06 September 2019                                                                     Post 476

Rumors of RMAF
This year's RMAF will be held in September, not the usual October, and at a different venue. I will post afterwards, as usual, but so that ye be not troubled I wanted to get something out beforehand.

Bastode-Based Split-Load Phase Splitter
A few posts back, I described transistor-based phase splitters for use in OTL amplifiers. These designs were not that different from tube-based split-load phase splitters, differing in that tubes do not come in the P flavor.

The bastode is a cascode circuit made out of N flavor and P flavor devices, such as a triode and a PNP transistor or MOSFET. Here is 12 possible arrangements.

Well, we can force the sixth bastode, into a split-load-phase-splitter circuit, as I did back in post 294.

This phase splitter offers the same gain as the triode-based split-load phase splitter, either unity or two, depending on your perspective. In contrast, what I am about to display is a different sort of phase splitter, whose major difference is that we take advantage of the bastode's potential of offering screaming gain and the possibility of employing negative feedback. For example, the following is a power buffer that uses the bastode transistors both to create two anti-phase signals and to compare the output to the input and make the needed adjustments to bring the output in line with the input.

The collector resistors, R2 and R3, see the same variation in current flow, so the same signal amplitude develops across them, but in the opposite phase. The two output MOSFETs are both N-channel devices, so at the top MOSFET draws more current, the bottom MOSFET draws less, as the top sees its gate voltage rise, while the bottom sees its gate voltage fall. The 1K µF bootstrap capacitor ensures equal source-to-gate voltage swings between the two MOSFETs. Resistor, R1, sets the idle current flow through the MOSFETs, albeit in an indirect fashion. The larger the resistor value, the less current flow will be; the smaller the value, the higher the current flow. We can use a mono-polar power supply.

If we are willing to run the output MOSFET in strict class-A, we can use an auto-bias circuit to set the MOSFET idle current.

The OpAmp-based DC servo amplifier at that top monitors the current flow the 0.5-ohm drain resistor to the voltage drop across the diode (another MUR410G) voltage drop on the left. If the output stage draws too much current, the voltage drop across the current-sense resistor will increase, which will prompt the OpAmp's output to swing negatively, thereby bringing down bias voltage to both the output MOSFETs.

On the other hand, we could go the anti-2gm route with a lower idle current.

The two MUR410G rectifiers only switch on when the opposing MOSFET cuts off, allowing the still engaged MOSFET's effectively transconductance to double, thereby ensuring constant-transconductance.

If you are wondering where the tubes are, here they are.

The ECC99 output triodes idle at a heavy 18.4mA and both get an unbypassed cathode resistor; yet, the output impedance is less than 1-ohm. How's that possible? The bastode stage offers a bunch of signal gain and all of it is returned as negative feedback, as the bastode's inverting input (the MJE340's base) connects directly to the output stage's output, resulting in a unity-gain output stage. The 12AU7 input tube realizes a gain roughly equal to its amplification factor (mu), which for a 12AU7 is between 17 to 18. The headphone amplifier's PSRR is also fine, coming in at better than -6dB in SPICE simulations. In addition, the current-swing balance between top and bottom triodes is excellent.

More Dang Crossovers
I planned on giving crossovers, both active and passive a good, long, rest. At the same time, I knew that topic was nowhere near exhausted. For example, we saw Yamanaka and Baekgaard crossovers with filler drivers and 2nd-order and 3rd-order crossover slopes for the woofer and tweeter, but what about a 4th-oder for the woofer and tweeter, with a 1st-order slope for the filler driver? Was that possible? It is.

Here is what I found worked fairly well.

The woofer and tweeter see 4th-order filters made from cascaded Butterworth filters with a Q of 0.707. Another name for this crossover type is the Linkwitz-Riley 4th-order or the LR4, as the final Q is equal to the 0.707² or 0.5, which results in the woofer and tweeter being down by -6dB at the crossover frequency. In addition, this crossover results in a doubling of the nominal speaker impedance at the crossover frequency. In other words, if 8-ohm drivers are used, the peak impedance equals 16 ohms. This explains why I specify a 2R impedance for the filler driver, as this will flatten the impedance back to R. Of course, it's hard to find good 16-ohm fullrange drivers, so the best procedure is to use 4-ohm woofers and tweeters, so that an 8-ohm filler driver can be used. In addition, the filler driver must deliver an SPL that is 6dB higher than the woofer and tweeter's SPL. Unlike a Linkwitz-Riley crossover, all the drivers must be wired in phase with each other.

The frequency response is flat, as is the phase response. Here is a three-cycle tone burst.

Note how cleanly the sum output starts and stops; also note how the filler driver puts out twice the signal that the woofer and tweeter do, yet the sum is equal to the woofer and tweeter output.

In general, high-order passive crossover are a pain, as the part tolerances must be extra tight and the driver impedance should be as close to ruler flat as possible. In contrast active crossovers are much more easily implemented, as 1% capacitors and resistors can easily be found.

This crossover uses a differential amplifier to create the required slopes for the filer driver. The differential amplifier's 5k and 10k resistors should be 0.1% types to ensure good performance. The differential amplifier allows us two break with quasi-two-way aspect and go for a more conventional three-way crossover, without having to change the differential amplifier resistor values.

Music Recommendation: Harp vs. Harp
Only a few new albums instantly thrill me. Gregoire Maret's and Edmar Castaneda's Jazz album, Harp vs. Harp, is one of the few. The idea behind this album, at first, sounded silly to me, but ultimately struck me as truly cool: why not combine a stringed harp and harmonica (aka a harp) so these gifted musicians can produce latin-flavored jazz? Béla Fleck and Andrea Tierra also contribute to the effort.

Needless to say, as you have no doubt figured out my MO by now, Harp vs. Harp is available at Tidal. Do give it a listen, as your speakers long to bounce to the jazzy Latin beats.

//JRB

User Guides for GlassWare Software
Just click on any of the above images to download a PDF of the user guides. By the way, all the links for the PCB user guides shown at the right now work.

//JRB