In article , John
wrote:
On Thu, 08 Oct 2009 15:33:16 +0100, Jim Lesurf wrote:
I am afraid you are confusing having extended gain response of the
overall design with the design being able to oscillate due to poor
design and loading.
I am aware of things like parasitic oscillations which depending on the
circuit could take place in the leg of a transistor before any low pass
filtering in a later stage for example.
Yes. One of the problems here is that at RF a complex power amp may
resemble the Shackleton aircraft. A set of RF sensitive components flying
in close formation. :-)
[snip discussing causes]
Even that isn't enough to rule out problems like an alteration in
distortion performance short of oscillation.
Well put. I have never considered any of my hi fi gear to be anything
more than what it was designed to be. Certainly not perfection. In fact
In this sphere I am satisfied with much less than perfection. I wouldn't
dream of or imagine that I could run some hi fi in any circumstances.
It's common knowledge that they are multivariate systems with many
parameters which change and are inter dependant.
Alas, the problem for the user or reviewer is that the willingness to
become unstable can depend on factors like the amplifier's inherent output
impedance *inside* the feedback loop. That isn't so easy to characterise
without altering or dissassembling the unit. So checking this can be
detective work.
I have used and tested premium op-amps when designing instrumentation
for medical research and didn't have any difficulty with rf springing
up anywhere along the line. A single high gain block of course is
asking for trouble.
I have spent some years designing and testing audio amplifiers -
commercially, for research, and for personal use. Plus various other
uses of analog amps, signal conditioning, etc, for research and
measurement. So my comments are based on practical experience.
The problem here is partly that domestic hifi amps have to deliver
high powers into undefined loads. The designer has no real idea what
the user will connect to the amplifier - particulary at RF.
Again the audio amplifier should not pass any rf and if it did and it
blew up then the consumer has every right to compensation (my opinion).
The problem here is the meaning of 'pass RF' for various reasons.
If the amp has a large power gain up to 20kHz then that can't vanish
abruptly above that *unless* you have a high order filter somewhere - which
then tends to produce swift changes in phase.
And the unit may have a subsection that can oscillate at RF when the
loadings and drive on other parts is in a given state.
And as I've pointed out, you don't have to explicitly inject RF to make
this happen. They system can do it to itself.
And as for defined loads he should know what is likely to be presented
to his amplifier or for what loads he is designing/selling his
amplifier? It's a cop out to say I'm making an amp but I cant say what
you can do with it. It's a total waste.
You would now need to explain how a designer of commercial domestic audio
power amps would know this. Have a look at a series of reviews of domestic
speakers. Note in particular some of the expensive ones that may have dips
down to an Ohm or so, and phase angles well above 45 deg whilst others go
to high impedances. Note that almost no data is provided for above 20kHz.
Note how much the behaviour varies from one model to the next. Note that
makers keep introducing 'new models' that aren't the same as their old
ones. Then note that a few meters of various unspecified cables will then
completely alter the presented impedance at RF. Out there, anything can
happen... and will.
You can guess what is 'likely'. But if you sell thousands, how many will
find themselves in an 'unlikely' situation over a use life of, say, 10
years. Some of the amps I've been involved with designing are still in use
after more than three times that period. Many have changed hands more than
once. During that lifetime some of the component values will have altered.
And if you assume, say, any phase angle which allows Real to be positive
and modulus bigger than 0.5 Ohms up to 30MHz then the design may be rather
more costly than using weaker assumptions. Perhaps enough so something else
is bought by the customers. So unfortunate if the unpublished unknown
actual results for most speakers were nothing like that demanding at RF.
So designers "should" do all kinds of things. But it might help them if
users realised some of the above. And if the data to base their decisions
upon were actually available. :-)
I know of one case where the user decided the (DIN) speaker socket was the
mains input socket. That even took the track of the board! =8-]
When I designed I ended up assuming that any impedance need to be 'safe' in
the sense of never doing anything worse than blowing a fuse, and the amp
had to be stable unconditionally. But being able to drive low impedances
(and high phase angles) meant I had to ensure the amp could drive very high
currents and powers. So +/- 60V into 2 Ohms, both channels sustained. No
problem - except when the test loads started to melt the benchtop. 8-]
Does put up the cost though, and may be a waste for most customers with
more reasonable speakers.
I think you will know the first rule of purchasing computers, - know
what you want it for then get one that will do the job. That was the
number one maxim before the pc became mainstream.
The problem with analogies to to determine if they are actually relevant or
not. :-)
The problem here is that in terms of your analogy, most sellers of
computers won't actually tell you what you need to know about what the
model can or cannot do. They may not even know. You just are expected to
buy the one that looks 'nicest' in the shop or magazine article and perhaps
the cheapest you think you can get away with using. Good luck. :-)
The situation is often very different in laboratory or
industrial/commercial cases. There you can spec the requirements and
sellers can be expected to give relevant specs for their products. But in
domestic audio most of that data is absent and people just assume it will
either work, or not, on the basis of some 'subjective comments' in a
magazine. Yer pays yer money and yer takes yer chance.
That was one of the main reasons I did the investigations that started this
thread. To raise awareness. I don't doubt that *most* amplifiers in *most*
cases will be fine. But where is the data to tell how many, in which cases,
etc?
I'm always moaning to my wife after a shopping trip how most things are
not properly designed.
I'm still baffled that so many hoovers have pipes that simply pull apart
when you pull the brush towards you. I can understand why they all seem to
be designed for midgets as I fear the designers presume the user will be
female, and female means small. But that becomes doubly daft as their
assumption when you find out how heavy many of the cleaners are. The target
user is presumable a midget strongman, preumably also a magician who can
magic the tubes to stay together. How many cleaners do they sell to
circuses, I wonder... :-)
Slainte,
Jim
--
Please use the address on the audiomisc page if you wish to email me.
Electronics
http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio
http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc
http://www.audiomisc.co.uk/index.html
New webpage on loudspeaker cables
Linear Mode
