| [ |
mood |
| |
 accomplished |
] |
I have finally been able to complete this latest project: a 36W monoblock that is based completely on cheap and readily available "Tee-Vee t00bz". The finals are Class AB1 6BQ6GTB horizontal deflection power amps designed for the wide scan, B&W CRTs that began appearing in the mid-1950s. Though rated like a 6V6 (PD= 12W for both types) the 6BQ6GTB can output a good deal more power without spec-busting. This is made possible since the 6BQ6GTB was designed to source big currents at low VPKs -- just what you want to move big currents through the horizontal deflection coils. Not only does it have a peak current way higher than a 6V6, it also has a much lower RL.
6BQ6GTB Loadline
So how do these sound? Based on listening experience, I'd say like a 6V6 on steroids. Even running open loop, with no negative feedback at all, there is very little of that pentode harshness in evidence. Except for some of the most difficult program material at the highest volumes, the nastiness is mainly audible by its abscence when you add the gNFB. Given how deep into Class AB, and the estimated h3, these sound surprisingly good. So good, in fact, that I dropped plans to include a local feedback loop.
Here are some open loop stats:
PO= 36W (8R) at onset of clipping
AVol= 55.56 (55 Calc)
VPP= 352V (350V Design)
VSGSG= 150V (150V Design)
VKK= -275V (-300V Design)
fh= 35.7KHz (-3.0dbv)
fl= 16.6Hz
h3= ~3.0% (5.0% Est)
Front End:
fh= 117.7KHz
fl= ~7.0Hz
12dbv gNFB:
fh= 50KHz
fl= 11.0Hz
AVcl= 13.6 (8R)
Front End
The front end is a cascoded differential based on the 6BQ7A: another TV tube, this one designed for use as a cascoded VHF, small signal amp. The 6BQ7A has the advantage of having a highish gm that doesn't collapse too fast with decreasing plate current -- desireable for a cascode. Though it is not such a great audio tube that you'd want to use it in more "conventional" audio small signal amps, the linearity isn't all that bad. The trick here is to get the RP up, and to draw a decent amount of current. This, of course, requires a high DC rail. Even if the THD estimate from the loadline analysis doesn't look all that great, a differential cancels the even order harmonics. The cascode connection also seems to improve linearity.
There isn't very much information concerning hollow state cascodes for audio -- mainly sites devoted to guitar amps. Of course, fidelity isn't such a concern for this particular service. It was a case of "let's try and hope for the best". Of course, the topology is used regularly with solid state designs, so why not for hollow state as well? It works just great there too. If your design doesn't need to swing a big output (the cascode's main limitation) then there is no reason that this shouldn't be employed more often. It beats hell out of all the paraphase-type splitters, gives more gain than a triode-based differential, and certainly more than a cathodyne. By using this design, I was able to ditch a second gain stage "up front".
The cascode phase splitter is capacitor coupled to cathode follower grid drivers made from the 6FQ7 dual triode. This is basically a 6SN7 in a nine-pin mini bottle. These grid drivers serve two purposes. First, this allows DC coupling to the grids of the finals. This eliminates the "blocking" problem when the inevitabe transient comes along that drives the grids positive. If a coupling capacitor is used here, turning on the grid-cathode parasitic diode will charge that capacitor negative. That will put the finals into a less linear part of the characteristic until that excess charge leaks off. (This is how Class C RF amps derive most, if not all, of their operating grid bias.) Even if you don't actually hear obvious distortion, the blocking will cause noticeable sonic degradation. Since there is no capacitor to accumulate a charge, the finals can slip transparantly a bit into Class AB2, as the driver can source grid current. Transients simply are not heard.
The drivers also serve to source current into the finals' combined Ci + CMiller + Cstray. At the higher frequencies, the grids don't see the true signal level until these capacitances are fully charged. Charging these quickly requires current. This is the slew rate problem, and it doesn't just occur with solid state designs. There is no way that a high gain, small signal triode pulling sub-milliamp currents can adequately drive the grids of most finals. 12AX7As connected to 6V6s is a bad design. Yet you see that all the time. YUCK!
The 6BQ6GTB finals are operated in straight pentode mode. Given the very low screen voltage limitation, both Ultralinear and Pseudotriode operations are not practical prospects. If you are going to get the least distortion, then you need to nail down the screen voltage, and operate them into the lowest possible impedance. A series dropping resistor certainly is not up to the task, and neither is a voltage divider. The screens are supplied from an active regulator. This consists of a 6AQ5 trioded as the series pass tube, a 6KE8 triode/pentode error amp, and a 0A3/VR75 gas discharge tube for the reference voltage. This being adjustable so that the screen voltage may be set correctly. This regulator helps to attenuate residual ripple, maintain a steady voltage under varying currents, and to provide a Lo-Z AC return. Active regulation is seldom ever seen in audio designs, though rather common in RF designs. You will never see it in commercial designs at all.
So how does this collection of Tee-Vee T00bz sound?
Actually, very good indeed. Even running open loop, the main sonic defect is sloppy bass due to underdamping of the woofer. The highs are a bit "edgy", but there is none of that irritating pentode harshness that quickly wears you out. Not even with the most difficult program material, unless you play it too loud anyway. This meets design criteria number one: make the best open loop implementation before adding any gNFB. This is opposed to what seems to happen with all too much regularity: skimp on the open loop design and sweep those defects -- along with most of the musical detail -- under the rug with excessive amounts of gNFB.
When dialing in the full 12dbv of gNFB, that pentode harshness is noticed by its abscence. The background becomes quite clear. Bass stands out in fine detail, now that the woofers are under control. However, compared to open loop listening, music sounds a bit "subdued". For a production amp, this would be OK. For some types of music, it might even be preferable. Doesn't do Ozzy or Nirvana justice, however. Cutting the gNFB back to 6dbv or so is much better. That pentode edginess is still tamed, especially in the highs. Bass still packs plenty of authority without the sloppiness, and the apparant sound stage is expanded. The reduced gNFB really lets the music "come alive". Even though adding a feedback control means another knob to twiddle, it is a good idea after all.
I'd stack this amp up against anything made with "audiophool" VTs. Can a $1.30 TV toob stack up against a 6V6/6L6/KT66/KT88 &c that cost $One helluvalot more? You bet it can!
Main Schemo
Screen Regulator
|
geeky
calm
