VAC 7070 problems

I have been working on improvements to a VAC7070 amp for months on and
off
in-between work for other folks.

I have just fitted active protection because the amp here had a nasty
habit
of spitting out 300B all too often, thus incurring horrble tube
replacement costs.

I have made many revisions to the circuit to reduce the horrid levels of
distortion and noise, and the woeful stability, and the terrible way the
tubes were
originally set up by VAC to run at 37 Watts at idle.
Despite the cathode bias on each 300B, some ran cool, and some ran hot,
with some slightly over the 42Watt Pda rating.

The owner got pissed off from tube failures.

So he lugs the all too heavy POS to my workshop,
and I find there's a spare tap on the HT windings, and that reduces Ea,
and then there
cathode resistors were too low a value, so I make them all 1.5K,
and finally have the tubes idling at 27W each and they might survive.

There's less max PO, but who cares? my customer only needs 45W, AB1, not
70W per channel from each quad of 300B.

I also fit RC filtering to the anode supply; VAC didn't have any B+
filtering,
just Si diodes into 4 x 330uF caps series - parallel to make 330uF,
but I add two more 470uF in series per channel, and have CRCRC with R
100 ohms x 40W
abd noise at the CT reduces to 0.04Vrms and this makes huge difference
to audible and measured
N&D even at low levels.

The driver stage is like a Williamson original, but I add a CCS tail to
the balanced amp
and reduce the drive amp N&D a huge amount and so far so good.

The owner bought 10 x Sofia Princess carbon plate 300B and I am doing
final tests
with these.

The problems I found are with stability just below clipping, and heater
bothers.

VAC in their wisdom or stupidity depending how you see things arranged
the
300B heaters so that the heaters on each side of the PP circuit
of one channel ( but not the other ) were connected effectively in
series.
So if you plug in just one 300B, hardly any voltage appears across the
cathode
but 10V appears across the socket grippers in the other empty socket.
I check out the hard to remove PS board where the rectified heater
supplies are,
and there are wrongly connected heater wiring. I rewired it so all
heaters are
truly floating, and not in "floating pairs".
I knew something was wrong when all the tubes were plugged in and I
meausured Rk.
I was getting 750 ohms instead of 1k5, because with tubes in and cross
coupled rectified DC
supplies, the heaters were interconnected. With new matched tubes the
amp was OK,
but once a pair of tubes drifted apart the divorce was ugly and smokey,
and a fire had started at some previous occasion, and burnt some of the
board
and a bodgy repair had been done, and the replaced Rk was a series
joined pair of R
where the solder gone dry, so more smoke..... ahahhhhhh, this story gets
worse and worse.

So the next problem is the way the blommin thing clips.

Its worst without any GNFB applied. There is switchable GNFB from 3dB to
12dB
in 5 positions. Originally, the critical damping done by VAC was a real
joke,
and BW was a worrying and unstable 200kHz because of deliberate
positive FB cross coupling with C from anodes to grids on opposite sides
of the PP circuit.
Kinda over done neutralisation. The slightest C on the output produced
RF oscllations.

Square waves over shot like a ***.

Anyway, I removed the output stage HF positive FB and added a zobel to
V1 anode load and
across the OPT sec and finally the open loop response with no GNFB is
10Hz to 34kHz, and 5Hz to 50kHz with only 6dB of GNFB.
At low levels up to 20W its now an extremely well behaved and good
sounding amp.

Clipping with the 4 good original Sovtek 300B out of the 8 in the amp
when it came to me
seemed to be OK, but with the Sofia Princess 300B,
the darn amp squegs badly about 2dB below clipping
with a 500Hz signal into any load.

Its as if the amp is allergic to working in class AB, and it was
supposed to though,
for how else were the makers ever to squeeze 70 watts from 4 x 300B?

Why such squegging? It gets to a certain nice undistorted output level
and then suddenly
the wave form gets asymetrical and one side of the PP circuit draws grid
current and vertical
lines appear in the distorted wave, and in bursts.

I tried bigger grid 300B stoppers, only 1k were originally fitted, 3k3
made no difference,
I tried R&C zobels across each half of the OPT primary, no difference.

So perhaps its the current cut off behaviour in class AB in the OPT.

I have not witnessed such crappy clipping behaviour before in any amp,
and boy,
I musta closely examined 100 amps in the last 7 years.

So I wondered just what OPT have been fitted to this expensive amp and
I'd say
they'd have to be plain old Hammond. Its very difficult to remove the
cover
over the 4 trannies in the amp; too many frightfully hard to get to
screws.

But peering through a few ventilation holes, all the trannies look like
bell ended Hammonds, and the wires all have what seem to me to be
Hammond
type colour coding.
There are UL taps at 40%, but not used of course,
and there are 4,8, and 16 ohm taps so 4 output terminals, and highest
turn ratio
is 20.8 : 1 and ZR = 432 : 1.

But it seems to me that maybe the half primaries
are not well coupled togther because there is probably not a
large amount of interleaving, which has the effect of effectively
coupling half prim's together.
This amp wasn't very easy to damp critically for unconditional stability
AND get a wide BW into any load, AND get a high amount of PO at F
extremes.

So has anyone else witnessed the bother of squegging
at near clipping in PP amps???

What was your cure to the problem?

I even tried a 0.01uF grid to grid of the output stage to seriously
reduce the HF response of the
driver amp, but it made no difference to squegging at medium F between
100Hz and 1kHz.

I have the other channel to test tomorrow.
After looking and measuring and trying things for 4 hours, I got
nowhere.

All the other nice amps just go up to clipping without any instability.
and then gradually clip to a square wave with a negative bias developed
to both
coupling caps on the driver and a higher Ek on output cathodes. No funny
business.
Is it something about Chinese made 300B?


Patrick Turner.
.

Quantcast