No.55
Hi-Fi Audio
EF86, 6N1P

The purpose of this memo is to report the data obtained from testing the Svetlana EF86 and 6N1P devices for equivalent input noise. I used the noise tester device detailed on my web page at http://members.aol.com/sbench101. This device is capable of resolving to less than 0.2 uV EIN in both flat gain and equalized (simulating expected performance for RIAA or NAB equalization).

We compared the performance of 4 Svetlana EF86s against 4 new-old-stock Amperex EF86/6267 devices. I also compared 4 Svetlana 6N1Ps against a mixture of 6DJ8, 6922 and 7308 types (total of 45 devices), mostly high-priced NOS types.

Results (EF86):

1: Finding the best (lowest noise) operating point. I did this with the EF86s operated in triode mode (screen grid connected to plate). Supply voltage was held constant at 200 volt DC. The plate load resistor and cathode bias resistor were varied (cathode resistor is bypassed for AC) to obtain best noise figure. This was a very “broad” optimum, but best performance occurred with a plate load resistor of 62k and a cathode resistor of 1.1k. This results in just under 2 volt bias and a plate voltage of about 110 volts. The Amperex “best point” was with a slightly higher cathode resistance, but all noise data shown used these values.

2: Establishing circuit gain.

Circuit gain for the device set up at this bias point was about 25 in triode mode. The tubes were also tested in pentode mode by connecting the screen grid directly to 120VDC. The plate and cathode resistor values were retained. This kept the idle point about the same. The plate voltage was about 120V in this mode. Circuit gain in this configuration was about 90.

3: Initial Results. (triode mode, flat response)

Part of the test capability is the ability to change the heater voltage. For some tube types, this has the effect of decreasing the noise. When I lowered the heater voltage, I noticed the noise dropped. But when I returned the heater to 6.3V, the noise dropped even further.

I then noted this effect on all the Svetlana devices, but not on the Amperex devices. However, there is some precedence for this. Reviewing the Electronic Designers Handbook (Landee, Albrecht, Davis, McGraw-Hill 1957), I noted a reference for low noise instrumentation amplifiers suggesting the “heaters should be cycled numerous times”.

Based upon this information, I cycled the EF86 heaters between 3.15V and 6.3 volts, on a 5 minute at normal and 5 minute at low heater voltage cycle. I repeated this for 12 cycles (2 hours). At the end of this period, the cathode material appears to be stabilized; I was not able to get any further improvement.

4: NOISE FIGURES FOR PRECONDITIONED SVETLANA EF86s VERSUS PRECONDITIONED AMPEREX EF86s

Legend:

TMF=Triode mode, flat. Vpp=200V, Rp=62k, Rk=1.1k, bypassed. This produces a bias voltage of about 2 volts, with about 110 to 120V on the plate. Noise is the input referred equivalent RMS noise. (=output noise/gain)

TMRIAA=Triode mode, RIAA (sort of) equalized. See my noise tester schematic for details. Operating point remains the same.

PMF=Pentode mode, flat. The only difference is the screen fed with 120V directly. All other characteristics and operating point remain the same; stage gain is higher.

PMRIAA=Similar to triode mode RIAA, except screen fed with 120V directly.

AVERAGE NOISE FIGURES:
SVETLANA EF86 0.70 uV
AMPEREX EF86 0.89 uV

These data are more-or-less in line with good triodes; you'll note that the dual triode tests (on my website) are with both sections in parallel, which gives the triodes an advantage by improving the reported noise by about 3 dB. As expected, the partition noise in the pentodes, operated in pentode mode, reduces their noise figure, although the extra gain may help the overall system noise figure. Notice how much the Svetlana EF86 have improved! They are now substantially better than the Amperex. (The device numbers are the same as reported above; device #3 improved over 7 dB.)

5: Other notes

The Amperex tubes show almost no microphonics and are surprisingly immune to external hum sources, almost not requiring a shield in low level applications. The Svetlanas are reasonably immune to hum pickup; more immune than most low level triodes. They exhibit some low level of microphony. In all, Svetlana has a really good product here.

Noise Results: The Svetlana 6N1P

I had obtained some Svetlana 6N1P dual triodes some time ago. These turned out to be quiet devices, worthy of consideration for low level signals. The data below summarize their performance when compared with a mixture of conventional 6DJ8, 6922 and 7308 devices. I noticed a similar effect on cycling the heaters that I reported above. However, the average improvement was only 1.5 dB. This is somewhat different, as the two triodes in each tube were paralleled in order to compare them with data previously collected on the 6DJ8 family.

The best operating point found was with a plate resistor of 25k and cathode resistor of 390 ohms. (200 volt supply). This is slightly different than the optimum point found for the 6DJ8 family which was 25k plate load and 320 ohm cathode resistor. The data represents the average of 4 sample 6N1Ps and the average of 45 assorted tubes of the 6DJ8 family.

Svetlana 6N1P Flat 0.35 uV
Svetlana 6N1P RIAA 0.34 uV
6DJ8 FAMILY Flat 0.35 uV
6DJ8 FAMILY RIAA 0.5 uV

Notice that in RIAA phono applications, Svetlana 6N1Ps are significantly better than the 6DJ8 family. In fact, only the out-of-production RCA red-base 5691 and GE 6072A are able to better them on an average basis, per the test results listed on my web site.

**The information provided in this application note is intended for general design guidance only. The user assumes all responsibility for correct and safe usage of this information. Svetlana Electron Devices does not guarantee the usefulness or marketability of products based on this material.

Back to Technical Bulletins Listing