PCB update
Last night, while picking up some take-out Chinese food, my fortune cookie read,

“In the end, no one will remember how quickly you completed a job, only how well you did the job.”

I was so taken back by the aptness of this wisdom-filled strip of paper to the long wait for the new PCBs that I completely forgot to append it mentally with the obligatory “in bed” ending. (Does everyone do that, or is that a local tradition?) At any rate, the wait is near over; all the new boards were made and shipped to me today. Now it will take about two days for me to fill and send out all the back orders.

In the meantime, you can prepare for your Aikido project. Below is the new circuit and heater schematic for the mono PCBs.

Note all the new jumpers. Why so many? The Rev. B PCBs are more flexible than Rev. A, as they now can accept many more output tubes, such as the 5687, 6900, 7044, 7119, 7370, 7892, and E182CC (and of course, the old standards, such as 6AQ8, 6BQ7, 6BS7, 6DJ8, 6CG7, 6GM8, 6H30, 6FQ7, 6N1P, 6N27P, 12AT7, 12AU7, 12BH7, 12DJ8, 12FQ7, 5963, 5965, E80CC, ECC81, ECC82, ECC83, ECC86, ECC88, and ECC99).

In addition, the heater arrangement has been redesigned so that many more input and output tube combinations are now possible. For example, the 9-pin mono boards can be used with 6.3V and 12.6V power supplies, as the heater jumpers allow tubes like the 12AX7 and 12BH7 to be wired as 6.3V tubes. In other words, it is now possible to use a 12AX7 as the input and a 6DJ8/6922 as the output tube (this has always been possible with the stereo 9-pin boards).

Click on image to see closeup

Speaking of the Aikido stereo 9-pin boards, they have been revamped into Rev. B as well. The changes were not as bold as in the mono boards, but the small changes will be welcomed by many. For example, all the tubes now find a few ventilation holes under their sockets; the layout has been tidied up a bit; two new jumpers have been provided, so that the two output coupling capacitors can be bridged on the board; and holes have been added to the sides of the capacitor C!, the large out coupling capacitor, so that a tie wrap can be used to secure the capacitor to the PCB. Moreover, the output stage’s bottom triode can now have its cathode resistor bypassed. Why would someone want to? Many have used 9-pin stereo boards as the foundation of a headphone amplifier and now it will be easy to build the following circuit.

Broskie OTL update

Soon after creating the Broskie OTL, my first thought was, "How good a headphone amplifier would it be?" The schematic below shows one possible arrangement.

The 6DJ8/6922 triodes are moderately stressed by the 120V cathode-to-plate voltage that each experiences.  But the more I thought about it, the more I wanted to design a lower-voltage version; I had been telling so many tube-based headphone amplifier builders to use a power isolation transformer to develop the B+ voltage, I felt I should heed my own advice. So the circuit was redesigned to work with only 170V worth of B+ voltage, as shown below.

Well I finally got around to running a few more SPICE simulations on the Broskie OTL circuit. Interesting, indeed. The distortion is quite low, as we would expect from the higher open-loop gain that it realizes compared to the simpler Broskie cathode follower, for example. What I didn’t expect was the harmonics to reveal a single-ended influence. Usually, a push-pull amplifier’s harmonics look like a saw’s teeth, with the odd-order harmonics peaking high above the even-order harmonics.

For example, (from a 6DJ8-based Broskie cathode follower circuit with a B+ of 170V and the same 300-ohm load impedance and 1Vpk output) note how the 3rd harmonic is up nearly 40dB compared to the 2nd harmonic in the graph above, but only +4dB in the graph below.

What is going on here? My guess is that since both top triodes work in tandem to drive the load, just as both bottom triodes work together, the combined efforts yield a more balanced distortion signature than one would expect from a balanced, push-pull amplifier, with the odd harmonics only a tad bit stronger than the even harmonics. In other words, something is slightly out of balance here. Now, the question is, "Should it be set to perfect balance or should it be left alone, as its present sonic signature is more likely to please the ear?" A good question indeed.

So for those who always wonder to what good use a new circuit in this journal can be put, the answer is that many high-end line amplifiers, DACs, and CD players sport two types of outputs, balanced XLR and single-ended RCA jacks. This headphone amplifier could be used with the balanced output, leaving the single-ended outputs for driving single ended power amplifiers; or, it could be used in the recording or radio studio, where balanced feeds need to be rendered single-ended for headphone listening. Moreover, this circuit would work well with voltage-output DACs that offered balanced outputs, as its distortion into high-impedance loads is almost nonexistent (well, at least as far as SPICE simulations are concerned; reality will—no doubt—differ).

Feedforward shunt regulators update
I know it has been only a week since I posted my last entry on feedforward shunt regulators, but I expected more than just three e-mails on the subject.

(Of course, there may have been many more that just could not get through my e-mail’s many layers of spam filtration. I have been on the sending end as well as the receiving end of this problem. This has become a huge problem for me and most other businesses. The statistic I have seen is that 80% of e-mail is spam. What happens when the percentage is 99.9%? Will any e-mail get through?

The additional big problem I face is that the anti-spam algorithms are prejudiced against foreign-sent e-mail, which is about 80% of the e-mail I receive. What’s to be done?

I am in favor of a $0.01 tax on each e-mail message sent. Your ISP would be charged by the taxman, so the IRS and ISPs would keep careful track. In addition, if your computer has been infected by a virus and has been transformed into a spam robot, you will quickly get the problem fixed, when your ISP hits you with a $500 monthly bill (ISPs could impose a $1 a day e-mail limit, with an e-mail notice that you have reached your limit). Now, $0.01 per message would be nothing for average person or business, but it would stop spammers cold, as they send out millions of spam messages a minute. Of course, some countries would not impose the tax, but then we could simply create filters that deleted all e-mail originating from that nation (besides, when has a government not jumped at the chance to create a new tax?). No doubt, organized criminals and lone hackers would hack into the network; but they would be only a few, compared to the millions sending spam now.)

Okay, back to circuits, the question was, "How do you set up an adjustable-idle-current feedforward shunt regulator?" The answer is easy. Just take some resistance from the top voltage-dropping resistor and add it to the bottom series resistor; then add a potentiometer and a few resistors, as shown below.

Or, what might prove to be a better solution: use fixed bias, instead of cathode bias. For those of you who switched over to solid-state rectifiers, here’s what you can do with the free 5Vac winding and the empty rectifier socket.

Yes, 300Bs are expensive, but no one would expect you to use an NOS WE 300B, just a cheap Chinese or East-European copy. (I love the idea of using three output tubes on a push-pull, monobloc amplifier; imagine a Dynaco MK-3 with three KT88s.)

The real advantage to an adjustable idle current through the feedforward shunt regulator is that tubes differ from each other and from themselves over time. In other words, the perfect null that was achieved today, may not be perfect six months from now, or when a new tube is used. In addition, the feedforward shunt regulator differs from the average sloppily-built single-ended amplifier in that deep nulls require precise series resistor values. Here’s why: a tube’s transconductance is not a constant; like its plate resistance, it varies with plate voltage and cathode current. The graph below shows the relationships between rp, mu, and gm in a 6SN7 with a fixed 250V on its plate. The X-axis displays the current as the key variable. Note how the mu, the least real attribute of a triode, is the flattest, while the transconductance bends about.

(By the way, a common complaint that I hear is that I always specify too high an idle current.  For example, most tube gurus run their 6SN7s with only 1mA to 2mA of idle current, whereas I never go below 4mA, preferring 8-10mA. Guilty as charged. Now, let’s put this in perspective: the same tube fancier who whines on and on about the higher current halving the life expectancy of his tubes is usually quite happy to pay some tube guru $800 to change the coupling capacitors and sprinkle some magic dust over his preamp, although the guru’s modifications cannot come near to delivering the sonic benefit that derives from raising the triode out of its low-current mire. Just look at the graph and at the mu plot. Where does the 6SN7 become linear, above 5mA, or below? Where is the rp the lowest, at 1mA or at 17mA? Let's get real, here: at 1mA, the 6SN7’s rp is almost 40,000 ohms and its transconductance is a measly 0.4mA/V. What price does sonic glory come at? Current and voltage. Expecting sonic glory at low currents and voltages is like expecting a race car to get good mileage. In addition, the dreaded triode sleeping disease infects the tubes that are run too lean, not the ones that are run too hot. So a tube may last twice as long at the left side of the graph, but if sounds twice as bad, do you really want to listen to it for twice as long?)

Back to the topic, to give you an idea of just how tweaky an adjustment is needed for a deep null, take a look at the graph below, which shows the results of a SPICE array of simulations with the series, voltage-dropping resistor’s value as the swept variable.

And here is the array setup:

Note that the resistor's value starts at 100.9 ohms and ends at 101.1 ohms. The deepest null occurred at 101.0 ohms and is represented by the thick red plot line. Note the almost 40dB difference for a lousy 1 ohms!

Since 1% resistors are not good enough for this degree of precision, adding a trimming potentiometer makes a lot of sense.

//JRB