As one tube becomes popular, i.e. expensive, we'll use something else. Because we can.
-- Fred Nachbaur
Output: 37W / 8 Ohms
Frequency Range: 11.0Hz -- 50KHz (-3.0dbv; 8 Ohms)
Variable Negative Feedback (0 -- 12dbv)
Active Screen Voltage Regulation
Low Noise, Hybrid Power Supply
2) 6BQ6GTB or 6BQ6GA
The type of test used to assess the performance involves subtracting the input test signal from the output. A perfect amp would produce a perfectly flat line, as there would be no difference. Whatever remains when doing this test must be distortion products. All three of the tests below were run at max output power, just below the threshold of clipping.
This particular frequency gives the deepest null, indicating minimal phase shift at the midrange. Most of what is seen on the screen is interference from a 50KW AM BCB station 30 miles to the north. Whatever distortion products that may be there are submerged beneath the interference.
Since there is more phase shift at the lower end of the audio band, the null isn't so deep. The residual is virtually pure sine wave, indicating minimal distortion. Here, there are some very small switching glitches, a natural consequence of running Class AB1, and the fact that no output transformer is perfectly balanced. There is no ringing in evidence, which would be a problem, as would obvious crossover distortion, such as occurs with Class B operation.
At this frequency, some h3 is becoming apparent, since no output transformer is perfect at the high end of the audio band. Of course, the third harmonic of 15.6KHz is 46.8KHz, and beyond the limit of audibility. The actual results are still pretty good.
These tests confirm that the design does indeed perform quite well, and agree with the listening tests.
The A Number One problem with audio design these days is the "cult" status attained by a good many VTs. The SET fraternity have driven the prices for directly heated power triodes such as the 2A3, 300B, 45, and to a lesser extent, the 6A3 and 845 through the roof. Nor have the "zero bias", high-u, RF finals, such as the 811A, escaped the attention of SET designers and builders. This doesn't apply to just the expected rare NOS specimens left over from the "good ol' days" that have been hiding in forgotten warehouses, stockrooms, attics, and the inventory of retired TV and radio repairmen. Even the new production types that ought to be no more expensive than any other new production types are out of sight. If the market can bear the price, you can bet that the highest possible prices will be charged.
This phenomenon doesn't just apply to a select few of DHTs, but to pentode and even small signal types. The prices for KT88s, KT66s, 6BQ5s, 6L6s, 6V6s, 12AX7As, 12AT7As, 6SL7s, 6SN7s, have been steadily creeping higher. Sometimes, prices reach some ridiculous proportions. Paying three figures for a small signal tube is crazy. So what can we do about that? The answer is to avoid audiophile fads in both tube types and circuit topologies. So we take another good look at the frequently neglected and often maligned "TV t00bz".
Originally intended for use as a horizontal deflection power amp, the 6BQ6GTB has a characteristic that's much different from that of the audio finals. Though rated like a 6V6 (PD= 12W for both types) the 6BQ6GTB can provide more than twice the power (6V6: 14W; 6BQ6GTB: 30W+) as it is designed to source large currents at lowish Vpk's. This characteristic is just what you need to pull big currents through the horizontal deflection coils of wide scan, black-and-white, TV CRTs. The high current operation also makes for a low RL, and therefore, easier to design, and less demanding to manufacture, output matching transformers. As with types such the 807, the low screen voltage rating pretty much precludes Ultralinear and trioded operations, the former of which is verging on attaining fad status (complete with a whole bunch of audiophile "wisdom" as to the topology's "magical" characteristics); the latter a necessity since the selection of audio power triodes is quite limited, due to the rapidity of the coming of the beam power revolution that reduced demand for audio power triodes to commercial non-viability. To help improve the audio linearity of the finals, active regulation of screen voltages, fixed bias, and grid drivers are included. The unusually large current demand puts considerably more stress on power transformers, making a monoblock design the preferred construction for this project.
The front end uses two 6BQ7As, a dual small signal triode originally designed for use as a VHF cascoded voltage amp. The front end is a differential (a.k.a. Long Tailed Pair) cascode, a topology "borrowed" from solid state practice, though rarely ever seen in VT audio amps. Without going into excruciatingly boring detail, the cascode is a direct coupled, two stage voltage amp which offers both the high voltage gain and the low CMiller of a pentode gain stage with a sonic signature more like that of a triode. Thus, the cascoded LTP is the only gain stage required. It also includes active tail loading for improved harmonic balance.
Another unusual feature not seen in commercial designs is variable gNFB. The listener can select anywhere from 12dbV of gNFB, to none at all. Thus, the amp can be tailored to different listening preferences and/or types of music.
The grids of the finals are driven with a 6FQ7 medium gain dual triode -- this type being essentially a nine-pin mini version of the octal 6SN7 -- configured as cathode followers. The use of a bipolar power supply allows for direct coupling to the 6BQ6 grids. Fixed bias operation improves linearity, doesn't waste power in a cathode bias resistor, provides stable grid bias under load, and does not require a coupling capacitor that would lead to nasty clipping when inevitably over-driven. A fast rising transient that drives the grids of the finals positive will turn on the parasitic diode formed by the control grid and cathode. The grid current that flows through this parasitic diode will charge the coupling capacitors negatively (This is how Class C RF amps derive most -- or all -- of their operating bias.) While that charge leaks off, the finals operate closer to plate current cutoff and in a more nonlinear part of the plate characteristic. Even if you don't hear obvious distortion, this tendency will still cause a noticeable overall degradation of sonic quality. This behaviour is made even worse when feedback is applied since the loss of the feed back signal will cause the closed loop gain to increase, deepening the clip. Since there is no capacitor to accumulate a negative charge, the operational bias is not disturbed. Since the cathode follower driver can source the current the grids need during the overdrive, the finals can transparently slip just a bit into Class AB2 to avoid outright clipping at the grids. Most of the clipping behaviour is due to OPT core saturation. This improves the overall sonic performance since the occasional clip is simply is not heard.
Fixed bias also avoids a related problem that occurs with cathode bias. A transient that causes more plate current to flow will charge the cathode resistor bypass capacitors to a higher voltage. This, too, results in operation closer to cutoff. In extreme cases, the amp can go silent as the finals are completely cutoff. This is the "blocking" problem. (Guitar players have their own term for this: "farting out".) Needless to say, it does nothing to improve the sonic quality. This can be even worse than excessive charge on coupling capacitors since cathode resistor bypass capacitors are even larger, and have longer time constants.
Cathode follower drivers are also more capable of driving current into the input capacitance of the finals. This becomes important at high frequencies since the control grids don't see the true signal value until the Ci + CMiller + Cstray is fully charged to that value. The higher the signal frequency, the faster you need to charge that input capacitance. The faster the capacitance charges, the greater the current demand. This is the slew rate problem, and it doesn't just apply to solid state amps. An amp that slews too slowly will distort and smear the high end of the audio band. Inadequate attention to this potential problem is all too evident in all too many VT designs out there. There is no way that a high-u, small signal triode that operates at sub-milliamp plate currents can deal with the capacitance of most power finals.
It has long been known that power pentodes of both sorts (beam "tetrodes" as well as pentodes with suppressor grids) operate with better linearity if the screen voltage is tightly regulated, and if the screens operate into as low an AC impedance as possible. Of course, Ultralinear, trioding, and audio finals that operate the screens and plates from the same power supply at equal voltages avoid the necessity for a screen supply. For lower cost of production of mass marketed consumer equipment, this is all good for the "bottom line" even if it does sacrifice sonics in the process. If you take a look at the plate characteristic, you can see how much the screen current varies with Vpk. Under static conditions, the screen current doesn't amount to more than a couple of milliamps. The screen current can rise to as much as 25mA under full load operation. A resistive voltage divider, even with a large bypass capacitor, isn't up to that task. A simple series dropping resistor certainly isn't either. Inadequate attention to screen regulation is all too common, and is a big part of the inferior sonics of pentode finals that the SET fraternity complain about. The active regulator also has the advantage of representing a very low AC impedance. This helps to get degenerative stray signal voltages off the screens that would otherwise lead to greater harmonic distortion. Screen regulation is rarely seen in commercial designs, even those from the Great Age of Hi-Fi that preceded the introduction of the first solid state audio amps during the mid-1960s. This is another case where the balance between sonic excellence and cost reduction was settled in favour of the latter.
The HV power is supplied via a full wave, balanced rectifier with a CLC ripple filter. The power diodes being the 5U4GB directly heated, dual large signal diode designed for this purpose. Since the power transformer has plenty of voltage reserve to accommodate the much greater forward voltage of the vacuum diode, this is a viable option for this project. The 5U4GB is one of the "beefiest" of the high vacuum diodes, and has more than enough current sourcing capability to meet the requirements of this particular project.
Since the 5U4GB, like all vacuum tubes, is a high voltage, low current device, it doesn't have the peak current rating that a silicon diode does. This puts a limit on the size of the first filter capacitor. This also tends to put a less stressful load on the high voltage winding. The larger ripple voltages that result are attenuated to acceptable levels by means of a second order LC ripple filter. Since the ripple choke provides dynamic current limiting, a much larger filter capacitor for ripple attenuation may be used behind it. The larger capacitance also serves to reduce the AC impedance at audio frequencies. The 5U4GB also provides a natural soft start operation that gives the rest of the VTs time to warm up. Since the full DC voltage is never applied to cold tubes, the problem of possible cathode stripping is avoided.
A solid state power supply provides a negative DC rail for bias. Since the current demand isn't so great, and since silicon diodes can work into much larger filter capacitances, a simple active decoupler works well here. The negative rail comes up first at power up, so that no VT operates at full voltage without bias. The negative rail also collapses more slowly at power down to preserve bias voltages as cathodes cool off. This reduces power up stresses on the tubes, and helps to prevent those annoying speaker pops that can sometimes occur with other designs.
With no feedback connected, the sound is surprisingly good, especially considering that the loadline predicts an acceptable, though not exceptional h3= 5.0% (3.0% measured, and close to the 6V6). You would expect the usual "pentode harshness", however, it isn't there. What you mainly notice is excessive "edginess" or "brightness" to the upper mids and highs. You can hear some nastiness with the volume turned nearly to the point of obvious distortion. Almost all the THD is h3, with relatively little higher order harmonics. Otherwise, the most obvious sonic defect is woofer under-damping due to the high rp of the power pentode. Even here, it is less noticeable than expected, and then most obvious with bass-heavy techno. The sound is quite clean and detailed, even with no gNFB at all. The 6BQ6GTB would appear to be every bit as good as the famous 6V6.
With some "spec busting", and biasing the finals hot by increasing the static plate current to 50mA, there is definite sonic improvement. The higher static plate current moves the operation closer to Class A, and reduces the cross-over distortion. This is possible since the 6BQ6GTB was designed to pump max RMS watts into TV deflection coils continuously for hours at a time while preserving a reasonable service life, and therefore the specs are much more conservative than those for audio finals. Audio amplification is a good deal less demanding, so there is no harm in biasing hotter than the rated PD= 12W. This also takes a load off the front end since max power is reached with a lower input voltage swing.
Listening with full gNFB applied reveals that there was a bit more of that pentode harshness present than without it. 12dbv of gNFB really takes the edge off the highs. As expected, bass performance improves with considerable additional detail revealed once the speakers are under control. If there is a problem here, it's that this level of gNFB is definitely tending towards a "solid state" sound. Hard rock and techno both sound "subdued". This might be an acceptable design compromise for a production amp, and might be preferred for listening to softer rock (e.g. Karen Carpenter, Joni Mitchell, Carol King) and Classical.
This level of gNFB seems best for hard rock and techno. There is just enough "edginess" to bring this type of music to life. Woofer damping is still good for tight bass with "authority". No weak bass here that has long been the major complaint against vacuum tube amps. Clarity of detail is excellent, so there isn't enough IMD to become a problem despite the lower level of gNFB. Le Renard remains a clear and detailed amp that doesn't add its own "coloration" to the music. It basically gets out of the way so that the listener can really hear.
It is not necessary to use audiophool expensive tubes and follow trendy fads in circuit topologies to get good sound.
There are several different variations of this basic type. The 6BQ6G types have the larger glass envelope, whereas the GT types use a thinner envelope. Appearance is not the only difference here. The 6BQ6GTA isn't quite so robust as its cousins, having smaller cathodes, and lower ratings. It will red plate if you attempt to use it in Le Renard. It doesn't have the tolerance for operating at above its rated PD. This explains why it was supersceeded by the 6BQ6GTB. As TV screens grew larger, that type was no longer up to driving the deflection coils. Of the larger 6BQ6s, the 6BQ6GA will work in this design. It operates the same as the 6BQ6GTB, biases the same, and will stand up to the hotter biasing used here. As for sonic performance, there is no difference between the 6BQ6GA and the 6BQ6GTB.
Copyright © 2009 All rights reserved