3 Nov
2002
3 Nov
'02
12:42 p.m.
On Sat, Nov 02, 2002 at 08:12:10 +0100, Tim Goetze wrote:
Did you try adding the 0th harmionic to the front of the table, and
dropping the last harmonic? That made it sound pretty good for low notes.
In the
instantaite block fixes it up more-or-less. Maybe even adds a bit
of compression (it boosts the gain to make it roughly 1:1 too).
have you modified the lut in the meantime? i don't seem to be getting
the right results with the log scaling you use. ah, i'm afraid that's only half the truth. try
the link i posted
to the list; somebody has been doing this before us and published.
i still haven't checked for frequency dependency of valve distortion,
but it seems they have, and found it to exist.
Figures, on both counts.
additionally, it seems that to tackle intermodulation
distortion it
is best to split the incoming signal to multiple bands (those great
theoreticians have been using FIRs in matlab) and shape each band
separately. they do in fact use two blended shapers per band.
OK, why do they use two shapers? Or is it one cheby and one non polynomial?
btw, the 'lower frequency' phenomenon occurs
with the real amp, too,
only it's not as pronounced.
OK, I didn't know that.
drat. i fear that we'll need to cool this box into
superconductance
to get the effect done *really* right.
Yeah, I'm thinking we're not tackling it right, if behringer can knock out
a virtual amp harware box for E150 it can't the that cycle hungry.
- Steve
3 Nov
3 Nov
3:11 p.m.
Steve Harris wrote:
two chebyshevs, blend factor depending on incoming amplitude.
Did you try adding the 0th harmionic to the front of
the table, and
dropping the last harmonic? That made it sound pretty good for low notes.
i still think the harmonic mix doesn't sound right, and it's
becoming obvious why.
additionally,
it seems that to tackle intermodulation distortion it
is best to split the incoming signal to multiple bands (those great
theoreticians have been using FIRs in matlab) and shape each band
separately. they do in fact use two blended shapers per band.
OK, why do they use two shapers? Or is it one cheby and one non polynomial? Yeah, I'm thinking we're not tackling it right,
if behringer can knock out
a virtual amp harware box for E150 it can't the that cycle hungry.
funny; now that i'm losing some skepticism about the approach
you seem to pick it up.
tim
7:14 p.m.
On Sun, Nov 03, 2002 at 07:48:55 +0100, Tim Goetze wrote:
Thats interesting, cos liiing at your waterfall plot it doesn;t seem liek
that should be enough.
I odnt think that Line 6 and co. are using chebys, I think there using
something more like our first attempt. That's not based on any knowledge
or anything, just a hunch.
- Steve
OK, why do they
use two shapers? Or is it one cheby and one non polynomial?
two chebyshevs, blend factor depending on incoming amplitude. Yeah, I'm
thinking we're not tackling it right, if behringer can knock out
a virtual amp harware box for E150 it can't the that cycle hungry.
funny; now that i'm losing some skepticism about the approach
you seem to pick it up.
4 Nov
4 Nov
9:02 a.m.
Steve Harris wrote:
On Sun, Nov 03, 2002 at 07:48:55 +0100, Tim Goetze
wrote:
Thats interesting, cos liiing at your waterfall plot it doesn;t seem liek
that should be enough.
afai understand the paper, they've been doing this as an academic
exercise. ivory towers and so on ...
OK, why do
they use two shapers? Or is it one cheby and one non polynomial?
two chebyshevs, blend factor depending on incoming amplitude. I odnt think that Line 6 and co. are using chebys, I
think there using
something more like our first attempt. That's not based on any knowledge
or anything, just a hunch.
yep, it seems unlikely the more i think about it. the peak
problem when lots of strong harmonics are generated could be
tackled with a phase-shifting filter. but we'd still need to split
the input into multiple bands to come up with purer sines (i've
done some more experiments with two mixed sines and a chebyshev
shaper: the output is badly garbled, with unpredictable peaks).
but pod distortion waveforms don't look like they use chebyshev
polynomials at all.
tim
10:06 a.m.
On Mon, Nov 04, 2002 at 01:54:56 +0100, Tim Goetze wrote:
[sorry for the bad-even-for-me typing it was very cold in my studio]
Well, IMHO the point of ivory towers is that you dont have to be connected
to reality, so you'd think they would just go for the most accurate
process. NB I haven't read the paper.
Thats
interesting, cos liiing at your waterfall plot it doesn;t seem liek
that should be enough.
afai understand the paper, they've been doing this as an academic
exercise. ivory towers and so on ... but pod distortion waveforms don't look like they
use chebyshev
polynomials at all.
I agree. I wonder if they've just worked out transfer functions for
different envelope amplitudes...
You could feed in sinewaves (or maybe saws, I'd have to think about the
consequences of that though), and construct a transfer function table by
mapping input amplitudes to output - you'd need to correct for the delay
caused by the filters in the circuit, but otherwise it should be fairly
easy.
The highpass filter might make things interesting, but you could probably
correct for it by averaging the output from the upcurve of the sine with
the output at the downcurve.
It still wouldn't be frequency dependent, but its another possible route.
It will be much cheaper than our other approaches, probably cheap enough
that you could carve up the input signal into frequency bands and process
them seperatly without it being slow... which would also make the alising
easier to deal with.
- Steve, feeling entusiastic again :)
11:45 a.m.
Steve Harris wrote:
Well, IMHO the point of ivory towers is that you dont
have to be connected
to reality, so you'd think they would just go for the most accurate
process. NB I haven't read the paper.
maybe the thought of amplitude-dependent harmonic relations
didn't occur to them, i don't know.
there should be a link to the pdf at the top of the page on
citeseer, it's a short but interesting read.
there's another interesting paper from DAFx '99: "discrete-time
models for nonlinear audio systems", you can fetch it somewhere
from http://echo.gaps.ssr.upm.es/COSTG6 iirc.
those guys are using a different approach; they seem to have
found a way to capture the nonlinearities of a system without
having to care how the system works at all. looks very cpu-
bound to me though.
tim
10:46 a.m.
The sine -> waveshaper mapping I talked about in the previous followup to
this mail isn't going to work (I think) because of the huge asymetry of
the output waveform, after looking at line-sine-fade Tim mailed.
It might still be possible with saws as the input, but then you have to
worry about phase shift was well :(
Dammit.
I think its probably worth a try with saws, at least it will make
extracting the transfer function simpler. I will try playing with
extracting a transfer function from a known nonlinear process and see how
much the phase shift messes it up.
- Steve
12:04 p.m.
On Mon, Nov 04, 2002 at 02:42:49 +0000, Steve Harris wrote:
I think its probably worth a try with saws, at least
it will make
extracting the transfer function simpler. I will try playing with
extracting a transfer function from a known nonlinear process and see how
much the phase shift messes it up.
This is actualy looking quite promising, you have to discard the transient
zero crossing of the saw wave and a few dB either side (the filtering
messes it up too much), but that is not a problem, you just overshoot the
desired input amplitude by a few dB.
I've attached some example code that generates a saw wave (p), applies a
nonlinear function (x, p^2 * 0.3 + p^3 * 0.73 - p^5 * 0.1), a highpass
filter and a delay (about 16.1 samples), then puts the transfer function
into a table (ignoring the delay). It can guess the correct input/output
alignment by looking at the shape of the out-of-sync transfer function.
There is an example here: http://plugin.org.uk/tmp/foo.png
The red line is the derived transfer function and the green line is the
actual function. They only really differ by amplitude (due to the filter)
and a bit of noise (cos of small input set, ~100ms at 44.1).
The code is in http://plugin.org.uk/tmp/foo.c but its not pretty.
- Steve
3:48 p.m.
Steve Harris wrote:
I've attached some example code that generates a
saw wave (p), applies a
nonlinear function (x, p^2 * 0.3 + p^3 * 0.73 - p^5 * 0.1), a highpass
filter and a delay (about 16.1 samples), then puts the transfer function
into a table (ignoring the delay). It can guess the correct input/output
alignment by looking at the shape of the out-of-sync transfer function.
There is an example here: http://plugin.org.uk/tmp/foo.png
not bad.
have you tried it with real-world data yet?
tim
6:31 p.m.
On Mon, Nov 04, 2002 at 08:42:26 +0100, Tim Goetze wrote:
Nope, dont have any yet. I'l try grabbing something of a distoriton pedal
or something, can you try one of your amp?
- Steve
I've
attached some example code that generates a saw wave (p), applies a
nonlinear function (x, p^2 * 0.3 + p^3 * 0.73 - p^5 * 0.1), a highpass
filter and a delay (about 16.1 samples), then puts the transfer function
into a table (ignoring the delay). It can guess the correct input/output
alignment by looking at the shape of the out-of-sync transfer function.
There is an example here: http://plugin.org.uk/tmp/foo.png
not bad.
have you tried it with real-world data yet? 
7:01 p.m.
On Mon, Nov 04, 2002 at 10:27:39PM +0000, Steve Harris wrote:
On Mon, Nov 04, 2002 at 08:42:26 +0100, Tim Goetze
wrote:
I'd like to get you some samples from my old Gibson at
various settings (it's very versatile, almost sounds like
a different amp depending on how you set it).
What would I need to run through it?
Sawtooth wave, right, but the amplitude & frequency -
they're swept, right?
what are starting & ending values? linear sweep?
Also, I don't have a way to tap the amp signal directly,
there was no "line out" on guitar amps in 1963. :)
I can only mic the speaker. Does that work for what you're doing?
--
Paul Winkler
http://www.slinkp.com
"Welcome to Muppet Labs, where the future is made - today!"
have you tried it with real-world data yet?
Nope, dont have any yet. I'l try grabbing something of a distoriton pedal
or something, can you try one of your amp?
7:58 p.m.
On Mon, Nov 04, 2002 at 02:58:03 -0800, Paul Winkler wrote:
I'd like to get you some samples from my old
Gibson at
various settings (it's very versatile, almost sounds like
a different amp depending on how you set it).
What would I need to run through it?
I just tried using a constant frequency, integer period saw wave.
We would probably need to grab several saws at different frequencies it
this works.
Also, I don't have a way to tap the amp signal
directly,
there was no "line out" on guitar amps in 1963. :)
I can only mic the speaker. Does that work for what you're doing?
No idea, I dont even know if it works if you can tap the signal directly
yet ;) I would think its hard to extract data from it if its gone through
cones and air. It would be interesting to see the results if you are
willing though.
- Steve
9:04 p.m.
Steve Harris wrote:
On Mon, Nov 04, 2002 at 08:42:26 +0100, Tim Goetze
wrote:
Nope, dont have any yet. I'l try grabbing something of a distoriton pedal
or something, can you try one of your amp?
sure thing. is a sawtooth from the virus ok or would you prefer
an integer cycle frequency? still have to do the rewiring if
you need it from the box.
tim
I've
attached some example code that generates a saw wave (p), applies a
nonlinear function (x, p^2 * 0.3 + p^3 * 0.73 - p^5 * 0.1), a highpass
filter and a delay (about 16.1 samples), then puts the transfer function
into a table (ignoring the delay). It can guess the correct input/output
alignment by looking at the shape of the out-of-sync transfer function.
There is an example here: http://plugin.org.uk/tmp/foo.png
not bad.
have you tried it with real-world data yet?
5 Nov
5 Nov
4:47 a.m.
On Tue, Nov 05, 2002 at 01:44:03 +0100, Tim Goetze wrote:
sure thing. is a sawtooth from the virus ok or would
you prefer
an integer cycle frequency? still have to do the rewiring if
you need it from the box.
It really needs to be integer period, and in sample sync.
- Steve
2:37 p.m.
Steve Harris wrote:
On Tue, Nov 05, 2002 at 01:44:03 +0100, Tim Goetze
wrote:
it's going to take some time until i get around to do this
due to a muted soundcard (unmuting fails because of an alsa
bug).
tim
sure thing. is a sawtooth from the virus ok or
would you prefer
an integer cycle frequency? still have to do the rewiring if
you need it from the box.
It really needs to be integer period, and in sample sync.
4 Nov
4 Nov
7:53 p.m.
On Mon, Nov 04, 2002 at 08:42:26 +0100, Tim Goetze wrote:
Just tried it, tested with a synthetic overdrive effect. The results look
plausible. but I haven't tried shaping anything with them.
http://plugin.org.uk/tmp/bar.png
The data is attached incase anyone wants to try building the waveshaper
out of the data. The ends of the curve are bogus, they would need to be
removed to get an accurate shape.
In the data file the first column is the input amplitude, the second is
the output amplitude.
I probably wont have any time to work on this tommorow, but I will have
plenty of brainwidth.
- Steve
There is an
example here: http://plugin.org.uk/tmp/foo.png
have you tried it with real-world data yet? 
5 Nov
5 Nov
1:10 p.m.
On Mon, Nov 04, 2002 at 11:49:59 +0000, Steve Harris wrote:
Just tried it, tested with a synthetic overdrive
effect. The results look
plausible. but I haven't tried shaping anything with them.
I mentioned our stumbling ardound in the dark on the music-dsp list, and
Jay Tomas came back with this:
---
hints for characterizing the amp:
1) don't try to characterize the whole amp at once with test signals..
characterize each preamp stage individually. Attempting to characterize
the clip curve of the whole amp will yield a frequency dependent clip
curve which isn't much use to you in a software implementation.
2) start with trying to measure the pre and post clip EQ curves with
test signals and then
use this info to deduce the effective static non-linear clip curve that
would approximately model what you are seeing.
---
So, it sounds like f we can get a curve off the seperate stages they
wont need fancy filter mangling, and saws are the way to go (well, noone
pointed and laughed ;)
Point 2 is a good one, by applying the EQ effect of the amp in reverse we
can counter out the filtering that makes the saws hard to analyse.
The bad news is that it means going inside someones amp with a probe, I
would have a go myself, but I've allready killed my last amp, and my
electronics skills are bad enough that I would probably fry myself ;)
Brave and/or stupid volunteers welcome ;)
- Steve
3:31 p.m.
Steve Harris wrote:
The bad news is that it means going inside someones amp
with a probe, I
would have a go myself, but I've allready killed my last amp, and my
electronics skills are bad enough that I would probably fry myself ;)
Brave and/or stupid volunteers welcome ;)
looking at a fender tube amp schematic shows +140 V after the first
12AY7 -- i don't have any equipment that would make this signal
digestable by a computer.
tim
7 p.m.
On Tue, Nov 05, 2002 at 08:26:40 +0100, Tim Goetze wrote:
Steve Harris wrote:
Yeah, ggjjjzap. I dont have anything that can handle that either. How do
those attenuating osciloscope probes work? Maybe you can build something
equivalent in an audio lead, though I wouldnt be happy putting anything
like that through a homemade attenuator into my desk.
- Steve
The bad news is that it means going inside
someones amp with a probe, I
would have a go myself, but I've allready killed my last amp, and my
electronics skills are bad enough that I would probably fry myself ;)
Brave and/or stupid volunteers welcome ;)
looking at a fender tube amp schematic shows +140 V after the first
12AY7 -- i don't have any equipment that would make this signal
digestable by a computer. 
6 Nov
6 Nov
1:10 p.m.
On Tuesday 05 November 2002 17:57, Steve Harris wrote:
On Tue, Nov 05, 2002 at 08:26:40 +0100, Tim Goetze
wrote:
> Steve Harris wrote:
> >The bad news is that it means going inside someones amp with a probe, I
> >would have a go myself, but I've allready killed my last amp, and my
> >electronics skills are bad enough that I would probably fry myself ;)
> looking at a fender tube amp schematic shows +140
V after the first
> 12AY7 -- i don't have any equipment that would make this signal
> digestable by a computer.
Yeah, ggjjjzap. I dont have anything that can handle
that either. How do
those attenuating osciloscope probes work? Maybe you can build something
equivalent in an audio lead, though I wouldnt be happy putting anything
like that through a homemade attenuator into my desk.
I have one of these:
http://www.velleman.be/Product.asp?lan=1&id=338488 (Velleman PCS64i digital
oscilliscope adapter for PC). It's good to 600V, I think. It attaches to
your parallel port. RadioShack carries them in the US. They run about $300.
As it also works as a transient recorder and spectrum analyzer (to 32MHz), it
is a great buy. Even though it is an 8-bit device, it is useful in showing
what the signal looks like, as your eye can't distinguish the difference on a
75dpi screen between 16 bit and 8 bit samples.
If you were close enough, I'd let you borrow my Tektronix 475 analog
oscope.... With a 10x attenuator probe it's good for signal amplitudes of a
couple hundred volts with a DC offset of 600V.
If I had the amp in question here, I would be happy to do the trace for you; I
routinely work on hot tube gear. (I'm a broadcast engineer; broadcast
transmitters that are tube output have 10kV or more floating around them.)
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
1:54 p.m.
On Wed, Nov 06, 2002 at 12:07:24PM -0500, Lamar Owen wrote:
As it also works as a transient recorder and spectrum
analyzer (to 32MHz), it
is a great buy. Even though it is an 8-bit device, it is useful in showing
what the signal looks like, as your eye can't distinguish the difference on a
75dpi screen between 16 bit and 8 bit samples.
The trick really is get get a sample off of it, 8bit might be enough, but
I suspect 16 will be easier to deal with.
If you were close enough, I'd let you borrow my
Tektronix 475 analog
oscope.... With a 10x attenuator probe it's good for signal amplitudes of a
couple hundred volts with a DC offset of 600V.
I have a scope, but I turned the probes into 1/4" jacks a while back ;)
I dont have a working amp either.
- Steve
3:13 p.m.
New subject: Insturmenting the amplifier for sampling (Was:Re: [linux-audio-dev] Re: Cheby amp code?)
On Wednesday 06 November 2002 12:50, Steve Harris wrote:
On Wed, Nov 06, 2002 at 12:07:24PM -0500, Lamar Owen
wrote:
> As it also works as a transient recorder and spectrum analyzer (to
The trick really is get get a sample off of it, 8bit
might be enough, but
I suspect 16 will be easier to deal with.
A series 1 microfarad capacitor in the 1kV breakdown range in series with your
line input should work, but you might want to use an old soundcard and
machine to do it... :-) Paper 1 mic caps in that voltage range are available
-- in a pinch 0.47 or even 0.2 would work, but the lower you go on
capacitance the more the probe will pollute the signal. Assuming the
soundcard's line input doesn't already have a low-voltage series coupling cap
already, in which case you have real problems, because the effective
capacitance or two series caps is equal to the reciprocal of the sums of the
reciprocals of the individual caps (the same as parallel resistors).
I have a couple of 50 microfarad oil-filled caps in the 12kV range, but you
don't want me to ship them (all 100 plus pounds) to you.... The capacitors
that are used in mag-reg HID arc ballasts are high value and high voltage,
but I'm not sure what their frequency response will be like. ESR for those
big oil-filled paper caps at 1kHz is probably going to be quite high due to
the inductance of the paper/foil rolls.
It just depends upon how precise you need it to be. A high voltage
nonpolarized electrolytic would be close to ideal -- tweeter caps, for
instance, for high-power speakers would be good, as at the typical input
impedance for a sound card they would pass the lower frequencies just fine
(they ordinarily work with a low impedance driver, thus decreasing the time
constant and increasing the cutoff frequency). Their ESR shouldn't be a
problem, as they're designed for low impedances.
Photflash electrolytics would be the next best thing, configured in a nonpolar
back-to-back arrangement, as they have extremely low ESR and inductance,
meaning they should be pretty linear.
I have a mostly working Fender Bassman available for testing. Does anyone
here have a schematic for that beast? I can get the particulars. It has a
nasty hum, and the bias on the 6L6's is quite bad (they run bright orange
instead of just barely red). There are probably many resistors that have
changed values in it, along with several dried out electolytics.
I also have a Dynaco three-tube kit special (A450 output with twin 8417's)
available for testing.
And I'm still wondering whether a SPICE model of the amplifier in question
would tell us anything about the waveforms.... :-)
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
3:51 p.m.
New subject: Insturmenting the amplifier for sampling (Was:Re: [linux-audio-dev] Re: Cheby amp code?)
On Wed, Nov 06, 2002 at 02:09:50PM -0500, Lamar Owen wrote:
On Wednesday 06 November 2002 12:50, Steve Harris
wrote:
I'm an electrical engineer and I've been following this discussion with
interest.
A series 1 microfarad capacitor in the 1kV breakdown
range in series with your
line input should work, but you might want to use an old soundcard and
machine to do it... :-) Paper 1 mic caps in that voltage range are available
-- in a pinch 0.47 or even 0.2 would work, but the lower you go on
capacitance the more the probe will pollute the signal. Assuming the
soundcard's line input doesn't already have a low-voltage series coupling cap
already, in which case you have real problems, because the effective
capacitance or two series caps is equal to the reciprocal of the sums of the
reciprocals of the individual caps (the same as parallel resistors).
Ahh, but like parallel resistors, if one is "big" the other resistance will set
the value. So choose a "big" balue for the external cap and you'll be
fine.
I have a couple of 50 microfarad oil-filled caps in
the 12kV range...
I'd like to see those reviewed in an audiophile "capacitor review!" Heh.
Photflash electrolytics would be the next best thing,
configured in a nonpolar
back-to-back arrangement, as they have extremely low ESR and inductance,
meaning they should be pretty linear.
But you don't need nonpolar caps... Just get the polarity right. Assuming your
soundcard is going to be grounded with the amp (grounding in tube amps is
itself... interesting), you will, in almost every circumstance, want the + side
to go to the amp.
I have a mostly working Fender Bassman available for
testing. Does anyone
here have a schematic for that beast? I can get the particulars. It has a
nasty hum, and the bias on the 6L6's is quite bad (they run bright orange
instead of just barely red). There are probably many resistors that have
changed values in it, along with several dried out electolytics.
I've seen schematics for the Bassman on the web. In fact, I used the Bassman for
my computer "wallpaper" for a while. Look around a bit. If you don't find
it I
may have a hard copy I can scan in. AFAICR, My Bassman 100 has a pot internally
to adjust the bias current and a sticker inside that gives the right current
value. (Clip an ammeter across the output transformer primary, adjust for
recommended bias).
I also have a Dynaco three-tube kit special (A450
output with twin 8417's)
available for testing.
And I'm still wondering whether a SPICE model of
the amplifier in question
would tell us anything about the waveforms.... :-)
Probably not much more than careful thought and intuitive mathematical
modeling... besides, SPICE probably doesn't consider things like capacitor
soakage and transformer hysteresis, which can give an amp some 'flavor'.
You can probably do this kind of testing without taking the amp apart... just
hook a wire around the appropriate tube pin. I hesistate to give instructions,
because you can end up seriously dead if you aren't careful with tube equipment.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
Regards,
Mark
markrages(a)mlug.missouri.edu
5 p.m.
New subject: Insturmenting the amplifier for sampling (Was:Re: [linux-audio-dev] Re: Cheby amp code?)
On Wednesday 06 November 2002 14:47, Mark Rages wrote:
On Wed, Nov 06, 2002 at 02:09:50PM -0500, Lamar Owen
wrote:
> cap already, in which case you have real problems, because the effective
> capacitance or two series caps is equal to the reciprocal of the sums of
> the reciprocals of the individual caps (the same as parallel resistors).
Ahh, but like parallel resistors, if one is
"big" the other resistance will
set the value. So choose a "big" balue for the external cap and you'll be
fine.
Which means the soundcard's cap would be the upper asymptote. Which is where
I was driving towards -- unless you have a good soundcard, that is. Typical
soundcards don't have really good input coupling caps, and that would 'color'
the waveform seen by the A/D.
> Photflash electrolytics would be the next best
thing, configured in a
> nonpolar back-to-back arrangement, as they have extremely low ESR and
> inductance, meaning they should be pretty linear.
But you don't need nonpolar caps... Just get the
polarity right. Assuming
your soundcard is going to be grounded with the amp (grounding in tube amps
is itself... interesting), you will, in almost every circumstance, want the
+ side to go to the amp.
Ah, but did you see the schematic of the amp in question? 'Ground' could be
the 120VAC LINE if you have the grounding switch set wrong. Best be safe,
since the amp's ground and the soundcard's ground could have 120VAC
differential.
Besides, a nonpolar coupling cap would be more useful in the generic case, so
that you could trace through both a negative and positive supply voltage
situation.
I've seen schematics for the Bassman on the web. .
Yeah, I'll google around for it. I just have to see which flavor of Bassman
this one is (twin 12's with separate head). It is very old, that much I know.
> And I'm still wondering whether a SPICE model
of the amplifier in
> question would tell us anything about the waveforms.... :-)
Probably not much more than careful thought and
intuitive mathematical
modeling... besides, SPICE probably doesn't consider things like capacitor
soakage and transformer hysteresis, which can give an amp some 'flavor'.
You can model those things easily enough with the port parameters SPICE
allows. And that's why I suggested such. SPICE is incredibly powerful -- I
have seen the pspice commercial flavor used for modeling HID lamp ballasts,
which operate on magnetic circuit principles.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
5:23 p.m.
New subject: Insturmenting the amplifier for sampling (Was:Re: [linux-audio-dev] Re: Cheby amp code?)
On Wed, Nov 06, 2002 at 03:57:05PM -0500, Lamar Owen wrote:
Yeah, this is what I was getting at. If the amp's ground is not at the
soundcard ground, there are several choices:
1) Audio transformer coupling. Will color signal to some degree.
2) Capactor coupling, but just flip the amp's ground switch till you
measure 0V between computer and amp.
3) Capacitor coupling, computer or amp plugged into isolation transformer.
4) Capacitor coupling, ground lift on computer or amp... not recommended.
In any case, a capacitor by itself is probably not going to be enough to
couple the signal.
Sweet.
On Wednesday 06 November 2002 14:47, Mark Rages
wrote:
I'll disagree here, because the input coupling caps are just going to gave
a 6dB/oct rolloff below frequencies we're interested in, anyway. If you're
speaking only about capacitor quality, then specifics are necessary to draw
a good conclusion.
On Wed, Nov 06, 2002 at 02:09:50PM -0500, Lamar
Owen wrote:
> cap already, in which case you have real problems, because the effective
> capacitance or two series caps is equal to the reciprocal of the sums of
> the reciprocals of the individual caps (the same as parallel resistors).
Ahh, but like parallel resistors, if one is
"big" the other resistance will
set the value. So choose a "big" balue for the external cap and you'll be
fine.
Which means the soundcard's cap would be the upper asymptote. Which is where
I was driving towards -- unless you have a good soundcard, that is. Typical
soundcards don't have really good input coupling caps, and that would 'color'
the waveform seen by the A/D. > Photflash electrolytics would be the next
best thing, configured in a
> nonpolar back-to-back arrangement, as they have extremely low ESR and
> inductance, meaning they should be pretty linear.
But you don't need nonpolar caps... Just get
the polarity right. Assuming
your soundcard is going to be grounded with the amp (grounding in tube amps
is itself... interesting), you will, in almost every circumstance, want the
+ side to go to the amp.
Ah, but did you see the schematic of the amp in question? 'Ground' could be
the 120VAC LINE if you have the grounding switch set wrong. Best be safe,
since the amp's ground and the soundcard's ground could have 120VAC
differential. Besides, a nonpolar coupling cap would be more useful
in the generic case, so
that you could trace through both a negative and positive supply voltage
situation.
True. My point is that voltages are generally positive in tube equipment.
Nonpolars are harder to find in high voltages.
I've seen
schematics for the Bassman on the web.
Yeah, I'll google around for it. I just have to see which flavor of Bassman
this one is (twin 12's with separate head). It is very old, that much I know. You can model those things easily enough with the port
parameters SPICE
allows. And that's why I suggested such. SPICE is incredibly powerful -- I
have seen the pspice commercial flavor used for modeling HID lamp ballasts,
which operate on magnetic circuit principles.
Glad that's not my job. It's hard to get convergence with hysteresis in
the system. I did some simulation of electirc motors on computer, but we
always linearized the equations first.
I'm afraid I've made a digression of topic here. But if we can figure a way
to sample the signals from a tube amp without killing the people writing
plugins, we have contributed to the Linux audio effort.
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11
Regards,
Mark Rages
markrages(a)mlug.missouri.edu
5:58 p.m.
New subject: Insturmenting the amplifier for sampling (Was:Re: [linux-audio-dev] Re: Cheby amp code?)
On Wed, Nov 06, 2002 at 03:57:05PM -0500, Lamar Owen wrote:
Yeah, I'll google around for it. I just have to
see which flavor of Bassman
this one is (twin 12's with separate head). It is very old, that much I know.
If the covering is tweed, it's a real collector's item...
likewise if the control panel plate is black.
If the covering is black vinyl and the control panel plate is silver,
it's probably just your garden variety great piece of gear. :)
--PW
--
Paul Winkler
http://www.slinkp.com
"Welcome to Muppet Labs, where the future is made - today!"
7989
days inactive
7992
days old
linux-audio-dev@lists.linuxaudio.org
25 comments
5 participants
participants (5)
-
Lamar Owen -
Mark Rages -
Paul Winkler -
Steve Harris -
Tim Goetze