Andre Jute wrote:
>
> Patrick:
>
> This is a superb post that I have saved to reread when I have a moment
> free of the ankle-nippers. Thank you for taking the trouble to share
> all this with us.
>
> Andre Jute
> Grateful
So what are ankle-nippers?
Giant pet fleas? children? fox terriers?
foot operated side cutters for low voltage high current wiring?
!
Patrick Turner.
>
> Patrick Turner wrote:
>
> > Andre Jute wrote:
> > >
> > > On Mar 20, 5:29 pm, Patrick Turner <i...@[EMAIL PROTECTED]
> wrote:
> > >
> > > > I favour the local cathode FB myself, but not with 50% FB as with
> > > > McIntosh or others.
> > > > Drive voltages are thus lower in my amps, and so the drive signals
have
> > > > less distortions
> > > > needing to be lowered by the global NFB.
> > > > so 20% CFB is all one needs.
> > > >
> > > > Quad-II uses 10%, which isn't really enough, but better than none
at
> > > > all.
> > > >
> > > > Parick Turner.
> > >
> > > I don't think it matters precisely how much you use, Patrick, unless
> > > you're stupid and go wild. What matters is that you have thought
about
> > > what you're doing and why you made the choice -- about what it will
> > > offer you in added value either electronically or sonically. Of
> > > course, that's a tautological statement: thinking stops you doing
> > > silly things. A "conservative amp" built by a copycat will never be
as
> > > good as someone who does things because he understands what he is
> > > doing and does whatever he does after deliberate thought.
> >
> > Its almost impossible not to be a copy cat even when you try to avoid
it
> > if you
> > design and build a tube amp. No matter what anyone comes up with,
> > its all been done before somewhere by someone.
> >
> > But the issue is to implement an idea that is chosen in as ideal
> > a way possible, lest you settle to reside besides the majority
> > of mediocre makers of tube gear.
> >
> > I recognised that the Acoustical connection as used in Quad-II
> > was a best way forward to reduce the multiplicity of odd numbered H
> > in a tube output stage and get the Ra effectively very low,
> > and to keep the power at the highest possible level and enjoy
> > a maximum efficiency without sacrificing sonic integrity.
> >
> > While the plain old ordinary simple ultra linear connection is a
decent
> > contraption
> > when lots of class A is used, and indistinguishable to triodes amoung
> > all the ppl I know,
> > it doesn't always suit all tubes, and leaves more to be had
> > in terms of lower Ra, less THD/IMD, and efficiency.
> >
> > >
> > > The QII example is particularly tricky. I once got into immense
> > > trouble on the Joenet for saying I would have used 15-20 per cent:
it
> > > was, according to the mouth-foamers, a heresy against Peter Walker
and
> > > an impertinence from a do-it-yourselfer to have an opinion. We soon
> > > discovered that the fulminators didn't understand that the cathode
> > > feedback must be counted in the total feedback, that it is isn't
> > > something separate.
> >
> > Well, they sure did foam at the mouth.
> >
> > But look, the week after someone invented a beam tetrode,
> > and used loop FB instead of the local FB between anode and grid in a
> > triode,
> > arguments raged as to which sounded better, tetrodes, triodes, or
> > pentodes.
> >
> > Everyone pushed their barrows full of produce, but the thing ppl
really
> > needed to be able to buy
> > after 1933 was good source material, because that was where major
sonic
> > defects abounded.
> >
> > Williamson gave the world his design which was far better than it ever
> > needed to be.
> > The UL then came out, along with the Quad-II, which was patented,
> > and thus locked out of most diyer efforts. Virtually nobody wound
their
> > own OPT, ever,
> > as is the case now. I would be an exception.
> >
> > And I have given greeat thought to it all and measured and listened.
> > I found an ally in Dennis Assenayev, who lurked here up to about 2003,
> > and he taught me about the tiode FB.
> >
> > Nobody works out how many DB of FB is applied in a triode output
stage.
> > Its because the calculation is unecessary, because the triode FB is so
> > intimately
> > applied that artifacts due to this app of FB are virtually non
existant,
> > and the tube becomes able to exist within a global loop of NFB without
> > the penalties
> > of instability, ie, oscilations that cannot be easily cured, and which
> > occur
> > noi matter what you do when placing a 0.22uF across the output.
> >
> > When a NEARLY intimate loop of FB exists such as cathode FB from an
OPT
> > to the cathodes of the output stage, there are parasitic reactances, L
> > and C in shunt or in series
> > with the signal that don't exist in a triode of triode connected beam
> > terode or pentode.
> >
> > So its wise to include the local CFB of a CFB output stage in the
total
> > FB applied.
> >
> > The Wiiliamson has 20 dB of global, and depending on the load, there
is
> > probably
> > an equivalent of 15 db in the triode connection of the KT66.
> > So 35 dB FB alltold.
> > Hence the low Rout for the amp.
> >
> > McIntosh have 50% CFB, and the 6L6 or 6550 run in pure beam mode
because
> > there is no
> > signal voltage between screens and cathode, and no FB voltage either,
as
> > in the case of QuadII
> > where screens are taken to a B+ and there is FB applied to between
> > cathode and g1, and cathode and g2.
> > McIntosh have a very low value RL as the load, and force the output
> > tubes to go a bit imto class AB2
> > to get high PO, and hence the open loop gain of the output tubes is
> > quite low even though they are in
> > beam tetrode. So although 50% of signal between anode and cathode is
at
> > the cathode as FB,
> > the actual amount of FB applied is about only 12dB to 15dB.
> > McIntosh paid special attention to the OPT design and have got bifilar
> > windings
> > of anode and cathode sections to ensure the AB switching artifacts are
> > supressed.
> > Paravicini addopted the GE way of doing it with capacitance shunts
> > across ends of OPT windings with the same ac voltages but at
> > widely different DC voltages. Either way works well.
> > So, effectively, McIntosh are using a similar amount of total NFB as
the
> > Williamson does.
> >
> > QuadII uses considerably LESS total NFB than a W amp IF you say there
is
> > 15dB FB in the triode connection.
> >
> > It would be very difficult to apply an additional 10dB of global FB
> > around a QuadII amp
> > because the OPT quality is very mild and modest indeed.
> >
> > In all my amps I have used 6550 or KT88 which have much higher gm than
> > the KT66 or 6L6,
> > AND I like to have 20% CFB AND I like to take the screens to a fixed
> > screen supply
> > at no more than about 300V above the cathode voltage.
> > So a CFB amp can be arranged for very high power if one wants, and you
> > are not at all worried with the
> > screen dissipation as you are when B+ is high and you try to run un
> > plain UL.
> > UL is limited, CFB isn't, so if anyone wishes to run KT88 with Ea =
> > 600V,
> > Eg2 = 300V, and get about 80 watts, fine, they can do it easily.
> > TT21, the transmitting version of the KT88 with anode top cap is ideal
> > in this audio app
> > and 140 watts is doable in class AB2 without arcing. But who would
want
> > to?
> > Such high PO means little class A PO, and its entirely wrong for real
> > music.
> >
> > So, with 20% of CFB and with tubes having a high value load, so class
A
> > is high,
> > the open loop gain of the output tubes is high compared to amps by
> > McIntosh
> > or the EAR 509 by Paravicini, so the 20% has as good a chance as is
> > needed to linearize
> > the tube and reduce RA without paying the unavoidable costs of rising
> > distortions
> > of the driver amp having to produce up to 150Vrms at each output tube
> > grid.
> >
> > Distortions in most driver stages having to make this much voltage is
4
> > times
> > what it is when making a maximum of 75Vrms which is as high as I ever
> > need to go.
> > Not only that, McIntosh use bootstrapping to fool the 12BH7 driver
> > triodes into thinking that
> > the anode load they have is a dozen time the actual dc RL which is
taken
> > to a winding on the OPT.
> > Bootstrapping like this feds back a small but significant amount of
> > artifacts in the form of
> > positive FB with distortion increasing effects, so McIntosh is a case
> > of 4 steps forward, and one step backward.
> >
> > In all my drive amps used for PP or SE amps I have a choke in series
> > with a resistance
> > to path the DC to the anode. At DC, at least the driver tubes have the
R
> > component,
> > and gain is fair. But at above 50Hz the dc carrying load becomes a
very
> > high impedance,
> > approaching a constant current source in behaviour. The main driver
tube
> > load is therefore
> > the grid biasing resistance of the following output tube/s.
> > Because the load line has effectively been made much closer to
> > horizontal
> > than with a pure R for DC, the distortions of the driver tube is
reduced
> > about at least 10dB, and so 0.3% of 3H at 75Vrms at each PP grid is
> > normal,
> > or 1.4% at 125V to drive an 845 in SET.
> >
> > My latest rendition of SECFB 13EI is to adopt the latest thinking
about
> > the whole matter.
> > Ie, use whatever amount of CFB that is feasable and useful, even if it
> > seems slightly too much,
> > and linearize the driver stage as much as possible, and reduce the
need
> > to rely
> > much on natural 2H cancelations to keep THD/IMD low.
> >
> > About 4 months ago I converted a pair of my single ended SEUL 13EI
amps
> > to CFB.
> > They were giving about 23watts into 8 ohms, with about 19W into 5
ohms,
> > the latter being
> > a bit of a mismatch with twice the THD/IMD as with 8 ohms.
> > Very nice amps though.
> > But I thought it was worth trying to get more without a sonic penalty.
> >
> > The OPT I wound back in 1997 had 3 anode P sections, so the central
one
> > was devoted to the cathode circuit.
> > CFB % is therefore 33%.
> > This CFB winding had taps along its length, including a CT, at 16.5%
ot
> > total P turns.
> >
> > The secondaries were re-arranged to give a load match of 2.8k:5 ohms,
> > and this would work properly
> > if the B+ was raised so that Ea = 480V instead of 380V for UL.
> > The screen was fed a supply and placed at only 200V above the cathode,
> > so screen current was real low,
> > and grid bias needed only to be +30V, instad of 80V with UL.
> > The screen was bypassed well with a cap to the 16.5% tap on the CFB
> > winding.
> >
> > There is a limit to just how you arrange the screen in some multigrids
> > because where there is too
> > much ac between cathode and screen the tube then starts to cut off and
> > reduce its power sort of like the
> > triode connection affects PO in beam tubes.
> >
> > The result of the whole re-arrangement gave me 35 watts of PO at the
> > anode,
> > not bad from a tube with 78 watts of idle dissipation at the anode.
> > With screen diss much reduced from the level found in the UL
connection,
> > the amount gained can be piled
> > onto the anode, and max efficiency rises to 45%, and.
> >
> > The OPT losses are about 10%, so I am now getting 30 watts instead of
19
> > into 5 ohms,
> > and THD/IMD is about 1/4 of what it was, and Rout only 0.3 ohms.
> > The global NFB was reduced from 15dB with the UL connection to only
8dB.
> > Total llop FB with CFB and global = approx 20dB, very similar to
QuadII,
> > but the Rout of this SE amp is 1/3 of the QuadII, and thd/IMD no
worse,
> > and there is 50% more PO.
> >
> > The use of the 33% of CFB in this little SE amp means that of the
> > maximum
> > 300Vrms possible between anode and cathodec, there will be 100Vrms max
> > at the cathode.
> > Open loop gain of the tube is about 15, so 20vrms is needed between g1
> > and cathode top make the 300Vrms.
> > So 120Vrms is needed at the grid.
> > So the gain reduction factor due to NFB in the output stage is about
20
> > / 120 = 1/6,
> > or about 14db of FB, but it isn't as simple as that because there is
> > some NFB applied between
> > screen and cathode, which tends to make the distortion spectra more
> > triode like; a slight but im****tant
> > issue nevertheless, and a bonus.
> >
> > To get the undistorted drive to the 13E1, I used an EL34 in triode
> > bypassed cathode,
> > DC RL = 25k and B+ = +700, so that Ea could be about 300V.
> > Gain is about 8. Input tube of paralleled 6SL7 drives the EL34 easily,
> > Miller effect is low.
> > I had no room to place a choke in series with the dc carrying RL.
> > 2H from EL34 is ths a little high, but then there is more 2H in the
> > output stage
> > becaus its only a darn beam tube to begin with, and they do have high
> > THD,
> > and when its reduced to a sixth of the open loop value by the CFB, it
> > still isn't negligible.
> > But at 5 ohms, there is near maximal natural 2H cancelations from low
to
> > high PO,
> > as the rates of increase of 2H in the driver stage and output stage
are
> > nearly equal.
> > So THD of the amp at 5 ohms resembles a good PP amp in pure class A.
> > This isn't to be sneezed at, and the 10dB reduction in artifacts at
the
> > nominal load
> > without using an extra 10db of NFB translates to terrific music, and
> > excellent conveyance
> > of the foundational nature of good bass in orchestral music.
> >
> >
> > > The same idiots assaulted me when I said that Mr
> > > Williamson's amp was marginally stable with his precise transformer,
> > > and not at all stable with the smallest variation in his transformer
> >
> > Indeed the W was barely stable at LF.
> > But gain step phase ****ft tweaker networks cure the problem entirely.
> >
> > Many ppl found out how bad it could be when they built
> > a W and refused to pay 3 times the normal cost of a crummy OPT with
too
> > little
> > pri inductance. they oscillated at LF without a load connected.
> > With the crummy OPT came very little interleaving, and a hint of C
> > loading at the output
> > made it oscilate at HF, so some samples squegged, producing bursts of
RF
> > at some LF rate, like an RF beacon.
> > Gain step reduction at HF was addopted by Wiilliamson himself, and a
4k7
> > plus 470pF
> > in series was strapped across the 47k RL of V1, 6J5.
> >
> > > -- I have never understood how the Willamson became the orthodoxy.
But
> > > those are two examples where anyone who unthinkingly builds a
chinese
> > > copy may claim conservative design but does the spirit of the
original
> > > inventers Walker and Williamson a serious injustice.
> >
> > Many makers in China mock western knowledge superiority.
> >
> >
> > Tube amp making to make money in China by direct marketing attracts
the
> > real dumbcluck
> > Chinese wannabe entrepreneurs whose other activities are to produce
> > Rorlex watches, and Goochi Handbags,
> > and to provide cheap DVDs of major film releases 5 minutes after
> > Hollywood releases them
> > for distribution.
> >
> > I've had to clean up after the mess they leave behind by rewiring and
> > re-engineerring their horrid contraptions.
> > But most I see ain't copies of the W or anything by Walker.
> > That'd be far too difficult. Usually they are slightly like a Leak,
but
> > with very
> > poorly implemeted and plain wrong alterations.
> > Silk amps based in Thailand is another gross offender.
> > Absolute ****ing KRAP!!!
> >
> > Don't get me started.
> >
> > The Chinese are nowhere near bright enough YET to make good OPT and
amps
> > on their own.
> > Quad amps may be Chinese made, but they are an exception because Quad
is
> > an old British trade name
> > which is worth gold, and ppl pay a huge sum for Quad amps, compared to
> > their cost of
> > Chinese factory production.
> > So there are funds for the careful supervision and quality control.
> >
> > But forget the rest which are gutter brands
> >
> > Perhaps the Chinese may enventually mature like the Japs, but I doubt
> > it, because when Japan matured
> > its electronics and began to maker better audio stuff than made in the
> > UK, US, Germany, Holland etc,
> > tubes were still mainstream, and SS was only at the crawler stage.
> > Digital amps now look set to replace class AB SS generics and they
won't
> > sound any worse.
> > They'll be cheaper, so the gap between SS amp prices and tube amp
prices
> > will widen,
> > placing pressure on Chinese tube amp makers to make them even more
> > cheaply and
> > unfortunately we know what that means, more smoke in the lounge.
> >
> > The first big amp with 4 x EL34 and CFB which I made in 1994 had a
> > Williamson input driver stage.
> > It had the widest possible BW and requires less input than an SE input
> > plus LTP
> > I normally use now. Sound is totally indistinguishable between the
two,
> > IMHO.
> > The wider BW occurs because the CPI acts like a buffer between input
and
> > driver,
> > and has its own lot of local current FB of about 15db.
> > Nobody adds that into the total amount of NDB used.
> > The W balanced amp works better with larger value of common cathode R.
> >
> > I quickly found that if I increased the CFB to more than 20%,
> > there was very little difference in Rout, or THD/IMD, and BW,
> > and no difference to the sound I could detect,
> > so going the whole hog to 50% like McIntosh wasn't needed.
> > And as long as you have enough P and S interleaving, the amp can be
> > stabilised easily.
> >
> > My 300 watters have 20% CFB. The massive OPT have 10 layers of P
> > windings,
> > split into 5 sections of 2 layers each.
> > One such two layer section is for the cathodes, and it is with a CT, 4
> > sections for the anodes.
> > There are 6 sections of S windings consisting of a 72t layer each.
> > Each of the 6 sections is split into 24t and 48t, to give a wide range
> > of Z matching.
> >
> > The effective Ra of each 6550 is thus reduced from its maximum of 18k
> > down to less than 600 ohms.
> > Running the 6550 in pure cathode follower would reduce Ra to 1/gM, ie,
> > about
> > 120 ohms at idle of 40mA per tube.
> >
> > The 20% CFB gives ENOUGH Ra reduction without allowing the output
stage
> > to become too difficult to drive.
> >
> >
> > Lincoln-Walsh made nice PP amps with CFB AND an IST driver tranny.
> > Back in maybe 1955. Its been done before.
> > But I only saw a LW schematic once, and have not seen an electronic
> > copy,
> > so LW might have gone broke or stopped production long ago
> > after not making very much.
> >
> > Anyone have a copy of the Lincoln-Walsh amps?
> >
> > Patrick Turner.
> >
> >
> >
> >
> > >
> > > Andre Jute
> > > Visit Jute on Amps at http://members.lycos.co.uk/fiultra/
> > > "wonderfully well written and reasoned information
> > > for the tube audio constructor"
> > > John Broskie TubeCAD & GlassWare
> > > "an unbelievably comprehensive web site
> > > containing vital gems of wisdom"
> > > Stuart Perry Hi-Fi News & Record Review
|
