add on ear joy

[Watt? Me worry?]

Hi RAT's!

I spliced in 50H chokes between the B+ PS filter and each OPT.
Followed each with a 22uF PP/oil cap.

No beauty was lost by this latest hatchet job ;)

The sound is stiil OK, too...

Happy Ears!

Al

[Ian Iveson]

Al wrote:

>
> I spliced in 50H chokes between the B+ PS filter and each
> OPT.
> Followed each with a 22uF PP/oil cap.

Goodness. 50H must be quite heavy.

> No beauty was lost by this latest hatchet job ;)

Not to the mains, for sure.

> The sound is stiil OK, too...

Result!

> Happy Ears!

Mine would be happier were it not for the cackling of crows
at the crack of dawn.

My neighbours have been told by their prophet not to waste
food, so they throw their leftover chapattis out for the
vermin, which I doubt is quite what the Great One had in
mind. An overpopulation of malnurished ferral pigeons leaves
dead bodies in the gutters, and hence the crows. And the
bluebottles in summer.

My ears would also be happier if the drunken Slovaks would
stop shouting sometimes.

Grumpy Ears!

Ian

[Bret L]

On Jan 19, 11:51 pm, "Watt? Me worry?" > wrote:
> Hi RAT's!
>
> I spliced in 50H chokes between the B+ PS filter and each OPT.
> Followed each with a 22uF PP/oil cap.
>
> No beauty was lost by this latest hatchet job ;)
>
> The sound is stiil OK, too...

How do you know?

[Bret L]

On Jan 20, 8:04 am, "Ian Iveson" >
wrote:
> Al wrote:
>
> > I spliced in 50H chokes between the B+ PS filter and each
> > OPT.
> > Followed each with a 22uF PP/oil cap.
>
> Goodness. 50H must be quite heavy.
>
> > No beauty was lost by this latest hatchet job ;)
>
> Not to the mains, for sure.
>
> > The sound is stiil OK, too...
>
> Result!
>
> > Happy Ears!
>
> Mine would be happier were it not for the cackling of crows
> at the crack of dawn.
>
> My neighbours have been told by their prophet not to waste
> food, so they throw their leftover chapattis out for the
> vermin, which I doubt is quite what the Great One had in
> mind. An overpopulation of malnurished ferral pigeons leaves
> dead bodies in the gutters, and hence the crows. And the
> bluebottles in summer.
>
You need a hawk, or at the least a good .22

[Patrick Turner]

On Jan 20, 4:51*pm, "Watt? Me worry?" > wrote:
> Hi RAT's!
>
> I spliced in 50H chokes between the B+ PS filter and each OPT.
> Followed each with a 22uF PP/oil cap.
>
> No beauty was lost by this latest hatchet job ;)
>
> The sound is stiil OK, too...
>
> Happy Ears!
>
> Al

So this would mean that the hum at the OPT CT would be probably now be
less than 1mV.
But you'll still see some slow up and down of the Vdc because of the
voltage changes in the levels of the mains.

But the 22uF is a very low value of capacitance and thus a high
impedance anchor between CT and 0V at low F.
So you need to rely on the class A operation lest you get serious
intermodulation if the amp were ever to stray into class AB operation.
But then Quad-II got away with only 16uF, with 17Vrms ripple at the
CT. But they filtered the screen supply well.

Quad 40 is a bit better; the samples I have here for a fix have the
5U4 charging 50uF though 20 ohms R and Vripple at CT is 6.8V. But
there are missing caps in the amps which is usual for chinese made
crap; places marked where they should be but ain't, and the value of C
between CT and 0V should be far higher considering how ****ing
expensive the amps are.
And BTW, one of the the 6SH7 in the Quad 40 is very microphonic.
Probably also chinese. But they sound OK.
0.03% THD at a watt with 8 ohms taken from the 4 ohm outlet.

I'd prefer to have a 470uF between CT and 0V. Nothing wrong with 50H
chokes if you obtained them at the right price, have suitable lifting
methods, have room on the chassis, don't have concerns about the
voltage drop across the winding resistance, and if they don't get hot
from the dc flow.

A choke input filter with a 50H choke and 470uF will work well if you
have tube diodes. Probably the bleed current will be less than the
usual 10% of output dc current, so you don;t have to waste much power.

Patrick Turner.

[Watt? Me worry?]

On Jan 21, 6:37*am, Patrick Turner > wrote:
> On Jan 20, 4:51*pm, "Watt? Me worry?" > wrote:
>
> > Hi RAT's!
>
> > I spliced in 50H chokes between the B+ PS filter and each OPT.
> > Followed each with a 22uF PP/oil cap.
>
> > No beauty was lost by this latest hatchet job ;)
>
> > The sound is stiil OK, too...
>
> > Happy Ears!
>
> > Al
>
> So this would mean that the hum at the OPT CT would be probably now be
> less than 1mV.
> But you'll still see some slow up and down of the Vdc because of the
> voltage changes in the levels of the mains.
>
> But the 22uF is a very low value of capacitance and thus a high
> impedance anchor between CT and 0V at low F.
> So you need to rely on the class A operation lest you get serious
> intermodulation if the amp were ever to stray into class AB operation.
> But then Quad-II got away with only 16uF, with 17Vrms ripple at the
> CT. But they filtered the screen supply well.
>
> Quad 40 is a bit better; the samples I have here for a fix have the
> 5U4 charging 50uF though 20 ohms R and Vripple at CT is 6.8V. But
> there are missing caps in the amps which is usual for chinese made
> crap; places marked where they should be but ain't, and the value of C
> between CT and 0V should be far higher considering how ****ing
> expensive the amps are.
> And BTW, one of the the 6SH7 in the Quad 40 is very microphonic.
> Probably also chinese. But they sound OK.
> 0.03% THD at a watt with 8 ohms taken from the 4 ohm outlet.
>
> I'd prefer to have a 470uF between CT and 0V. Nothing wrong with 50H
> chokes if you obtained them at the right price, have suitable lifting
> methods, have room on the chassis, don't have concerns about the
> voltage drop across the winding resistance, and if they don't get hot
> from the dc flow.
>
> A choke input filter with a 50H choke and 470uF will work well if you
> have tube diodes. Probably the bleed current will be less than the
> usual 10% of output dc current, so you don;t have to waste much power.
>
> Patrick Turner.


Hi RATs,

I have had these chokes so long, I do not remember the price. They are
ElectraPrint 4601. They were not expensive, Jack is never going to
retire, neither. They do drop some voltage - the 6BQ6 as triodes are
now drawing 46mA at 315V. It ain't just the Chinese, I have seen lots
of boards in retail boxes with lots of empty cap positions. Cost
reduction by trial and error, no doubt :)

There was a mfgr in the 1950s that designed TVs that way. They didn't
work well nor last long, but the margins were good. Everybody has some
guiding light. In Nam we learned the light at the end of the tunnel
was just a fungus infection on the butt of the guy in front of you.

This amp is, or once was, a DIY HiFi Supply "Ella". The kit version of
the M100 from Hong Kong.

The 6BQ6-GTB is a nice plug in relacement for the 6CA7, if you cut the
juice about half. It is an old horizontal amplifier. Bit more grunt
that the fabled EL34 ;) You have to run a wire from the top cap to pin
3. NOT for Audiophiles. "What if I put the wire in backwards? Oh, it
could be disasterous!"

The Caps are Obbligato PP/oil in metal cans. 630V. They are huge
compared to the Black Gate 'lytics. Sound exactly the same - to many
recent graduates of Engineering Universities... sigh.

Happy Ears!

Al

[Watt? Me worry?]

Hi RATs!

Each choke drops ~20 volts. Not enough to heat them much, only
slightly warm after about a week of ABBA to Zappa :)

Al

[Patrick Turner]

On Jan 22, 2:50*pm, "Watt? Me worry?" > wrote:
> On Jan 21, 6:37*am, Patrick Turner > wrote:
>
>
>
>
>
> > On Jan 20, 4:51*pm, "Watt? Me worry?" > wrote:
>
> > > Hi RAT's!
>
> > > I spliced in 50H chokes between the B+ PS filter and each OPT.
> > > Followed each with a 22uF PP/oil cap.
>
> > > No beauty was lost by this latest hatchet job ;)
>
> > > The sound is stiil OK, too...
>
> > > Happy Ears!
>
> > > Al
>
> > So this would mean that the hum at the OPT CT would be probably now be
> > less than 1mV.
> > But you'll still see some slow up and down of the Vdc because of the
> > voltage changes in the levels of the mains.
>
> > But the 22uF is a very low value of capacitance and thus a high
> > impedance anchor between CT and 0V at low F.
> > So you need to rely on the class A operation lest you get serious
> > intermodulation if the amp were ever to stray into class AB operation.
> > But then Quad-II got away with only 16uF, with 17Vrms ripple at the
> > CT. But they filtered the screen supply well.
>
> > Quad 40 is a bit better; the samples I have here for a fix have the
> > 5U4 charging 50uF though 20 ohms R and Vripple at CT is 6.8V. But
> > there are missing caps in the amps which is usual for chinese made
> > crap; places marked where they should be but ain't, and the value of C
> > between CT and 0V should be far higher considering how ****ing
> > expensive the amps are.
> > And BTW, one of the the 6SH7 in the Quad 40 is very microphonic.
> > Probably also chinese. But they sound OK.
> > 0.03% THD at a watt with 8 ohms taken from the 4 ohm outlet.
>
> > I'd prefer to have a 470uF between CT and 0V. Nothing wrong with 50H
> > chokes if you obtained them at the right price, have suitable lifting
> > methods, have room on the chassis, don't have concerns about the
> > voltage drop across the winding resistance, and if they don't get hot
> > from the dc flow.
>
> > A choke input filter with a 50H choke and 470uF will work well if you
> > have tube diodes. Probably the bleed current will be less than the
> > usual 10% of output dc current, so you don;t have to waste much power.
>
> > Patrick Turner.
>
> Hi RATs,
>
> I have had these chokes so long, I do not remember the price. They are
> ElectraPrint 4601. They were not expensive, Jack is never going to
> retire, neither.

Does Jack have to fear the dude with hour glass and sythe?

They do drop some voltage - the 6BQ6 as triodes are
> now drawing 46mA at 315V. It ain't just the Chinese, I have seen lots
> of boards in retail boxes with lots of empty cap positions. Cost
> reduction by trial and error, no doubt :)

I've seen many a CD player which seems to lack dozens of parts which
are just not on the board.
There is the local model for the Japanese, then there is the model
they sell to folks who ain't so fussy.....
>
> There was a mfgr in the 1950s that designed TVs that way. They didn't
> work well nor last long, but the margins were good. Everybody has some
> guiding light. In Nam we learned the light at the end of the tunnel
> was just a fungus infection on the butt of the guy in front of you.

I thought the L at EOT was a headlight on a freight train coming
towards you....


> This amp is, or once was, a DIY HiFi Supply "Ella". The kit version of
> the M100 from Hong Kong.
>
> The 6BQ6-GTB is a nice plug in relacement for the 6CA7, if you cut the
> juice about half. It is an old horizontal amplifier. Bit more grunt
> that the fabled EL34 ;) You have to run a wire from the top cap to pin
> 3. NOT for Audiophiles. "What if I put the wire in backwards? Oh, it
> could be disasterous!"

I ain't seen a 6BQ6. But we made 6DQ6/12DQ6 here in Oz which were not
good hi-fi tubes.
But a 6CM5 tetrode in triode mode is a nice sub for the guy who cries
for a week over the cost of a 2A3.

>
> The Caps are Obbligato PP/oil in metal cans. 630V. They are huge
> compared to the Black Gate 'lytics. Sound exactly the same - to many
> recent graduates of Engineering Universities... sigh.

In the app you have, maybe any old cap of 22uF would sound the same.

But last year I rebuilt a guy's speaker crossovers and banished all
the electros and replaced them with Obigato caps.
These are mainly well dressed motor start or motor run caps packaged
for audiophiles which means they have hard wire leads and are sealed
with epoxy potting compount within neatly cut lengths of copper or
brass water pipe. They do look so chic!
Anyway, they also made an inprovement to the sound, and possibly there
may be some improvement in THD/IMD measurements, because the Obligato
caps are presumably polypropylene with high current ability, and are
thus more a real cap than an electro which is a kind of battery. The
currents being handled by the crossover caps is much higher than
filter caps. So where there is only tiny cap currents in filtering or
coupling apps then perhaps what cap one uses makes little sonic
difference. Now I will get dumped on for saying that, eh.

I've got access to unbranded motor start with ratings up to 400Vdc at
least and C values up to 60uF.
I used a few to replace the caps in series with midrange speaker drive
units. I have no idea how they are made except that they are not
bipolar electros. Maybe they are polyester. OK. Thar's still better
than an electro.

In a rail filter situation, methinks the more C you have the better.
If anyone is worried about tiny non linear currents then I suggest ppl
run an amp up to 1dB below clipping and have a look at the waves at
the filter caps. Its NOT alarming.
With a square wave, maybe you see some spikes. OK, then bypass the
electros with plastics, and the HF spikes will dissappear.

Patrick Turner.

>
> Happy Ears!
>
> Al- Hide quoted text -
>
> - Show quoted text -

[Patrick Turner]

On Jan 23, 4:20*pm, "Alex" > wrote:
> "Patrick Turner" > wrote in message
>
> ...
>
> I'd prefer to have a 470uF between CT and 0V. Nothing wrong with 50H
> chokes if you obtained them at the right price, have suitable lifting
> methods, have room on the chassis, don't have concerns about the
> voltage drop across the winding resistance, and if they don't get hot
> from the dc flow.
>
> A choke input filter with a 50H choke and 470uF will work well if you
> have tube diodes. Probably the bleed current will be less than the
> usual 10% of output dc current, so you don;t have to waste much power.
>
> Alex:
>
> Is it healthy for a tube rectifier to charge a large cap on warm-up?

Its never healthy. Tube rectifiers have very low peak current ability
and there are strict rules to follow with regard to the cap value
being charged directly by a tube diode. Eg, in most apps, 47uF is max
for GZ34.

But a 47uF has 34 ohms of reactance at 100Hz, so there is the argument
that you can have a 470uF cap and as long as you have a series 34 ohm
R between GZ34 and the 470uF, then the peak current won't be more than
with a 47 uF cap.

There are strict liminations to the dc flow from the reservoir cap.
One has to be sensible so about 170mA is max for a GZ34. But if one of
a pair of OP tubes goes faulty and draws say 200mA, with the other on
70mA, then you have 270mA, and the GZ34 might arc internally.

The same will happen with 470uF and the series 34 ohm R.

But when an amp is turned off, then on again soon while the OP tubes
are still hot then the Idc is large for a second or two with a 47uF,
and surely will be for longer with a 470uF cap and 34 ohms series R.

If you ask very little from a tube rectifier, they are fine. But the
amps I build might have 450V at 800mA, and I can use a bridge with 8 x
6A rated Si diodes and it just won't fail. I sure can have 470uF, but
I still always have some series R to limit the *peak* charge current
to a low value.

> While cathode has not reach normal temperature its oxide coating resistance
> is large. Current flowing causes power dissipation inside the oxide coating.
> Unfortunately this dissipation is not uniform. If some area of the cathode
> gets a bit hotter than others, electron emission is getting larger from this
> spot. Growing current creates more power dissipation in the oxide and causes
> the local temperature to rise and emission to increase further, causing more
> power generation, etc in an avalanche fasion. This spot is getting hotter
> not by the filament, but by the cathode emission current. The whole cathode
> current tends to concentrate in a small hot spot since its resistance is
> lower than the rest of the cathode.

Well yes, but in practice tube diodes are fairly rugged and withstand
temporary overloads beyond the known SOA.

>
> Needless to say in the hot spot the oxide gets evaporated (until the filter
> cap charges and the "avalanche" dies out). Thus every power-up destroys a
> bit of cathode.

Indeed, this is the case if you had a GZ34 charging 470uF. To avoid
the high peak charging current, use some series R.
>
> Is it correct to say that a rule of thumb is not to exceed about 0.5...1uF
> per each mA of the rectifier rating (at 250V; derate proportionally for
> higher voltages). For 6X4 do not exceed say 35...70uF total (before + after
> a choke), for 5Y3 -- 50...100uF, etc.?

Just follow the maker's data sheets. Nobody used 470uF caps in power
supplies when the data was written.
Quad had GZ32 charging only 16 uF which has XC = 100 ohms at 100Hz.
Now we'd want to use 150uF at least which is only 9 ohms. So to
compensate you'd add a series 100 ohm R.
If you draw the wave forms of the charge current and ripple voltage it
all becomes clear.

With huge C values, the discharge rate from 470uF is very slow, and
Vripple is very low. Therefore the time during which the cap can be
charged tends to become short. However, the time to charge the caps on
every wave crest is lengthened by the tube diode's internal "on"
resistance but its not enough to rely on and peak current is too high.
Hence the need to increase series R with large C.

Of course the best way to get the most out of a tube diode is to have
a choke input filter. The diodes conduct right through the cycle and
peak current is very low. And just to make things interesting one may
add a series RC network across the choke where the C is parallel
resonant with the choke at Fripple. This effectively about doubles the
Z of the choke at Fripple, and thus halves the bleeder current needed.
The tunes choke input supply is a marvel.
C value can be very large.

But Si diodes arestill better.

Patrick Turner.

[Ian Iveson]

Patrick Turner wrote:

> Is it healthy for a tube rectifier to charge a large cap
> on warm-up?

Its never healthy. Tube rectifiers have very low peak
current ability
and there are strict rules to follow with regard to the cap
value
being charged directly by a tube diode. Eg, in most apps,
47uF is max
for GZ34.

But a 47uF has 34 ohms of reactance at 100Hz, so there is
the argument
that you can have a 470uF cap and as long as you have a
series 34 ohm
R between GZ34 and the 470uF, then the peak current won't be
more than
with a 47 uF cap....

***But it's not a very good argument, is it? Peak current
would be considerably greater into your recommended
combination of 470uF and 34 ohms. Impedances don't add like
you're adding them, perhaps, and there are higher mains
harmonics present, too. Consequently your advice could
easily lead to early rectifier failure.

Ian

[Alex]

"Patrick Turner" > wrote in message
...

I'd prefer to have a 470uF between CT and 0V. Nothing wrong with 50H
chokes if you obtained them at the right price, have suitable lifting
methods, have room on the chassis, don't have concerns about the
voltage drop across the winding resistance, and if they don't get hot
from the dc flow.

A choke input filter with a 50H choke and 470uF will work well if you
have tube diodes. Probably the bleed current will be less than the
usual 10% of output dc current, so you don;t have to waste much power.

Alex:

Is it healthy for a tube rectifier to charge a large cap on warm-up?

While cathode has not reach normal temperature its oxide coating resistance
is large. Current flowing causes power dissipation inside the oxide coating.
Unfortunately this dissipation is not uniform. If some area of the cathode
gets a bit hotter than others, electron emission is getting larger from this
spot. Growing current creates more power dissipation in the oxide and causes
the local temperature to rise and emission to increase further, causing more
power generation, etc in an avalanche fasion. This spot is getting hotter
not by the filament, but by the cathode emission current. The whole cathode
current tends to concentrate in a small hot spot since its resistance is
lower than the rest of the cathode.

Needless to say in the hot spot the oxide gets evaporated (until the filter
cap charges and the "avalanche" dies out). Thus every power-up destroys a
bit of cathode.

Is it correct to say that a rule of thumb is not to exceed about 0.5...1uF
per each mA of the rectifier rating (at 250V; derate proportionally for
higher voltages). For 6X4 do not exceed say 35...70uF total (before + after
a choke), for 5Y3 -- 50...100uF, etc.?

[Ian Iveson]

Alex wrote:
>
> "Patrick Turner" > wrote in
> message
> ...
> Its never healthy. Tube rectifiers have very low peak
> current ability
> and there are strict rules to follow with regard to the
> cap value
> being charged directly by a tube diode. Eg, in most apps,
> 47uF is max
> for GZ34.
>
> But a 47uF has 34 ohms of reactance at 100Hz, so there is
> the argument
> that you can have a 470uF cap and as long as you have a
> series 34 ohm
> R between GZ34 and the 470uF, then the peak current won't
> be more than
> with a 47 uF cap.
>
> There are strict liminations to the dc flow from the
> reservoir cap.
> One has to be sensible so about 170mA is max for a GZ34.
> But if one of
> a pair of OP tubes goes faulty and draws say 200mA, with
> the other on
> 70mA, then you have 270mA, and the GZ34 might arc
> internally.
>
> The same will happen with 470uF and the series 34 ohm R.
>
> But when an amp is turned off, then on again soon while
> the OP tubes
> are still hot then the Idc is large for a second or two
> with a 47uF,
> and surely will be for longer with a 470uF cap and 34 ohms
> series R.
>
> Alex:
> My concern was not about a steady state operation (where
> of course one should not try to exceeed maximum values)
> but about power-up. Consider:
>
> 470uF cap charged at 300V holds about 20J of energy. The
> same amount will have to be dissipated in a rectifier to
> charge the cap (given that power transformer has zero
> output impedance and no other series resistors are in
> circuit).
>
> If charging takes 2 seconds, then on average the tube will
> be dissipating 10W during these 2 seconds. 10W for two
> seconds -- is nothing, but only if dissipated on the
> plates of the recifier.
>
> But during power-on, while cathode is cold, a significant
> portion of these 10W dissipates in poorly conducting oxide
> coating, and not even uniformely, but in tiny hot spots.
> That is what I believe is very damaging to the tube's
> cathode, evaporating it.
>
> In steady state operation hot spots are not formed because
> the whole cathode area has low resistance and shares
> emission.
>

A paradigm shift, in this case from our normal
considerations of voltage and current to energy, can seem
like a nonsequiter to the process of understanding.

But in this case our usual calculation methods would be
horrendously complicated because the cap charges through one
impedance and discharges through another.

Although your method demonstrates that Patrick's suggestion
may be unwise, it isn't conclusive because diode datasheets
don't quote their limits in terms of energy. AFAIK they just
give a maximum capacitance. Perhaps it would be safest, in
general, to stick to that?

However, clearly, as resistance before the first cap is
increased, there will come a point when the diode won't care
about the capacitance. Similarly, in a CRC filter, what
value of R is required for the assumption to hold good that
we can forget about the second C?

Simulation offers another way of avoiding awkward
calculations, and will show how your energy flow is
distributed in time. The "PSUD" design software from Duncan
Munro's site is perfect for the job, free, and compact.

http://www.duncanamps.com/software.html

A graph of capacitor current over time after power-up, for
each of Patrick's two cases, can be produced in a jiffy.
They are *very* different. As you imply, settling time is
much greater for Patrick's alternative, and total current
during that time much greater. Simulation should show that
the peaks are considerably higher, too.

Capacitor current, once settled, may be less in Patrick's
alternative circuit, possibly because the resistor
attenuates the higher harmonics more than the smaller
capacitor does.

Ian

[Alex]

"Patrick Turner" > wrote in message
...
Its never healthy. Tube rectifiers have very low peak current ability
and there are strict rules to follow with regard to the cap value
being charged directly by a tube diode. Eg, in most apps, 47uF is max
for GZ34.

But a 47uF has 34 ohms of reactance at 100Hz, so there is the argument
that you can have a 470uF cap and as long as you have a series 34 ohm
R between GZ34 and the 470uF, then the peak current won't be more than
with a 47 uF cap.

There are strict liminations to the dc flow from the reservoir cap.
One has to be sensible so about 170mA is max for a GZ34. But if one of
a pair of OP tubes goes faulty and draws say 200mA, with the other on
70mA, then you have 270mA, and the GZ34 might arc internally.

The same will happen with 470uF and the series 34 ohm R.

But when an amp is turned off, then on again soon while the OP tubes
are still hot then the Idc is large for a second or two with a 47uF,
and surely will be for longer with a 470uF cap and 34 ohms series R.

Alex:
My concern was not about a steady state operation (where of course one
should not try to exceeed maximum values) but about power-up. Consider:

470uF cap charged at 300V holds about 20J of energy. The same amount will
have to be dissipated in a rectifier to charge the cap (given that power
transformer has zero output impedance and no other series resistors are in
circuit).

If charging takes 2 seconds, then on average the tube will be dissipating
10W during these 2 seconds. 10W for two seconds -- is nothing, but only if
dissipated on the plates of the recifier.

But during power-on, while cathode is cold, a significant portion of these
10W dissipates in poorly conducting oxide coating, and not even uniformely,
but in tiny hot spots. That is what I believe is very damaging to the tube's
cathode, evaporating it.

In steady state operation hot spots are not formed because the whole cathode
area has low resistance and shares emission.

Regards,
Alex

[Ian Iveson]

Alex wrote:

>>> My concern was not about a steady state operation (where
>>> of course one should not try to exceeed maximum values)
>>> but about power-up. Consider:
>>>
>>> 470uF cap charged at 300V holds about 20J of energy. The
>>> same amount will have to be dissipated in a rectifier to
>>> charge the cap (given that power transformer has zero
>>> output impedance and no other series resistors are in
>>> circuit).
>>>
>>> If charging takes 2 seconds, then on average the tube
>>> will be dissipating 10W during these 2 seconds. 10W for
>>> two seconds -- is nothing, but only if dissipated on the
>>> plates of the recifier.
>>>
>>> But during power-on, while cathode is cold, a
>>> significant portion of these 10W dissipates in poorly
>>> conducting oxide coating, and not even uniformely, but
>>> in tiny hot spots. That is what I believe is very
>>> damaging to the tube's cathode, evaporating it.
>>>
>>> In steady state operation hot spots are not formed
>>> because the whole cathode area has low resistance and
>>> shares emission.
>>>
>
> Ian:
>> A paradigm shift, in this case from our normal
>> considerations of voltage and current to energy, can seem
>> like a nonsequiter to the process of understanding.
>>
>> But in this case our usual calculation methods would be
>> horrendously complicated because the cap charges through
>> one impedance and discharges through another.
>>
>> Although your method demonstrates that Patrick's
>> suggestion may be unwise, it isn't conclusive because
>> diode datasheets don't quote their limits in terms of
>> energy. AFAIK they just give a maximum capacitance.
>> Perhaps it would be safest, in general, to stick to that?
>>
>> However, clearly, as resistance before the first cap is
>> increased, there will come a point when the diode won't
>> care about the capacitance. Similarly, in a CRC filter,
>> what value of R is required for the assumption to hold
>> good that we can forget about the second C?
>>
>> Simulation offers another way of avoiding awkward
>> calculations, and will show how your energy flow is
>> distributed in time. The "PSUD" design software from
>> Duncan Munro's site is perfect for the job, free, and
>> compact.
>>
>> http://www.duncanamps.com/software.html
>>
>> A graph of capacitor current over time after power-up,
>> for each of Patrick's two cases, can be produced in a
>> jiffy. They are *very* different. As you imply, settling
>> time is much greater for Patrick's alternative, and total
>> current during that time much greater. Simulation should
>> show that the peaks are considerably higher, too.
>>
>> Capacitor current, once settled, may be less in Patrick's
>> alternative circuit, possibly because the resistor
>> attenuates the higher harmonics more than the smaller
>> capacitor does.
>
> Alex:
> I do not argue with the above.
>
> My point was that Patrick's sort of a hint to use a series
> resistor to be able to use very large filter capacitance
> is false and dangerous.
>
> ===quote from Patrick Turner=====
> But a 47uF has 34 ohms of reactance at 100Hz, so there is
> the argument
> that you can have a 470uF cap and as long as you have a
> series 34 ohm
> R between GZ34 and the 470uF, then the peak current won't
> be more than
> with a 47 uF cap.
> ==== wnd quote =======
>
> Having read the above quote one might think that with
> 34ohms he can get away with connecting 470uF, and with a
> larger resistor, say 50ohms, the capacitance can be
> unlimited!!!!!
>
> My point was to warn a reader not to take Partick's advice
> for, though a series resistor will indeed relieve a tube
> in steady state (Partick is right here), nevertheless a
> large cap will be ruining the tube's cathode on EACH
> power-up.
>

Yes. If it hadn't seemed dangerous advice, I wouldn't have
commented. We only differ in the way we explain why.

The chance of me convincing Patrick of anything is zero, and
the chance of him saying so in public is less. A business
run by a "self made man" with little formal education is my
worst-case scenario, communication-wise.

Maybe you'll have more luck, if he hasn't yet dismissed you
as "high-fallutin".

While you're at it, you might have a go at introducing the
concept of vectors, and complex impedance. See if you can
break through the "imaginary numbers" barrier. Good luck!

I have the opposite problem, incidentally, being formally
educated in engineering and philosophy, but with neither
education nor professional experience in electronics. I
thought Patrick should be a good match but he's having none
of it.

AFAIK, there's only four or five ppl left here, so I doubt
we have much traction on the Weltgeist anyway, sigh...

> If you use large caps, please use Si diodes with 10-20ohms
> wirewound in series to limit inrush current to 10A.
> Alternatively if one loves GZ34, design some delay circuit
> which will be shunting say 5kohm series resistor out after
> the cap has settled, leaving only 34ohms in circuit. To do
> make such a circuit one might use relays, thermistors,
> MOSFETs, etc.

Many, including Patrick, use some delay circuit. I use
Morgan Jones', with a 4040 timer and SCR. 34 ohms comes from
nearly nowhere, BTW, because phase wasn't taken into account
in Patrick's arithmetic, and only 100Hz was considered.
Impedances should be summed as vectors, and there are many
higher mains-frequency harmonics present.

> But admit, Si diodes are so much simpler and better.

Simpler, yes, but I don't know about better. Some argue that
amps can sound better with valve rectification, and small
filter caps. AFAIK the rationale is based on how the amp
reacts to, and recovers after, being occasionally
overdriven. I've never heard an audio amp with a valve
rectifier, so I can't argue about how they sound. These days
there is a tendency to build amps big enough so that limits
are rarely reached, so it probably doesn't matter what shape
they are.

Ian

[Alex]

"Ian Iveson" > wrote in message
...
>> Alex:
>> My concern was not about a steady state operation (where of course one
>> should not try to exceeed maximum values) but about power-up. Consider:
>>
>> 470uF cap charged at 300V holds about 20J of energy. The same amount will
>> have to be dissipated in a rectifier to charge the cap (given that power
>> transformer has zero output impedance and no other series resistors are
>> in circuit).
>>
>> If charging takes 2 seconds, then on average the tube will be dissipating
>> 10W during these 2 seconds. 10W for two seconds -- is nothing, but only
>> if dissipated on the plates of the recifier.
>>
>> But during power-on, while cathode is cold, a significant portion of
>> these 10W dissipates in poorly conducting oxide coating, and not even
>> uniformely, but in tiny hot spots. That is what I believe is very
>> damaging to the tube's cathode, evaporating it.
>>
>> In steady state operation hot spots are not formed because the whole
>> cathode area has low resistance and shares emission.
>>

Ian:
> A paradigm shift, in this case from our normal considerations of voltage
> and current to energy, can seem like a nonsequiter to the process of
> understanding.
>
> But in this case our usual calculation methods would be horrendously
> complicated because the cap charges through one impedance and discharges
> through another.
>
> Although your method demonstrates that Patrick's suggestion may be unwise,
> it isn't conclusive because diode datasheets don't quote their limits in
> terms of energy. AFAIK they just give a maximum capacitance. Perhaps it
> would be safest, in general, to stick to that?
>
> However, clearly, as resistance before the first cap is increased, there
> will come a point when the diode won't care about the capacitance.
> Similarly, in a CRC filter, what value of R is required for the assumption
> to hold good that we can forget about the second C?
>
> Simulation offers another way of avoiding awkward calculations, and will
> show how your energy flow is distributed in time. The "PSUD" design
> software from Duncan Munro's site is perfect for the job, free, and
> compact.
>
> http://www.duncanamps.com/software.html
>
> A graph of capacitor current over time after power-up, for each of
> Patrick's two cases, can be produced in a jiffy. They are *very*
> different. As you imply, settling time is much greater for Patrick's
> alternative, and total current during that time much greater. Simulation
> should show that the peaks are considerably higher, too.
>
> Capacitor current, once settled, may be less in Patrick's alternative
> circuit, possibly because the resistor attenuates the higher harmonics
> more than the smaller capacitor does.

Alex:
I do not argue with the above.

My point was that Patrick's sort of a hint to use a series resistor to be
able to use very large filter capacitance is false and dangerous.

===quote from Patrick Turner=====
But a 47uF has 34 ohms of reactance at 100Hz, so there is the argument
that you can have a 470uF cap and as long as you have a series 34 ohm
R between GZ34 and the 470uF, then the peak current won't be more than
with a 47 uF cap.
==== wnd quote =======

Having read the above quote one might think that with 34ohms he can get away
with connecting 470uF, and with a larger resistor, say 50ohms, the
capacitance can be unlimited!!!!!

My point was to warn a reader not to take Partick's advice for, though a
series resistor will indeed relieve a tube in steady state (Partick is right
here), nevertheless a large cap will be ruining the tube's cathode on EACH
power-up.

If you use large caps, please use Si diodes with 10-20ohms wirewound in
series to limit inrush current to 10A. Alternatively if one loves GZ34,
design some delay circuit which will be shunting say 5kohm series resistor
out after the cap has settled, leaving only 34ohms in circuit. To do make
such a circuit one might use relays, thermistors, MOSFETs, etc. But admit,
Si diodes are so much simpler and better.

[Patrick Turner]

On Jan 23, 5:17*am, "Ian Iveson" >
wrote:
> Patrick Turner wrote:
> > Is it healthy for a tube rectifier to charge a large cap
> > on warm-up?
>
> Its never healthy. Tube rectifiers have very low peak
> current ability
> and there are strict rules to follow with regard to the cap
> value
> being charged directly by a tube diode. Eg, in most apps,
> 47uF is max
> for GZ34.
>
> But a 47uF has 34 ohms of reactance at 100Hz, so there is
> the argument
> that you can have a 470uF cap and as long as you have a
> series 34 ohm
> R between GZ34 and the 470uF, then the peak current won't be
> more than
> with a 47 uF cap....
>
> ***But it's not a very good argument, is it? Peak current
> would be considerably greater into your recommended
> combination of 470uF and 34 ohms. Impedances don't add like
> you're adding them, perhaps, and there are higher mains
> harmonics present, too. Consequently your advice could
> easily lead to early rectifier failure.
>
> Ian

And this is why I don't use tube rectifiers AT ALL unless somebody
begs for one.

Adding 34 ohms in series with something which has 3.4 ohms of
reactance means that the real impedance is the sq.rt of the sum of the
squares of R and XR.
So the resultant Z is only a tad above 34 ohms.
With 47uF, XR at 100Hz = 34 ohms and with added 34R the ZCR = 48 ohms.

I still have series R between Si diodes and the large cap sizes I like
to use; most of the time anyway; and these are there to reduce peak
charge currents by about 1/3 during normal consant operation. The R is
chosen to achieve this and no more reduction is needed. The
transformer usually runs a little quieter, and if a fault develops the
R might fuse open well before a diode fuses to a solid short circuit
or before a winding overheats or fuses open. There is a benefit to not
trying to get the absolute maximum B+ voltage theoretically possible,
and there is seldom any need to have a perfectly regulated low
resistance B+ rail.

If you examine the wave forms of the Ripple voltage, with SS diodes at
least the charge rise part of the "saw tooth" ripple wave is often
only 1/4 of the discharge fall time. It means the diode has to pass
enough current to charge the cap in a short time to supply the Idc
draw during the longer discharge time. All one can do is make the
charge and discharge times about equal, and as you approach this the B
+ will fall, so you don't want excessive series R.

I suggest you examine the waves using a full wave GZ34 and 33 ohm R
feeding a 470uF at +400V and with Idc = 150mA,
and let me know what you find.
Try replacing the 33 ohms and 470uF with just 47uF.

What does your simulating program say about this??????

Tube diodes are remarkably rugged, but main thing that causes longer
term excessive peak cathode currents is where a circuit has a faulty
output tube or shorted electrolytic. The faulty OP tube can cause
currents to be high, but often not enough for a fuse to blow. Usually
a fuse will blow if a cap is shorted.

At turn on when the amp is cold, the ramp up time of B+ through a
34ohms to charge a 470uF cap is probably slower than if charging up a
47uF cap directly. But just after turn on and with cathodes only
partially heated, there are some high currents produced to get the B+
upwards, even with a GZ34 feeding a 47uF.

Think about this longer and imagine a 470uF cap at +400V and GZ34
trying to keep it charged with 150mA of draw.
With Si diodes, the Vdc : Vac ratio is about 1.35:1, so 296Vrms is
needed to make 400Vdc.
With a GZ34, the ratio is maybe only 1.2 :1 so 333Vrms is needed. So
peak ac V = 469V.
Where you had 69V between the 400V at the cap and the peak winding
voltage, you have the 69V in series with the diode Ra and the winding
Rw, and if that totalled 69 ohms the peak I would be 1 A. Reasoned
like this nothing makes sense.

The Vripple will be 0.7Vrms. IR = 0.2Arms, but peak I due to non
linear wave form might be 0.5A or more.??

The added 33 ohms is a welcome additional amount of resistance which
does reduce the peak charge I. It also drops the Vdc that you will get
without the added series R.

Anyway, assume for both cases peak acV = 469V, and Vdc = +400V exist
with GZ34 feeding 47uF directly, or with 33 ohms plus 470uF.
What makes you suppose the peak charge currents are going to be
destructively higher in the 470uF + 33 ohm case?

But just build & measure and all will be revealed. Never use fuses
that tend not to blow. Don't ever build any amp without active
protection against excessive Ia in any one or more OP tubes. Always
have a resistance in series with the mains input winding and a delay
circuit to shunt the resistance once the B+ has risen to 2/3 its full
value.

This is my method with SS, but with a tube diode the tube cathode
should have fast heating AND then a longer delay.
This enables a fairly low value mains fuse to be used which should not
give nuisance blows.

In the Quad40 though, there is a thermistor in series with the 5U4 5V
cathode to make it warm slowly. This makes the B+ rise as slowly as it
does with a GZ34. The Quad 40 also has a series R to the C1 = 41uF.
The biggest strain on the 5U4 is when someone turns off the Quad40 and
then turns it back on again after 10 seconds when everything is still
hot. The B+ must be raised up fast without the restricted cathode
ability.

Patrick Turner.

[Patrick Turner]

On Jan 24, 1:56*pm, "Ian Iveson" >
wrote:
> Alex wrote:
> >>> My concern was not about a steady state operation (where
> >>> of course one should not try to exceeed maximum values)
> >>> but about power-up. Consider:
>
> >>> 470uF cap charged at 300V holds about 20J of energy. The
> >>> same amount will have to be dissipated in a rectifier to
> >>> charge the cap (given that power transformer has zero
> >>> output impedance and no other series resistors are in
> >>> circuit).
>
> >>> If charging takes 2 seconds, then on average the tube
> >>> will be dissipating 10W during these 2 seconds. 10W for
> >>> two seconds -- is nothing, but only if dissipated on the
> >>> plates of the recifier.
>
> >>> But during power-on, while cathode is cold, a
> >>> significant portion of these 10W dissipates in poorly
> >>> conducting oxide coating, and not even uniformely, but
> >>> in tiny hot spots. That is what I believe is very
> >>> damaging to the tube's cathode, evaporating it.
>
> >>> In steady state operation hot spots are not formed
> >>> because the whole cathode area has low resistance and
> >>> shares emission.
>
> > Ian:
> >> A paradigm shift, in this case from our normal
> >> considerations of voltage and current to energy, can seem
> >> like a nonsequiter to the process of understanding.
>
> >> But in this case our usual calculation methods would be
> >> horrendously complicated because the cap charges through
> >> one impedance and discharges through another.
>
> >> Although your method demonstrates that Patrick's
> >> suggestion may be unwise, it isn't conclusive because
> >> diode datasheets don't quote their limits in terms of
> >> energy. AFAIK they just give a maximum capacitance.
> >> Perhaps it would be safest, in general, to stick to that?
>
> >> However, clearly, as resistance before the first cap is
> >> increased, there will come a point when the diode won't
> >> care about the capacitance. Similarly, in a CRC filter,
> >> what value of R is required for the assumption to hold
> >> good that we can forget about the second C?
>
> >> Simulation offers another way of avoiding awkward
> >> calculations, and will show how your energy flow is
> >> distributed in time. The "PSUD" design software from
> >> Duncan Munro's site is perfect for the job, free, and
> >> compact.
>
> >>http://www.duncanamps.com/software.html
>
> >> A graph of capacitor current over time after power-up,
> >> for each of Patrick's two cases, can be produced in a
> >> jiffy. They are *very* different. As you imply, settling
> >> time is much greater for Patrick's alternative, and total
> >> current during that time much greater. Simulation should
> >> show that the peaks are considerably higher, too.
>
> >> Capacitor current, once settled, may be less in Patrick's
> >> alternative circuit, possibly because the resistor
> >> attenuates the higher harmonics more than the smaller
> >> capacitor does.
>
> > Alex:
> > I do not argue with the above.
>
> > My point was that Patrick's sort of a hint to use a series
> > resistor to be able to use very large filter capacitance
> > is false and dangerous.
>
> > ===quote from Patrick Turner=====
> > But a 47uF has 34 ohms of reactance at 100Hz, so there is
> > the argument
> > that you can have a 470uF cap and as long as you have a
> > series 34 ohm
> > R between GZ34 and the 470uF, then the peak current won't
> > be more than
> > with a 47 uF cap.
> > ==== wnd quote =======
>
> > Having read the above quote one might think that with
> > 34ohms he can get away with connecting 470uF, and with a
> > larger resistor, say 50ohms, the capacitance can be
> > unlimited!!!!!
>
> > My point was to warn a reader not to take Partick's advice
> > for, though a series resistor will indeed relieve a tube
> > in steady state (Partick is right here), nevertheless a
> > large cap will be ruining the tube's cathode on EACH
> > power-up.
>
> Yes. If it hadn't seemed dangerous advice, I wouldn't have
> commented. We only differ in the way we explain why.
>
> The chance of me convincing Patrick of anything is zero, and
> the chance of him saying so in public is less. A business
> run by a "self made man" with little formal education is my
> worst-case scenario, communication-wise.
>
> Maybe you'll have more luck, if he hasn't yet dismissed you
> as "high-fallutin".
>
> While you're at it, you might have a go at introducing the
> concept of vectors, and complex impedance. See if you can
> break through the "imaginary numbers" barrier. Good luck!
>
> I have the opposite problem, incidentally, being formally
> educated in engineering and philosophy, but with neither
> education nor professional experience in electronics. I
> thought Patrick should be a good match but he's having none
> of it.
>
> AFAIK, there's only four or five ppl left here, so I doubt
> we have much traction on the Weltgeist anyway, sigh...
>
> > If you use large caps, please use Si diodes with 10-20ohms
> > wirewound in series to limit inrush current to 10A.
> > Alternatively if one loves GZ34, design some delay circuit
> > which will be shunting say 5kohm series resistor out after
> > the cap has settled, leaving only 34ohms in circuit. To do
> > make such a circuit one might use relays, thermistors,
> > MOSFETs, etc.
>
> Many, including Patrick, use some delay circuit. I use
> Morgan Jones', with a 4040 timer and SCR. 34 ohms comes from
> nearly nowhere, BTW, because phase wasn't taken into account
> in Patrick's arithmetic, and only 100Hz was considered.
> Impedances should be summed as vectors, and there are many
> higher mains-frequency harmonics present.
>
> > But admit, Si diodes are so much simpler and better.
>
> Simpler, yes, but I don't know about better.

I'd say Si diodes are better, as well as simpler to use.

In my last pair of SE pure class A amps I made for a customer there
was 470V at 350mA for the anode supply.

Probably I'd have needed a quad of GZ34 to handle that, and probably
series R as well to help them pass equal current.
If you don't have them all working with even loading, he who works the
hardest bows out first.

Tube rectifiers will sure sound different in aguitar amp. You strike a
big chord, overload the amp grossly, the OP stage goes into temporary
class C operation to make a square wave, and the B+ sags about 25% and
the poor filtering means you have the signal modulated by the large
ripple voltage. And because of the low value C between OPT CT and 0V
the bass notes modulate the higher F.
The idea of musical heaven is to operate the guitar amp just on the
verge of gross overload to ensure maximum intermodulation between all
F which may be possibly involved, not just the high levels of
fundemental F which initially drown out everything else if the note is
excessively high.
Square waves fed to a speaker have high power sound, and some folks
find it very pleasing ans exciting. The more hamonics the better. The
odd numbered THD due to wave squaring can drown out most of the
harmonics of a string note except of course the even numbers.

Do not musos say the more distortion and grunge in a musicians amp,
the better.

Better? by just whose criteria? I had a "Dark Metal" bandsman as a
client who tortured his amps to bits and he brought a large combo
around with an amp. So he give a demo with his axe and I put on the
ear muffs. It sounded exactly like a procession of 747s were crashing
into my work shed. He preferred the 2 x 6550 in triode mode.

Just what effect a tube rectifier might have on the sound under such
circumstances is open to debate.

But its rare now to find a modern made guitar amp with tube
rectifiers. Si diodes are used and the reservoir C is usually a lot
larger value and the sound is cleaner. The gravelly sound on overdrive
due to ripple voltage modulating the music tone is different in the US
to UK because the ripple frequencies are 120Hz and 100Hz respectively.

Some old muso amps had CLC filtering to the anode supplies. But Fender
and Marshal often left out the choke and C2 on export models. They
could afford to ignore anyone who cmplained in foreign countries.


In hi-fi amps the B+ rails are supposed to be like batteries, and
clean as a whistle, and you never overload anything any time, but the
amp spends 99% of its life producing an average of 10% of maximum
clipping power where that is say 30W.
maybe 3% where PO max = 100W.

Quad-II only had 16uF at the OPT CT for a pair of KT66 inclass A.
The grunge in the signal from reactions between signal and Vripple was
about equal to the THD.
So the grunge was kept low because of the common mode rejection due to
class A working.

If higher C values are used with Si diodes the idle bias in Quad-II
maybe dropped and less class A and more class AB becomes permissable
and no worse sound than when the amps are in pure class A all the
time.

I've used bias Ia = 55mA with mods to power supply with Si diodes and
the amps sound OK. One does not need to use the original 70mA Ia if
you clean up the PSU. Tubes last longer.

Patrick Turner









Some argue that
> amps can sound better with valve rectification, and small
> filter caps. AFAIK the rationale is based on how the amp
> reacts to, and recovers after, being occasionally
> overdriven. I've never heard an audio amp with a valve
> rectifier, so I can't argue about how they sound. These days
> there is a tendency to build amps big enough so that limits
> are rarely reached, so it probably doesn't matter what shape
> they are.
>
> Ian- Hide quoted text -
>
> - Show quoted text -

[Watt? Me worry?]

On Jan 21, 1:23*am, Bret L > wrote:

>
> > The sound is stiil OK, too...
>
> *How do you know?

Poopie said so. He Knows ;)