Discussion in 'SBAF Blogs' started by purr1n, Dec 26, 2018.
2018-12-26 AN HONEST DISCUSSION ABOUT TUBES
AN HONEST DISCUSSION ABOUT TUBES
I recently got this Amplitrex AT1000 tube tester. With the number and quality of vintage tubes in decline and the sheer number of tube amps owned by myself and many friends in the El Lay area, I felt this was a necessary expense. There is also a geek side to me that really wants to understand what is really going on with these tubes. I mean, what the heck does 2100 or 1900 mean on the Hickok meter for those tubes sold on eBay? So really, I'd be happy if I could recoup half the cost of the unit by findings the gems and selling them for a premium. As an aside, I've put a few aside for SBAF that I will sell at reasonable cost taking into account my time and expenses.
The Amplitrex can do plate curves; but so far, I have been using it to get the most basic of measurements: emission and transconductance. There are two ways to get these results, fixed biased and auto bias. I used auto bias because the majority of tube amps out there, at least in the personal audio hobby, are auto biased. I’m not going to go too deep into the process, but I will quote a few things from the manual below. The manufacturer of the AT1000 has its own way of measuring these things, so take it for what it is.
The tube is biased into conduction in one of two modes. In Auto Bias Mode, the grid voltage is adjusted downward automatically until the specification plate current is achieved ... Once the plate current is flowing, the unit will proceed automatically to test the transconductance ... The actual measured transconductance is shown on the right side of the third line of the display. This value is determined by actually making the tube amplify a 1 kHz sine wave signal, and then measuring the AC plate current that results.
Transconductance normally reads close to the specification value. If transconductance reads low, this means that tube gain may be less than that expected in its application circuit. If it reads higher than the specification, more gain can be expected in the target circuit. If emission was low (low measured grid voltage or low measured plate current), the transconductance can be expected to possibly (but not always) be low as well.
In Auto Bias Mode, if the actual grid voltage was higher than the specification value, this means that the tube has strong emission and can be expected to have a good amount of life remaining. If the actual grid voltage is very low or is zero, the tube is bad and should be replaced.
In the center of the fourth line, a percentage of Transconductance is shown. 100% means that the GM of the tube was the same as shown in the data sheets. A reading of less than 100% means that the tube’s transconductance is only that percentage of what the data sheet shows. As for Emission, a GM percentage over 100% shows that the tube is performing better than specification. The percentage values shown makes it easy to tell at a glance whether the tube is good, weak or excellent.
You may ask, what constitutes a “good” tube? Well, this is difficult to say, and depends more upon your own criteria .. Tubes showing emission or GM less than about 70% can be classified as bad ... However, you must be the one to determine for yourself if a tube with 70% emission or GM is suitable for use or not. For example, a tube showing 70% may perform very poorly in one piece of equipment, yet be OK in another.
So here I have two real-world examples, the results for a well-used vintage Hytron 2A3 and a new production Shuguang 2A3. I like the sound of both of these tubes. The Shuguang is dirt cheap and sounds and drier than the vintage Hytron.
New production 2A3
So these results are a little bit interesting. The well-used vintage 2A3 actually measures closer to spec for the emission and transconductance tests than a very little used new production 2A3! Perhaps there were lost secrets to the materials science back in the day after all. This is not at all unexpected because there was a time before transistors when tubes were the only game in town. Does this explain why the vintage tubes "sound better" than the new production tubes?
Not being satisfied with only the results from the tube tester, I decided to take a very basic measurement on the AverLAB analyzer: a 1kHz sine wave at -3dBU into a 330 ohm resistor load from the EC Aficionado (no feedback version using WE396A driver). 330 ohms is about nominal impedance of a Sennheiser HD600 / HD650. -3dBU is approximately 0.55V, which is fricking loud on the Sennheisers. The volume knob to get zero gain on the amp is near the 1 o'clock position.
1kHz sine wave at -3dBU into 330 ohms
Eddie Current Aficionado (no feedback version using WE396A driver)
New production 2A3
1kHz sine wave at -3dBU into 330 ohms
Eddie Current Aficionado (no feedback version using WE396A driver)
Holy cow, the vintage tube looks horrible! Actually, it’s not, but it does explain quite a bit about the differences in sound between these two tubes, particularly the "richness" that I spoke of with the vintage tube. To put things in perspective, the second order distortion product at 2kHz, is about 55db below the fundamental. That’s like the difference between the sound levels of a jet plane flying overhead at 1000ft and a quiet library. I don’t want to get too deep into measurements and how they correlate to human hearing right now, but I at least wanted to point this out. I will be honest and say that I do like the sound of distortion, not easily audible distortion, but distortion on the cusp of human hearing, and with a pattern of each successive harmonic being progressively lower. Please keep in mind that this measurement is very specific and does not even get close to quantifying the entirety of audio reproduction and human hearing. Anyone who does not understand this will be referred to another more suitable web forum and be permanently banned from SBAF.
One thing does immediately come to mind: Tube weirdos who speak of the special sonic qualities of specific vintage tubes, e.g. "1931 yellow lettering, chrome flashing on lower quarter of glass, production batch BFG", as if they were vintages of wine may be correct for their own individual tubes, but not necessarily correct for the type of tubes in general. That is unless we know how the tubes measure on the tube tester or in distortion characteristics. As we near 2019, there is really no such thing as NOS (new old stock) tubes anymore. We can’t assume anything unless we know the condition of the tubes.
Finally, with the few tubes I have tested so far, I have noticed that the Hickok tester results for eBay items do usually translate to good emissions and transconductance results with the AT1000. There have been a few duds, but every seller I have dealt with has been excellent.
never buying new production tubes of any ilk, and always dependent on NOS tube testing measurements of some sort can leave one queasy based on ebay sellers- or more comfortable if it was Uptone or Brent Jesse.
it's good to see a thread about independent findings so better questions can be asked/understood.
there is some location in Denver that has a tester (IIRC) and allows you to bring-yer-own-tubes in for testing.
have been too busy (or trusting of ears?) to warrant a trip and full afternoon of electron measurement calculations.
Excellent work, Marv!
Do you by any chance have some EML-s laying around?
That distortion data is frightening!
Perhaps more effort in the new tubes manufacturing went into mechanical damping.
The new prod tubes do 'sing' less with those HF heaters.
I remember this article from a while back on how amplitrex doesn't do DHT testing well
What happens if the spectrum measurements are repeated at -10 and -20 dBu?
Also note the difference in residual mains noise, especially at 240 Hz.
Those graphs really illustrate the point well, thanks. I'm actually surprised how much stronger the harmonics are on the old tube. I wonder if that is something that can be generalized to new vs old tubes overall. Have advances in tube manufacturing technologies created tubes with less harmonic distortion, and therefor less of that desirable component of the sound?
That first graph reminded me of one I first saw here; https://kenrockwell.com/audio/why-tubes-sound-better.htm The second one not so much......
Let's not excessively fault the cheapest new 2A3 on the market.
Kudos for doing this.
I wish I had more time to contribute, but I can support Merv’s results with data based on our own tube matching and testing experience:
1. The distortion characteristics of tubes vary widely in circuit, not just from manufacturer to manufacturer, but batch to batch and sometimes tube to tube. The tubes that vary tube to tube you don’t want—not NOS, not new, not no way, no how. We choose NOS and new production tubes based on consistency...and even then they all go through a computerized matcher.
2. As AtomicBob pointed out, the hum characteristic of tubes varies widely in circuit as well. Some old production and new production tubes have vastly different hum characteristics (we’re talking 30dB here!). Again, these are the ones we avoid and sort out.
I’m looking forward to more results!
It should be noted that the SAME tubes may exhibit different performance at different operating levels.
ZDT Jr with stock tubes at 0 dBu and 300R load
ZDT Jr with stock tubes at -10 dBu and 300R load
ZDT Jr with stock tubes at -20 dBu and 300R load
At 0 dBu the 2nd harmonic is higher than the third.
At -10 and -20 dBu the third harmonic is dominant.
Also note how many other harmonics drop off faster at lower levels.
I have bought tubes from tctubes.com and also had them check tubes I bought off eBay. Tyler also uses an Amplitrex 1000 to test and match tubes before shipping. He routinely uses the Amplitrex 1000 in fixed bias mode:
Tyler goes into more detail on why he feels most tubes should be tested using fixed bias at Interpreting Amplitrex AT1000 tube tester results.
AN HONEST DISCUSSION ABOUT TUBES - ADDENDUM
As far as microphonics or the effects of Craig's high-frequency AC filament heaters (which run near 50kHz), it's hard to say anything specific regarding new production or old stock tubes. I've heard both new and vintage tubes exhibit that physical grinding noise with Craig's amps. I suspect that the metal structures inside the tubes are resonating or beating against the 50kHz signal.
Here is an analysis of the microphonics of the new production Shuguang and vintage Hytron tubes about two seconds after I tap the glass of the tubes with my finger.
New Production Shuguang 2A3 vs Vintage Hytron 2A3 Microphonics
The vintage 2A3 is the one with higher distortion, the new production has more ringy behavior
The vintage tube's microphonics settles very quickly. The new production Shuguang microphonics are still evident after about eight seconds.
New Production Shuguang 2A3 Microphonics
Two seconds and eight seconds after finger tap
We can see some residual ringing around 700Hz still after eight seconds
On the topic of two tubes of the same type measuring differently in terms of distortion or having different hum characteristics, we can see this being a little evident with this so-called matched pair of vintage Hytron 2A3s. One tube has about 7db higher harmonic distortion, lower noise floor, and a small evident 120Hz hum spike compared the other.
Vintage Hytron 2A3 Match Pairs
1kHz at -3dBU into 300 ohms, EC Aficionado
As far as how to measure and match tubes, we can debate the merits between using auto-bias and fixed-bias. I would argue that the best way would be to pick a method and see where the accumulated data leads (with both good, not so good, and bad tubes in the sample set). Or perhaps utilize both methods for the sake of completeness. The tubes in the matched vintage 2A3 pair above actually measure very close to each other in both the auto-bias and fixed bias tests. However, there is quite a bit of difference in harmonic distortion.
This, of course, brings up the question: should a simple distortion test on the analyzer also be added after checking for emission and gain?
@purr1n Marv this is cool (and interesting)! I wanted an Amplitrex once upon a time. TBH I never did see a reason to prefer HF heating to good low noise DC. I guess it'd be really interesting to see the FFTs with Craig Uthus vs Rod Coleman filament heating, all else being equal. Obviously not realistic with our once resident AF super modder now absent.
@atomicbob do we care about -90dB 240 Hz mains residuals in the "bad" case other than for academic purposes?
It depends. -90dB from a loud rendition of music with content at 240 Hz won't be audible due to psycho-acoustic masking. But note that these graphs are calibrated with dBu on the y-axis. That has a reference of 0.775V. So a steady hum of -90dBu at 240 Hz will be audible during quiet musical passages played at low volumes on electro-acoustic transducers with high sensitivities in a quiet listening space.
Denon D7000 will emit 32 dB SPL at -90 dBu and Campfire Audio Solaris will emit 35 dB SPL at -90 dBu. Listeners of music with high dynamic range may find such mains noise with those cans / IEMs annoying. For my preferred HD800-Jmod headphones, -90 dBu will only emit 11 dB SPL and not be audible even in my quiet acoustic lab. I would likely prefer the harmonic content displayed by the new production 2A3 over the vintage version.
To add, the thing with amplifier hum (and noise) is that it tends to be sticky, at the same level regardless of the output / position of the volume knob.
Harmonic distortion however goes up or down according to increasing or decreasing output respectively, and will be masked by the fundamental or other signals if the distortion is sufficient low (-60db to -80db).
I didn't want people susceptible to fundamentalist objective religion thinking that because one can hear hum at -80db with sensitive headphones such as the ATH-D2000, they could hear distortion at -80db. Two different things really.
Separate names with a comma.