--- In ham_amplifiers@..., "pentalab" <jim.thomson@...> wrote:
If u scroll through the messages... u will see that # 863 has been
blown out ! I responded to it last night... then it comes back
as "message 863 can't be retrieved" Did the fellow blow it out
himself... or was it censored... or what ?
Yahoogroups allows the originator of a posting to delete it. Or the
moderator can. I don't fundamentally like that - once posted, a
posting ought stay as a matter of record.
The 2 tone test is flawed, and isn't used anymore in the
commercial world. By juggling the spacing of the 2 x tones, you
can hit a .."sweet spot" and come up with really good IMD numbers.
That's usually with very wide spacing. At very narrow spacing, you
challenge the power supply bypassing, so the worst numbers often occur
at that test condition.
They all use the noise test these days... pump the xcvr with
bandwidth limited white noise.... that will drive the xcvr to full
pep output.... and simulate voice or many data tones. In the
commercial world, it's common to use 4 x multiplexed ssb
channels... or as many as 16 x mark/space combo's... sometimes
interleaved.
Noise Power Ratio puts band-limited white noise through the rig, with
a notch in the middle of it. IMD products will tend to fill in the
notch. Very harsh test.
You will read the arguments of referencing IMD to one tone of
a 2 x tone sig instead of PEP. For several multiplexed ssb
channels or many data tones... this is valid. For a single
voice channel on ssb, like we use.... it's not needed.
The ARRL USED to ref IMD to one tone.... now they ref to
PEP. S-meter's on xcvr's are all PEAK reading devices. We all
hear.... "your signal is 10 over S-9.... but ur splatter is S-
6.... 4 khz away" The ref here is IMD to PEP. We don't
hear...." gee, I gotta ref off freq splatter to on freq signal
strength MINUS 6db".
Here's a quote from "SSB systems and circuits"
"ANOTHER deficiency of the 2 x tone test is that the "3rd order"
products observed on a spectrum analyzer are actually the sum of
the 3rd and ALL higher ODD order components. Typ, the 5th order
component is OUT of phase with the 3rd, which tends to PRODUCE
distortion CANCELLATION. This leaves the FALSE impression that the
IMD is better than it really is."
Sad. It's in error. The 5th order component is at a different
frequency than the 3rd order component, therefore they are completely
distinguishable from each other.
Ex: 5MHz and 5.001MHz are the incident sigs. The 3rd ord will be
4.999 and 5.002 MHz. 5th order will be 4.998 and 5.003 MHz. They
don't sit on each other, and don't add to each other in any way.
They go on to say a better method for single channel ssb voice
is to look at plane voice on a digital storage spectrum
analyzer... over a long time period... whereby all the out of
band IMD products are held in a "peak hold" mode.... the
concept being to look at long term spectral power densisty.
A 2 x tone test imo... is pretty lame duck... it won't dynamically
exercise HV and bias + fil supplies either.... since the 2 x
tone puts everything in a .."static state".
Yes, quite. The difficult is in coming up with a repeatable test.
It's not scientific unless it's repeatable - two guys in two different
locations, using two sets of test gear should come up with the same
results. The BW-limited noise test is a good one in this regard.
I tried the white noise test... and it def produces more IMD than
plane voice ever will.... since the white noise looks like
thousands of tones... all beating against each other.
BTW... when running white noise into the xcvr.... the plate current
on the linear is EXACTLY 1/2 the key down value.... so u can
use white noise to tweak the tune/load caps to max pep output.
You're driving it too hard. Peak to average ratio of BW-limited white
noise (the voltage distribution of BW-limited white noise is closer to
a Rayleigh distribution) is approx 16:1. The word "approx" is
important beause in truth, there is some percentage probability of ANY
power level being present at some time. 16:1 is the diff between 50%
probabiliy and 1%. Adequate for communications-grade amps.
Even EESB comes out looking better than a white noise test... OR
using W8JI's convoluted 2 x tone test, where he uses 2 x
extremly wide spaced tones.... like 100 hz and 3100 hz.... then
sez the total IMD BW is 9 khz wide.
In U.S. dialects, a vocalized sibiliant contains both low and high.
E.G. "Z" or "J". However, the wide-spaced test is actually one of
those that does not stress the power supply much, therefore you tend
to get artificially good values.
Try as hard as you want... you can't find ANY combo of words or
phrases that will produce SIMULTANEOUS highs and lows.
Sure can...any vocalized sibilant. To be exact, "Z" is a "voiced
alveolar sibilant" and is often used as a "torture test" for system BW
commercially when test gear is unavailable.
Has anybody tried adjusting the Zsac on these big tubes from one
extreme to the other.. and actually measured the imd ?? Kinda a
moot point... unless the xcvr is the same or better IMD wise,
nobody would ever hear any benefit.
The tendency is for IMD to vary in only a small amount. You can
improve it at low levels by sending the ZSAC very high, but that does
not change IMD at high levels. When you get ZSAC to a too-low level,
IMD shoots up quite a bit. That's why even in commercial gear,
fixed-voltage bias is common, it just doesn't matter much as long as
you've got some ZSAC.
Later... Jim VE7RF
Z
