[Andy Evans]

Eeyore wrote:
That video on YouTube got me thinking about them and why they might have
different characteristics to indirected heated triodes.

Clearly the heater/cathode has a voltage gradient across it. Turn the idea
'upside down' and it's equivalent to a grid with a non-equipotential effect over
its length. The most similar effect I can think of would be a variable mu valve
where the grid spacing varies.

Anyway, it occurs to me that a DHT will have a characterisitic similar to the
sum ( or composite ) of several triodes with displaced grid potentials. This may
possibly make it more linear. I haven't looked into it that far yet but I'm sure
it could be readily modelled.

Graham

I'm glad you are taking an interest in DHTs, Graham. You should try a
build - let me know if you want to try any, I have 100s. Can't help
with the theory - I have no idea how they're built nor why they sound
better but I can give you the results of a years listening tests with
four different DHT users if you want to know what they sound like. One
thing - the topology and the filament supply are absolutely crucial to
the best sound, though they'll work OK with clean DC off bench power
supplies. Andy

[Chris Hornbeck]

On Thu, 21 Dec 2006 01:49:30 -0500, "Arny Krueger"
wrote:

Ironically, preamps built with 12AX7s have never come close to providing the
low distortion operation that we take for granted with their solid state
predecessors.

This, and lots of the rest of the post, pass beyond
being merely wrong.

Physicist: "But, surely, Pauli, you don't think what I've said is
completely wrong?"

Wolfgang Pauli: "No, I think what you said is not even wrong."

Religion doesn't belong on a technical newsgroup.
Thanks anyway,

Chris Hornbeck
"Too soon oldt; too late schmardt."

[Chris Hornbeck]

On 21 Dec 2006 03:41:30 -0800, "Peter Wieck" wrote:

Surely you're not saying that tolerances for semiconductors
are lower than tolerances for electron valves? That would
be wacky.

Not at all. Semi tolerances are much tighter than those for tubes
(Parameters wide enough for the QE2 at speed does not suggest tight
tolerances). But also of an entirely different nature. Best analogy
that comes to mind is that a tube does function similarly to a valve, a
transistor similarly to a switch.

Similarly....

Not Exactly.

I still have no idea what you're saying.

Vacuum tubes were routinely built in large quantities with
production tolerances in the 10% range for the three big
numbers.

Discrete semiconductors are made with such large tolerances
that equipment manufacturers buy them in color groups with
2:1 (or sometimes more!) ranges, pre-selected by the device
manufacturer.

This is a large driving force in the trend towards
integrated circuits, which take advantage of device
matching on a die and design around device absolute
spec tolerances which are terrible by the standards
of much physically larger devices like electron valves.

Much thanks, as always,

Chris Hornbeck
"Too soon oldt; too late schmardt."

[Eeyore]

Chris Hornbeck wrote:

Discrete semiconductors are made with such large tolerances
that equipment manufacturers buy them in color groups with
2:1 (or sometimes more!) ranges, pre-selected by the device
manufacturer.

The only parameter with that much variation is hfe. I can't recall seing a group
with 2:1 range btw.

The simple fact is that hfe isn't *that* important usually anyway.


This is a large driving force in the trend towards
integrated circuits

Trend ? Discrete semiconductor design is almost exclusively the province of
specialists now !

Graham

[Chris Hornbeck]

On Fri, 22 Dec 2006 02:40:50 +0000, Eeyore
wrote:

Discrete semiconductors are made with such large tolerances
that equipment manufacturers buy them in color groups with
2:1 (or sometimes more!) ranges, pre-selected by the device
manufacturer.

The only parameter with that much variation is hfe. I can't recall seing a group
with 2:1 range btw.

And I've never even heard of anyone offering *any* active
semiconductor with anything remotely approaching 10%
tolerances, other than by special after-production
selection.

The only semiconductors with transfer exponents low enough
to be remotely called "linear", junction (or, in special, narrowly-
defined high-bias cases, MOS) FET's are supplied in color
groups with huge variations, selected in batches after
production. 2:1 at least. Everyone check it for themselves.

This is only raised as an issue because of the astounding
level of serious bull**** on the newsgroup lately. Please
don't contribute to the noise level; things are difficult
enough.


The simple fact is that hfe isn't *that* important usually anyway.

It *can't* be, because it *can't* be specified. Modern
semiconductor amplifier designs *must* design _around_
device specification, 'cause it ain't happenin.

No big thang; good work can be done other ways. But
let's all not introduce bogus arguments about our
favorite Gods.

This is a large driving force in the trend towards
integrated circuits

Trend ? Discrete semiconductor design is almost exclusively the province of
specialists now !

Now?

Much thanks, as always,

Chris Hornbeck
"This life has been a test. If this had been an actual life,
you would have received instructions on where to go and what to do."
- Angela Chase (Claire Danes)

[Chris Hornbeck]

On Fri, 22 Dec 2006 03:17:18 GMT, Chris Hornbeck
wrote:

Actually, I take it all back. Had I paid due diligence
to the thread's title, I wouldn't have responded at all.

rec.audio.tubes belongs to all of us.

Chris Hornbeck
"This life has been a test. If this had been an actual life,
you would have received instructions on where to go and what to do."
- Angela Chase (Claire Danes)

[Chris Hornbeck]

On Thu, 21 Dec 2006 12:50:37 GMT, Patrick Turner
wrote:

But without the local emitter R current NFB the bjt is very non linear
in its voltage amplifying
ability even though the Ib to Ic ratio is a substantially linear
character for any bjt.



Let us see who is asleep or not!!!

Assuming you mean the above paragraph. Ho, ho, ho.

Much thanks, as always,

Chris Hornbeck
"This life has been a test. If this had been an actual life,
you would have received instructions on where to go and what to do."
- Angela Chase (Claire Danes)

[Arny Krueger]

"Chris Hornbeck" wrote in
message
On Thu, 21 Dec 2006 01:49:30 -0500, "Arny Krueger"
wrote:

Ironically, preamps built with 12AX7s have never come
close to providing the low distortion operation that we
take for granted with their solid state predecessors.

This, and lots of the rest of the post, pass beyond
being merely wrong.

Try to be fair Chris, no matter what your biases are.

The OP was well beyond being merely wrong, containing factually wrong, not
to mention irrelevant comments like:

"inspect the sand at the base. That's the stuff amplification devices are
made of, tubes and transistors both"

The major irrelevancy is that the materials that amplification devices are
made out of is moot. How they work is far,far more relevant.

The factual error is the statement that tubes (and transistors) are made out
of sand. Tubes are routinely made that contain no sand at all. Tubes and
transistors don't work unless they contain other metals than just silicon.

My comment was more relevant and factual than much if not all of Jute's OP,
yet you seem to take no exceptions to it.

[Patrick Turner]

Chris Hornbeck wrote:

On Thu, 21 Dec 2006 12:50:37 GMT, Patrick Turner
wrote:

But without the local emitter R current NFB the bjt is very non linear
in its voltage amplifying
ability even though the Ib to Ic ratio is a substantially linear
character for any bjt.

Let us see who is asleep or not!!!

Assuming you mean the above paragraph. Ho, ho, ho.

hee hee hee, Yea, Ah doo mean thaat laarst parrograffy...

The collector current is substantially linear with base current, a nice
but often
useless charactersistic of a bjt. The fact there is ANY base current is
an irritation to us all, and forces
us to use darlington pairs, or darlington triples, like in Crown amps.

But just considering a lone iddy biddy signal bjt, we find they have
about 30 times the THD/IMD
that is produced in a garden variety triode for the same 5Vrms of
output.
This is because of the non linear turn on behaviour of the PN junction.
And maybe you have noticed, but if you set up a little bjt in a common
emitter gain stage,
and EVERYONE should do this so they don't remain ignorant which is the
case if they don't,
then if you supply the base with a signal from a current source rather
than a voltage source,
( and shock horror, ppl will be FORCED to distinguish between a voltage
and current source to
prevent total ignoramitis from ****ing their brain completely )
then the voltage output linearity compared to voltage input at the
current source is a lot better.
But nearly all discrete bjt stages have an unbypassed emitter resistor
to give oodles of local current FB.
It also has the drmatic and relieving effect of raising the otherwise
utterly attrocious
LOW input impedance of the base circuit of the bjt with a solidly
grounded emitter.
Also lucky for us is that the bjt has oodles MORE transconductance than
any other devices.
The bjt needs an Eb to Ee change of from about 0.4V to 0.8V to go from
mA change to perhaps 30A change,
and all bjts operate to produce all their collector current change with
this small Eb-e change.
Its such a non linear relationship that engineers are taken out and shot
if they refer to bjts having Gm, and an amplification factor µ, although
collector resistance, Rc, is
recognised, but hell, the Gm and Rc are wildly different depending on
the Ic that flows.

But Gm, Rc, and µ can all be worked out for a bjt, they most certainly
can and should be measured
when farnarcling around with the devices.
They are of interest if we determine the quiescent Ic conditions and
consider a very
small current change in the bjt. This can be quite handy where we have a
Voltage Amplifier Stage
in a power amp where the main voltage gain of the amp is determined by
the VAS between the input differential pair, IDP, and output
transconductance stage which
unlike the VAS stage puts amps of current flow into the voltage signal.
Its not uncommon for a signal bjt to have Gm = 0.15A/V at only 5mA of
collector current,
so using a 2k load produces a gain of about 300, a huge amount more than
if we hooked up
a 6AU6 to the 2k load, because its Gm is only about 0.004A/V at 5mA of
Ia.

So with an open loop gain of 300, which may be piles more than we want
if its a line stage we are making,
then we have the opportunity to reduce it with local current FB to make
closed loop gain = 10,
by placing an emitter resistor approximately 2k / 10 or about 180 ohms,
( a standard value ).
For 2Vrms of collector output, there is 1mA of load current, and
0.18Vrms at the 180 ohms.
The base-emitter voltage will be 2/300 = approx 0.0066Vrms, so Vb input
is 0.18 + 0.0066 = 0.1866Vrms,
and closed loop gain is slightly more than 10, and the gain reduction
factor due to NFB = 10/300 = 0.033,
which is roughly -30dB.
So if THD was an appalling 2% at 2Vrms out putput with no FB then with
FB
it will be 2/30 % = 0.066%, and closer to what a triode without loop
current or loop FB will provide.
Input resistance is raised by the inverse of the gain reduction factor,
ie, in this case 30 times,
so that if Rb in with grounded emitter was say 1kohm, its 30k with CFB.
But this is still appallingly low by vacuum tube standards, ( or J-fet
standards )
and so a darlington pair make a sensible solution with the Rb input
raised in this case by a factor equal to the hfe of the driving input
bjts of the pair,
which acts like an emitter follower on the base of the following bjt.
If the hfe is say 100, not uncommon, then Rb in would be 3,000,000 ohms
by rough calculation,
but only a measurement will confirm exactly what you'll get when you use
these lil varmints
because of the variations in device parameters; bjts with the same data
can be wildly unmatched,
with samples varying in a range about equal to the differences between
say a 12AY7 and 12AX7.

So one must have a flexible mind with solid state productions, and
realise that
although trying to find matched devices for each of 2 channels will help
the gains of the 2 channels to be equal,
what has by far the most dominating effect is the amount of NFB and
resistor accuracy, something
we don't have to worry about these days because resistors within 1% are
easily found, and they stay
within this tolerance over time and with varying temperature which btw
is another darn thing that has a big effect
on the gm of a bjt.
But where we see a load of 2k2 and Re of 220 ohms, we can say the stage
gain will be about 10;
it won't be 100, or 3, it'll be about 10, its that simple, and
regardless of variations in gm hfe etc.


Now I must do a little more work...

Patrick Turner.




Much thanks, as always,

Chris Hornbeck
"This life has been a test. If this had been an actual life,
you would have received instructions on where to go and what to do."
- Angela Chase (Claire Danes)