Discussion of Dave’s Results
Once again the NOS tube wins the power race. The EH average is 11% lower and the JJ is 32% lower. The Hickok gm numbers show only 65% of nominal for the EH and 93% of nominal for the JJ’s. One should bear in mind though, that tube tester gm figures are not very good indicators of actual power tube performance.
- The new manufacture 7591’s fall short of true NOS by 11 to 32%.
- The JJ-7591S is substantially weaker than the EH-7591A. However, in a less demanding application like the ST-40, it might perform as well or better.
- The EH-7591A shows higher distortion and low tube tester gm. It is unclear whether the distortion is due to actual nonlinearity or reduced gm lowering feedback factor.
- If I had to choose between the JJ and the EH, it might be the JJ for the ST-40 and the EH for the ST-70a. The mods in the ST-70a increase feedback factor, allowing it to handle the issues with the EH better.
- Note that the large physical size of the EH might preclude its use in some equipment, such as the Scott LK-72.
I would like to thank my good friend David Gillespie for the measurements and other information he contributed for this article.
Addendum #1 — Improved testing using class AB1 bias
Since this article was published, Dave has developed an enhancement of his method of testing output tube power. The basic approach is unchanged but now a bias current is added. This better represents the demands of class AB1 operation, whereas what was being tested before was essentially class-B operation. Most commercial tube audio power amps use class AB1 bias.
The effect is to increase the average plate current, at which power is measured. Normally, fresh tubes have far more cathode emission than they need to sustain maximum plate current. That is required to protect the cathode. Such tubes should show about the same power output, with or without the AB1 bias. However, if cathode emission is weak, the additional DC bias current is more burden and power output could suffer. The new AB1 test results are shown at right, performed on the same tubes tested before. In addition, test results for four more samples of the EH7591A are included.
The new test method slightly increased the output power measured for the reference Sylvania NOS tubes and for the Electro-Harmonix samples, tested before. The minor increase is attributed to additional cathode heating provided by the higher plate current. The original two EH samples had shown over 20% difference in power output, so we were happy to test the additional four, to clarify their true performance. It turns out that the originals are at the extremes of the six tested. Average power output of the EH’s is 91% of the historical average of NOS tubes. Not a bad showing in terms of power output.
Unfortunately, the JJ-brand 7591S tubes didn’t fare well in the new test. Their original weaker performance became still weaker with the additional bias current, delivering only an average of 63% of the power that NOS tubes deliver. You might be tempted at this point to dismiss the JJ’s as a weak contender, next to the EH’s. However, recall from the previous testing, that the EH’s showed unusually low Gm in a Hickok tube tester. While tube tester readings should be regarded with suspicion, if low Gm is confirmed, it could explain why the EH’s showed higher distortion in amplifier operation. Bear in mind too, that only two JJ’s were tested. That isn’t much of a sample. Both Dave and I are working on new testers which will accurately measure Gm. Stay tuned for further developments...
Thanks to Dave Gillespie for the additional test data.
Addendum #2 — High accuracy Gm tests refute Hickok data. EH tubes partially vindicated!
It is well known that almost all commercial tube testers from the golden era give poor quality Gm measurements. For one thing, the grid drive levels used are generally far too high. Also, the plate voltage is pulsating DC, resulting in a blur of characteristics across a wide range of voltages.
To improve this situation, Dave has built a new tube tester, designated the TT-10, which operates tubes under controlled DC bias conditions, drives the grid with a reasonably low AC voltage and measures AC plate current. The results are accurate Gm values which can be compared directly to published databook standards.
Using the TT-10, the same 7591 tubes which were tested above, were retested. Here are the new Gm results, along with the original Hickok Gm figures and the original power output measurements:
As you can see, the Sylvania NOS tubes closely match the ideal databook Gm values, giving confidence in this approach. While the true Gm of the EH tubes is comparable to ideal, the Hickok reads them as less than half of normal. The true Gm of the JJ tubes is at least 80% of ideal, yet the Hickok reads them far lower. Interestingly, the Hickok reads the EH’s much lower than the JJ’s when, in fact, the true Gm of the EH’s is higher. So the Hickok tube tester cannot be trusted even for relative comparisons.
It is interesting to observe that output power tends to vary in the same direction as true Gm. However, this is not to say that it is proportional. Clearly, there is still a place for maximum output power testing, for power tubes.
In conclusion, these tests exonerate the Gm performance of the EH tubes, which had been called into question by the Hickok results. However, there still remains the fact that EH 7591’s showed higher distortion than others in actual amplifier operation. To summarize, we can say that:
- There is no longer reason to think that the EH 7591’s have insufficient Gm.
- The higher EH distortion is probably due to tube nonlinearities.
- Hickok test results for these tubes are nearly worthless.
Addendum #3 — Evaluation of the new Tung-Sol 7591A, including real-world and lab test results
In earlier sections of this article, we examined new-manufacture, 7591 tube offerings under the JJ Electronic and Electro-Harmonix (EH) brands. Note that New Sensor Corp. owns at least the following brands: EH, Tung-Sol, Sovtek, Genalex, Mullard and (in North America) Svetlana.
Within the past few years, a new 7591A tube has been released under the Tung-Sol brand. You can see it at right next to a classic, Westinghouse, Eico-branded 7591A. For the first time in the modern era, we have newly-made 7591 tubes which actually look like the original! Even the internal construction looks reasonably similar. This comes as a relief to folks who have had trouble fitting other new 7591s into their equipment. For example, it’s been reported that the case doesn’t fit with previous new tubes in the H.H. Scott LK-72 and 299D and the Fisher X-200 models. There should be no problem with the new Tung-Sol tube, assuming that the minor increase in height seen above isn’t an issue. Aside from solving the fit problem, vintage audio collectors are no-doubt pleased that the new 7591A doesn’t look out of place in their equipment.
How Well Does It Perform?
As we saw above, the EH 7591A and, to a lesser extent, the JJ 7591S can perform adequately but they didn’t quite match original NOS in terms of output power. First, a few caveats about that statement:
- We can only afford to test a small sample of tubes, so this is not “scientific” and should be taken as an indication, rather than hard fact.
- It’s been a long time since we tested the EH and JJ products, so they might have changed. (If only we could afford to buy more tubes...)
- Our vacuum tube power testing is an evolving art and methods have improved over the years. (Check out the companion article about this.)
Performance in the Eico ST-70a
To get a real-world look at the new 7591A, we ran three pairs of them in the same amplifier and against the same RCA and Sylvania tubes that were used above, 7 years ago. Only the power amp of the ST-70a was used, as before. DC bias, DC balance and AC balance were adjusted for each tube pair. Details of the ST-70a are covered near the beginning of this article and in the ST-70a mods article. In the new table below, we’re seeing output power higher by about 3% for the RCA and 7.6% for the Sylvania, relative to the original readings. The primary filter capacitor was replaced recently, which could account for part of the difference. Distortion of the new tube at half output voltage is equal to the RCA, so linearity is excellent.
More detailed (but less reader-friendly) data is available by clicking here. In any case, we need to be careful about comparing this table to the previous one above, by using performance relative to the RCA tubes. Previously, the JJ7591S produced 89% of the RCA power and the EH had 96% (1kHz, 1% distortion). Output power results in the new table show the Tung-Sol at 92% of the RCA, so it’s higher than the JJ but lower than the EH tube. The advantage of a real-world test is that it shows the actual results you get in a real amplifier. The disadvantage is that it’s just a point-sample. A different amplifier, with different biasing and different power supplies, might perform quite differently. Before we close the book on output power performance, let’s look at the lab power and vacuum tube analyzer tests.
Laboratory Power Performance
Power testing in the lab was performed in a manner similar to the method used by Dave Gillespie, except this approach uses standard lab equipment instead of a custom-built vacuum tube power tester. I’m grateful for him sharing the techniques used. The test procedure is feasible for serious hobbyists and it’s covered it in detail with a companion article here. The power test operates the tube in a relay socket with a load resistor to represent the output transformer primary. We drive the tube grid with a sine wave source to cause maximum plate voltage swing. The peak-to-peak value of the swing gives us the maximum power that one side of a push-pull pair can deliver. Test results from eight new Tung-Sol 7591A tubes, along with a reference NOS RCA and four EH 7591A tubes are shown in the table below. More detailed (but less reader-friendly) data is available by clicking here.
You can see that the RCA tube is indeed delivering the full power defined by the databook plate curves of the original 7591A. It’s remarkable that the tube comes as close as it does to ideal. One of the Tung-Sol tubes, TS-7, was apparently defective in some way, as its output was only 81% of ideal versus the 96% average of the others. Discarding the bad one, I consider the 96% score to be very good, though not quite as good as the EH value of 98%. I guess one could spot them one bad egg. An occurrence of one count cannot be considered indicative of a trend. Still, if it’s included in the overall Tung-Sol power average, it pulls it down to 94%. In view of the fact that this new tube is so much smaller than it’s new-manufacture competition, I guess a slightly lower output power result can be accepted. I’m just thrilled that they finally produced one that’s physically similar to the original. By the way, no correlation between output power and Gm was found among the Tung-Sol tubes.
Gm Performance in a Vacuum Tube Analyzer
Measurements of Gm were performed with a newly designed vacuum tube analyzer, designated as the VTA-101, seen at right. This is the first time that this instrument has been mentioned publicly, so you can consider this a sneak preview of the upcoming article which will cover it in detail. The design attempts to achieve something like “HP quality.” [Referring to the instrument company.] Accuracy and stability were paramount.
In the table above, the term, “True Gm” is used as a column header to distinguish these readings from those given by many of the vintage tube testers. Very few were actually capable of measuring accurate Gm, as illustrated in Addendum #2 of this article. Anyway, the Gm of the Tung-Sol tubes varied from 91 to 98% of the ideal value, averaging about 94%. I consider this a very good showing, although the EH tubes averaged 99.9%. Small signal Gm isn’t a very critical parameter for power tubes.
Summary of the Tung-Sol 7591A
From the limited sample analyzed it was found that:
- It delivered 92% of “NOS” * power in an Eico ST-70a amplifier, versus 89% for JJ and 96% for EH.
- Lab power tests showed 96% of ideal power, excluding the impaired unit or 94% with it included, versus 98% for EH.
- Gm averaged about 94% of ideal versus 99.9% for EH.
- Medium level distortion in the ST-70a was 0.031%, exactly the same as the RCA “NOS”. The EH tubes weren’t retested for this but showed 0.11% in the earlier tests.
Overall, the Tung-Sol 7591A performed pretty well in these tests. The weaker power output in the amplifier is balanced somewhat by a very good showing in the lab power tests. Gm was fine, if not stellar, and distortion was excellent. One thing that should be taken into account with the power tests is, we really are looking at small variations which would not be audible. For example, 92% power represents a loss of only 0.4dB relative to 100%. To me, the advantages of the smaller size and having a tube that doesn’t look out of place in vintage equipment could far outweigh such minor variations among products. As far as the significance of the single impaired unit, it’s hard to judge. One would need a bigger sample to take that seriously, since the other seven tubes showed good performance. I have to give this critter a big ‘ol thumbs up!
I would like to thank Dave Gillespie for his sage advice and encouragement for this 7591 investigation, particularly in the area of vacuum tube power testing.
* The RCA and Sylvania tubes used in the ST-70a were originally NOS and have been lightly used. Since they are used, we refer to them as NOS in quotes.
Add your comments here...