04 November 2006

The Tube Scream

Oh, the irony of it all.

In the October 22nd blog, I mentioned how I fear schematic typos. And then, in the October 31st blog, I made a big one. At least it was a big one in my eyes, though many will never have spotted it.

Let me explain. The 6BQ5-based OTL headphone amplifier circuit I originally drew up left out a key Aikido-like touch to enhance the amplifier’s PSRR figure. I didn’t spot it, because I didn’t go through the usual step-by-step explanation of how the circuit works; instead, I only offered a link to an article that I had written that had explicated the circuit. Oops.

And while I bemoaned the e-mail burden with which I daily struggle, I didn't explain that I had fallen particularly far behind of late because two weeks ago, my son’s teacher and one of his classmates drowned. A week spent trying to explain to a three-and-a-half-year-old how something like this could happen and what it means, along with attending various funeral-related events, put me far behind in work in general and answering e-mail in particular.

I reasoned that in just a little more time than it would have taken me to answer the original lost e-mail, I could write a blog entry addressing the same issue.  But I didn't have time to do it complete justice last week.  So I decided to write a second part; and when I gave the original entry a second reading, I spotted the schematic typo. (October 31’s blog now holds the corrected schematics.)

Okay, now in greater detail.

6BQ5-based OTL headphone amplifier circuit revisited
In spite of the little time spent writing the blog entry, the 6BQ5-based headphone amplifier was actually carefully designed, with the part values not casually chosen. The result is that the headphone amplifier acknowledges and anticipates the power supply’s failings, presenting the requisite amount of injected power-supply noise into the output stage to cancel the power-supply noise from the output. Below is the bad schematic that I had posted.

Notice the parts in red. This capacitor serves as a power supply decoupler, filtering away much of the 510-volt power supply rail’s noise. So what’s wrong? Isn’t this what 99% of tube gurus would do? (I was tempted to write 99.9% of tube gurus, but upon some reflection, I realize that percentage would imply that the world holds at least 1,000 tube gurus, a number I doubt.) The problem is that if this decoupling capacitor terminates into ground, the negative power supply rail noise will be superimposed upon the top output triode’s grid signal. How’s that? The differential amplifier’s common cathode resistor attaches to the negative power supply rail and any current variations through this resistor that result from the resistor seeing a varying voltage potential will be relayed to the plate resistors, as they all fall on the same current path.

But in the correct schematic below, we see the decoupling capacitor terminated into the positive-power-supply rail, not ground.

Now the negative power-supply-rail noise will meet and cancel with the positive power-supply-rail noise, as the cathode and plate resistors values were chosen to ensure a noise null at the output. If the cathode resistor were larger, then the output would see positive power-supply-rail noise leaking out; if the resistor were smaller, then the output would see negative power-supply-rail noise leaking out instead. (Ideally, we want injected negative power supply rail noise to equal the positive power-supply-rail noise, so that cancellation is complete.)

How does this trick work? Both the positive and negative power supply rails deviate from being pure DC sources, carrying the rectification noise residue. Fortunately, the noise on each rail is nicely related, equal in magnitude, differing only in being 180 degrees out of phase to each other. (This only holds true when both rails see the same current load, as they do in this amplifier.)

The complete headphone amplifier is shown above. By the way, speaking of noise reduction, I just reread something that I had written in the second issue of the Tube CAD Journal, which is worth repeating every seven years:

Noise, its causes and elimination, will continue to be explored in future issues. The motive is not hard to find.

Noise constitutes a large part of what is condemned as the "tube sound," by those who favor solid-state electronics. Few tube supporters would defend hiss or hum as being euphonic. Noise stands guilty as charged, as it robs the musical experience of subtle nuances. Much of what is magical in a great recording is the air, minute details, and weak hall reverberations that make the recording come alive in our living rooms.

By masking these subtleties, noise reminds us that what we are hearing is not live, but only electronically reproduced. Of course, live performances are often marred by coughing, hacking, talking, creaking chairs, and glass clinking. But this noise is random and sharp, unlike the continuous and dull drone of power supply hum; no one in the audience imagines that he is hearing a reproduced musical event in his living room because of this live noise. (In fact, a few rare recordings so perfectly capture this live noise that its reproduction actually helps to convince us that what we are hearing is live.)

Given the choice between low distortion and low noise, choose low noise. The difference between .1% distortion and .01% distortion cannot be heard; a reduction of noise from -60 dB to -80 dB can. If you want to improve the sound of your tube equipment, lower the noise. Of course, low noise and low distortion are not mutually exclusive and it would definitely be better to have both.

Here is the irony of the situation. Most solid-state gear is very quiet, because of the low distortion race among stereo manufacturers. In order to please the total harmonic distortion meter, an amplifier must be both clean and quiet, as the meter cannot distinguish between distortion and noise. Whereas, the tube audio designer, having no faith in total harmonic distortion measurements, holding to his belief that no meter can beat his ears, forgoes the requirement of extremely low noise operation.

Some tube audio designers are very pessimistic about the noise given off by their designs. They tell us that the vacuum tube is an inherently noisy device and that it is simply not possible to match the solid-state gear in low noise operation. We might believe them, if it were not for counterexample offered by some of the best tube gear: tube microphones, preamps, and amplifiers that do not hum along with the music or hiss like tea kettles. (I built a direct coupled electrostatic headphone amplifier that is so quiet that a friend used to borrow it to evaluate the noise of hand picked transistors for an ultra-low noise preamp he was building. The headphones and amplifier revealed more information about the amount and quality of the noise than did looking into an oscilloscope.) There is simply no excuse for one tenth of a volt of hum from the output of a tube power amplifier.

We are not saying that the tube is just as quiet as the transistor or FET, as, ultimately, it is not; only that much of what is built with tubes could be much quieter.

Amplifying low output MC cartridges reveals the absolute noise floor of a tube. The quietest tube is much noisier than the quietest FET. The 2SK147 FET is eerily quiet; after spending years looking at the 6DJ8's hash noise on an oscilloscope, the flat thin line of noise from the 2SK147 makes one check to see if it is actually connected to the probe. This praise for the FET's low noise does not constitute a wholesale endorsement of the FET's use for all audio circuits, only a recommendation for this limited use: amplifying the micro-volts out of an MC phono cartridge sufficiently to give the vacuum tube a chance to take over.

Making tube equipment quieter requires a solid grasp of where and how the noise is interjected into the output signal.

6BQ5 issues
John Atwood has set me straight about the old Svetlana SV83. They are relabeled 6P15P pentodes, a Russian equivalent of an EL83, whereas the 6P14Ps are the Russian equivalent to the EL84. In John’s own words:

I was looking at your latest Tube Cad Journal posting, and I noticed that you implied that the SV83 is equivalent to the EL84. It isn't. It is really a Russian version of the EL83, the 6P15P, which is a video pentode, more similar to a 6CL6 or 6AG7. It may work in a 6BQ5 socket, but the bias points are somewhat different. The pin-outs are also somewhat different, but if one was aware of both pinouts, the sockets could be wired to be compatible with both.

I have used both the conventional (i.e. consumer-grade) and mil-spec versions of the Russian 6P14P (the mil-spec version has the suffix "EB" in Cyrillic), and both are good, although I especially like the EB version. They sound good, and last a long, long time. I had a quad in my old Heathkit SA-2 amplifier, which I used as my computer sound amp, which ran every day. Before these, I used US-made 6BQ5/7189s and they would last a year or two at the most before one "went critical" and the others were loosing gm. I had the 6P14P-EB tubes in for over 2-1/2 years before I took the amp out of service, and they all tested perfect! These tubes are available on Ebay for about $2.50 each from Russian and Ukrainian sellers. I have gotten some from these sources and have had essentially no problems with the purchases.

And here is an interesting link SV83 Output tube warning! Below is the pinout for the 6P14P and 6P15P pentodes.

Speaking of small, 9-pin power pentodes, I love the 6CW5/EL86, which is a powerful, low-voltage pentode. (The EI version cost only $5 and I am sure that you can still pick up NOS American 6CW5s for just a little more.)

                ET-T3021

POWER-AMPLIFIER PENTODE

DESCRIPTION AND RATING

FOR AF POWER AMPLIFIER APPLICATIONS
The 6CW5 is a power-amplifier pentode designed for use in the audio-frequency power-output stage of television and radio receivers and in high-fidelity amplifiers.

GENERAL

ELECTRICAL
      Cathode—Coated Unipotential
       Heater Voltage, AC or DC                  6 3 Volts
       Heater Current                                0.76 Amperes

Direct Interelectrode Capacitances*
       Grid Number 1 to Plate                        0.6 pf
       Grid Number 1 to All                             12 pf
       Plate to All                                         6.0 pf

MECHANICAL

       Mounting Position—Any
       Envelope—T-6½, Glass
        Base—E9-1, Small Button 9-Pin Outline Drawing—EIA 6-4
        Maximum Diameter 7/8 Inches
        Maximum Over-all Length 3-1/16 Inches
        Maximum Seated Height 2-13/16 Inches

MAXIMUM RATINGS

DESIGN-CENTER VALUES
      Plate Voltage                                                         250 Volts
       Screen Voltage                                                       200 Volts
       Plate Dissipation                                                      12 Watts
       Screen Dissipation                                                1.75 Watts
       Peak Screen Dissipation                                         6.0 Watts
       DC Cathode Current                                               100 Milliamperes

       Heater-Cathode Voltage
           Heater Positive with Respect to Cathode             100 Volts
           Heater Negative with Respect to Cathode
                DC Component                                            150 Volts
          Total DC and Peak                                             300 Volts
       Grid-Number 1 Circuit Resistance
          With Cathode Bias                                             1.0 Megohms

TERMINAL CONNECTIONS
       Pin 1—Internal Connection
        Pin 2—Grid Number 1
        Pin 3—Cathode and Grid Number 3 (Suppressor)
        Pin 4—Heater
        Pin 5—Heater
        Pin 6—Internal Connection
        Pin 7—Plate
        Pin 8—Internal Connection

CHARACTERISTICS AND TYPICAL OPERATION

AVERAGE CHARACTERISTICS
     Plate Voltage                                                    170 Volts
       Screen Voltage                                                 170 Volts
       Grid-Number 1 Voltage                                    -12.5 Volts
       Plate Resistance, approximate                       23000 Ohms
       Transconductance                                         10000 Micromhos
       Plate Current                                                     70 Milliamperes
       Screen Current                                                 5.0 Milliamperes

CLASS A1 AMPLIFIER
    Plate Voltage                                                    170 Volts
     Screen Voltage                                                  170 Volts
     Grid-Number 1 Voltage                                     -12.5 Volts
     Peak AF Grid-Number 1 Voltage                          9 9 Volts
     Zero-Signal Plate Current                                     70 Milliamperes
     Maximum-Signal Plate Current                            70 Milliamperes
     Zero-Signal Screen Current                                  5.0 Milliamperes
     Maximum-Signal Screen Current                          22 Milliamperes
     Load Resistance                                             2400 Ohms
     Total Harmonic Distortion,approximate                  10 Percent
     Maximum-Signal Power Output                           5 6 Watts

6N1P input tube
One reader asked if the 6N1P could be used instead of the 12AX7. It can, but at a price. First of all, it will offer less gain, which means less feedback will be available, which might be important to those seeking the lowest output impedance and distortion. Second, the 6N1P should be run much hotter than a 12AX7, 5mA per triode, not the 1mA that a 12AX7 can get away with. So what’s the big deal, 10mA versus 2mA? The big deal comes in the form of bigger decoupling capacitors. Where the 12AX7-based headphone amplifier only needed 2µF capacitors, the 6N1P version will require five times larger capacitors, 10µF decoupling capacitors. In addition, the two bipolar power supplies may become unbalanced, as the negative power supply rail will be taxed greater than the positive rail, as the current through the differential amplifier starts at the negative power supply rail and ends at the 510V rail. (The workaround might be to use more capacitance on the negative power supply rail than the positive rail.)

In practice, this amplifier may work well enough that these added costs are worth paying. Remember, the 6N1P can be much quieter than the 12AX7.

Another input tube that I would experiment with is the 12BZ7, which also holds a mu of 100, but with half the rp of a 12AX7.

//JRB