notes
The DC operating point was very similar for all tubes, despite the obvious difference in construction. Noise had a greater spread of about 6dB between tubes of the same type. Gain was better, with about a 3dB spread, or 1.5dB per gain stage, a very small number. There is almost no difference in frequency response. Microphonics is very different between tubes, both in level and character – some tubes sounded quite bassy and others jangley or brittle.
With the caveat that more tubes should be measured and by this scoring, the ranking is:
- The best tube overall is the Electro-Harmonics 12AX7.
- The best for THD is the Sino 12AX7, but wow does that filament glow bright for a second or so after switch on!
- If you want gain, go for a Sovtek 12AX7LPS or a Mullard 12AX7
- For lowest microphonics, Tung Sol 12AX7
- If low IMD is your thing, buy a JJ ECC803S, just make sure the microphonics aren’t going to be a problem.
- For first stage use, where you want low noise and microphonics, pick an EH 12AX7.
Thoughts
I couldn’t reliably tell any difference between the music samples from different tubes. It seemed a little easier with the 440Hz tone but I wasn't able to translate that into better or worse sounding. You can see from the spectral plots that the biggest difference in harmonics is at the high end. It is known that THD is not a good indicator of pleasant sounding and that it’s better to use a square weighting of harmonics. Perhaps we could go further and assign a greater weight to the dissonant harmonics. It’s easy to do the calculation but difficult to get the listening tests to correlate them. In the end the differences are small. For certain, I didn’t hear anything like the differences that are ascribed to these brands by the distributors or in folk lore on Internet sites. I saw terms like ‘rich’, ‘lush’, ‘musical tone’ and ‘detailed’. But maybe that's just me. One site claimed ‘superb linearity’ for the Tung Sol when in fact it’s one of the worst. Again, I caution, most of the differences are tiny. Anyway, you don’t have to take my word for it and I may well be applying my own biases. You can listen for yourself.
More Things I'd Like to do
1) Effect of Output Stage
It would be helpful to remove the effects of the output stage, just because the artifacts are so small. To do this would require a high impedance buffer e.g. a cathode follower. Maybe use more cascaded stages.
2) Like Tube Spreads
The data doesn’t tell us if the variations are due to tube brand or manufacturing spreads. Many more tubes would need to be tested to do this. I did happen to have a total of six Tung Sol 12AX7’s. I found they were all very similar as far as THD, Noise and IMD but I didn’t run the full range of tests on them
3) Microphonics
A more repeatable method needs to be devised and an RMS value computed.
4) Hum on Heater supply
The heater supply needs to be improved. The current one had about 300mV of ripple on it. I’d like to be able to switch between AC or low ripple DC so the sensitivity of a tube to heater hum would be quantified.
5) Listening Tests
Listening tests need to be done and added to the scoring table. It seems that this could be very effectively accomplished by an Internet community encompassing a wide variety of individuals.
6) Reliability
There are no tests here for reliability; I cannot see how this could be done without considerable expense. Anecdotal evidence is not likely to be useful. I also think that reliability is terribly important thing to know and there is a dearth of data on this for new production tubes.
7) Clipping
It seems that behavior while clipping is crucial to a complete investigation. This needs careful thought on how such a test might be contrived as circuit topology has an effect and it might be unwise to try to generalize.
8) Power Tube Tests
Again, these would greatly affected by circuit topology and need careful thought. It seems certain at the very least that the operation in single ended and push-pull modes would have to be tested. There are also many variables, drive level, operating point, plate load and supply voltage to name a few.
I hope you've found this interesting.
Nick
Comment