I’ve been buying tubes for a long time, and arguably I buy more tubes than most folks. In the last 10 years alone, I’ve documented several hundred tube amp repairs, many of them requiring tubes. Over time, I’ve read a lot of opinions and evaluations on tube brands and sound. In particular, the mysterious conversation often is centered on: “which ones have what characteristics?”
Over and over I’ve been asked by truly curious customers, “which brand of tube should I get for this amp?” or something similar. I’ve always bristled at the question for many reasons, but when pressed, I usually answer with the name of the brand of tubes I have most recently had the least issues with. It’s as honest of an answer as any. And it might just be the best answer. Quality seems to vary tube-to-tube, and not so much brand-to-brand. By the way, has anyone else noticed that Lot Codes for tubes are rare? This is in contrary to most all electronics production for the last 45 years. Look at ANY transistor or IC…
The reason for that reply is that most tubes sound about the same. I know it, and most experienced musicians, engineers or honest tube guys know it too. The problem is that I really don’t like claims that aren’t backed up, by me or anyone else. This has been rolling around in my mind for literally years. Just recently, I finally did something about it that I had been threatening to do for a long time: The TET.
TET stands for Triode Evaluation Tool. It looks like this: A special test circuit which is carefully built to represent the most-often used preamp topology, the Resistance-Coupled Amplifier stage. In this version, everything is built for linear response, large quality plastic film coupling caps, a very bypassed regulated high voltage supply, regulated DC filament circuits, a symmetrical layout, shielded cables for all signal paths, star-grounded, and enclosed in a recycled all-metal enclosure.
The software to support it is an audio Real Time Analysis program called TrueRTA, which has been around quite a while.
How TET works: The idea is that TrueRTA can use a USB audio I/O interface (“sound card”) to send audio out (as Pink Noise) and at the same time read back the audio in, and perform the FFT analysis, giving the frequency response it sees in real time. It can also perform sweeps, and importantly, it can calibrate out any system irregularities. This is the key to the TET function.
Two step calibration: to calibrate the TET, I use a “dummy tube”, which is nothing more than a tube base with a couple of very good film caps, one connected for each triode element, so the grid of section A is coupled directly to the plate output of section A, and likewise for the B section of a typical dual triode. Ins connected to outs. Easy. It’s essentially a tube with zero gain and no noise. The value of the cap is chosen to allow good low frequency response.
For TrueRTA, you can calibrate out any of the USB sound card’s effects by simply performing a test-cable loopback of output to input, which performs a frequency response for the sound card itself. In doing so, you normalize that response out (flatten it) with a correction factor (automatically by the SW), and you now have a system that will tell you the frequency effects of anything placed into that cable loop. This is very similar to how an RF Network Analyzer is used.
Anything put into the signal path that deviates the calibrated system response from flat will be the fault, or effect, of that device. I use this for examining DI boxes, testing for frequency response at various drive levels to see what they are capable of. In the case of the TET, we go a step further, and include the entire fixture with the dummy tube installed and the system powered up and use that normalization to flatten out the whole TET system. That way, any tube that is installed will have its effect on the frequency response made clear, with the addition of showing the signal gain of the tube, a nice bonus that you can measure while you are at it. This allows you to compare the relative gains of one side of the triode with the other. Not surprisingly, internally the sections of a tube are usually a little different between each other by a dB or so. You can also check gain over changes in plate voltage doing this.
Another bonus measurement we can make is to evaluate system noise and distortion characteristics. I have a switch installed at the tube input that shorts the grids to ground, so anything the TrueRTA “sees” coming back from the fixture is just the garbage made by the tube itself, mostly hum and noise.
By sending in a single tone, say 100 Hz, you can clearly see the harmonics generated by the tube and, importantly, how many dB down they are from the 100Hz signal and how strong the harmonics are relative to just each other. This is a great way to see a true coloring of the sound that some hi-fi enthusiasts get very excited about when they notice one tube sounds a bit different than another in controlled critical listening tests.
(note the B+ choke and input shorting switch were added after these pics were taken)
Mythbusting: Let me note here, that I’ve always said that tubes have a pretty flat response, and the TET easily proves it. The results I am typically seeing for most triodes I’ve tested are flat within +/-1dB across the 20Hz-20kHz bandwidth. That’s pretty damn flat. It’s way flatter than other things you do to a signal, like the circuit you use it in, the transformers you choose, the cables, the tone controls you are probably addicted to, and most of all, the speakers and room you are using. Relatively speaking, the frequency response of the tube plays almost a full 0% in all of that equation, meaning changing brands of tube is not going to make for “leaner mids”, “offer a more robust bass”, deliver “extra glassy treble articulation” or any other marketing nonsense. It won’t happen. Now… are there systems where you might notice the slight differences in a tube’s harmonic character or noise? Sure! Especially with familiar sound sources and when the entire signal chain is good enough not to mask the very, very subtle differences by their own larger issues (noise, phase, reflections, distortions) that are also present. Most of us don’t get to enjoy that kind of listening equipment and environment, in fact it’s dead-rare. Your speakers and room will probably contribute much larger effects than the tube will, truth be told, but marketing persists. Funny that they never publish any real data. Did you ever notice that?
Noise: Probably the most useful aspect of the TET that I see is in being able to really see system noise generated by the tube. I added the input short test when I considered the noise floor and wondered what the connection to the sound card might be contributing. For me, the hardest part about building the TET has been battling the noise floor inherent to the system, I find that even with very careful grounding, layout, extra filtering and screening, there is still 60Hz hum, but I found in the end that it’s coming from my regulated 300V power supply. There is no ground loop there, it’s just that the regulation, while good, is not really great. The plots I’m getting show it around 75 to 80 dB down from the test signal level in the best cases. Considering a CD has about a 90 dB S/N ratio, that’s not a lot of hum, you wouldn’t likely hear it under normal conditions, but it’s there along with it’s harmonics at 120, 180, 240, 300-Hz etc. I’m looking into getting a better power supply, but adding in a HV power choke to the fixture with some extra filtering after it did help a lot.
Conclusion: Well, building this was a fun bit of research and development, but most importantly it’s a great tool for evaluating popular triodes in a meaningful, reproducible way. I’ve already been able to very clearly measure the issues with a few noisy tubes I had laying about. In one recent case, a customer had excess noise in a tube-buffered FX loop used in a boutique hand-wired amp. My new replacement tube had no affect whatsoever on the issue, whereby we spent another solid hour or two looking at everything else (including RF spectrum analysis with a loop probe to prove it wasn’t oscillating) only to discover that my “new” tube was just as noisy as the one of his I had removed. A third tube (also new) in the same circuit was dead quiet. According to the TET, that first “new” tube had a very clear-to-see 15 dB of extra wide-band noise coming out of it at all times. The TET displayed it perfectly. I love it. Now… all I need to do is build one of these for the big beam power tubes. And I will.