|
If you just want to measure amplitude, wouldn't a power meter be easier ?
|
Well yes it might be, but if you don't have one ...
Also for the uninitiated (i.e. me) how accurate are power meters?
Dave
________________________________
From: TekScopes@... [mailto:TekScopes@...] On Behalf Of HankC
Sent: 15 February 2013 14:50
To: TekScopes@...
Subject: [TekScopes] Re: Frequency response flatness in conventional sampling (say 7S11/S
If you just want to measure amplitude, wouldn't a power meter be easier ?
|
Albert, I wouldn't use a sampling scope or even an analog one pushing its BW to calibrate RF generators, except maybe for comparing one against another. A power meter like HP848X heads with HP43X meter is a good way to go, but expensive.
You can also roll your own with modern power level detector ICs - they can also be used directly in the generator leveling loops to improve performance. I'm familiar with some from Analog Devices, but I'm sure the usual companies have plenty to choose from too. These are very common ICs used in huge quantities for all the wireless gear out there.
Here's a place to start looking:
It would take quite a bit of investigation and experimenting, but the trick is mostly to get the desired broadband flatness in the coupling of the signal to the IC, and of course, the right dynamic range. Logarithmic and linear responses are available. With linear, you can expand the measurement for better resolution over a small range for calibration purposes, but have less overall dynamic range.
These devices usually measure true RMS voltage over the entire bandwidth, so if your waves have lots of distortion, the harmonic power will be counted too, up to the BW, then diminished. This also could give a quite different result than looking at the p-p envelope even on a perfectly flat scope - another reason to not use a scope, unless p-p voltage is what's wanted.
Ed
toggle quoted message
Show quoted text
--- In TekScopes@..., "David C. Partridge" <david.partridge@...> wrote:
Well yes it might be, but if you don't have one ...
Also for the uninitiated (i.e. me) how accurate are power meters?
Dave
________________________________
From: TekScopes@... [mailto:TekScopes@...] On Behalf Of HankC
Sent: 15 February 2013 14:50
To: TekScopes@...
Subject: [TekScopes] Re: Frequency response flatness in conventional sampling (say 7S11/S
If you just want to measure amplitude, wouldn't a power meter be easier ?
|
I think using a sampling oscilloscope for flatness calibration is a great idea. The sampling heads are both very high bandwidth and have a very predictable frequency response. The weakest links will be the SWR match and termination but that applies to any system. You can do away with cable losses by using a sampling head extender. If you take the 3db bandwidth numbers I posted earlier and divide by 4, that is the point where the sample head output will be down by 2%. The second number shown is where they will be down by 1%: S-1 260 MHz 190 MHz S-2 1.18 GHz 869 MHz S-4 3.60 GHz 2.61 GHz S-6 2.99 GHz 2.17 GHz I think this is the first time I have had a need to do math involving a sin(x)/x function. On Fri, 15 Feb 2013 17:45:49 -0000, "Ed Breya" <edbreya@...> wrote: Albert, I wouldn't use a sampling scope or even an analog one pushing its BW to calibrate RF generators, except maybe for comparing one against another. A power meter like HP848X heads with HP43X meter is a good way to go, but expensive.
You can also roll your own with modern power level detector ICs - they can also be used directly in the generator leveling loops to improve performance. I'm familiar with some from Analog Devices, but I'm sure the usual companies have plenty to choose from too. These are very common ICs used in huge quantities for all the wireless gear out there.
Here's a place to start looking:
It would take quite a bit of investigation and experimenting, but the trick is mostly to get the desired broadband flatness in the coupling of the signal to the IC, and of course, the right dynamic range. Logarithmic and linear responses are available. With linear, you can expand the measurement for better resolution over a small range for calibration purposes, but have less overall dynamic range.
These devices usually measure true RMS voltage over the entire bandwidth, so if your waves have lots of distortion, the harmonic power will be counted too, up to the BW, then diminished. This also could give a quite different result than looking at the p-p envelope even on a perfectly flat scope - another reason to not use a scope, unless p-p voltage is what's wanted.
Ed
|
Hi Ed,
No thanks for your discouraging words about using a sampling 'scope, but thanks for the other references. I knew about those HP thermo-couple methods which are also mentioned in cal procs of predecessors of SG503/SG504. It's a pity that most background info at the site of Analog Devices requires registration, even if the first screen says that such is not required. I suppose these RF ICs also measure temperature changes?
The presence of harmonics indeed raises the question what to use, peak-peak or rms, and how different the results could be. I don't want to think of that since I still try to stay away of spectrum analyzers. As far as I know the SG504 and the like all use p-p detection for leveling. Maybe a challenge for David to create an rms variant of his leveling head?
You mention (as does AD) the importance of correct coupling between signal and measurement device. What about VSWR of the detector input? I suppose the correction factors supplied with Tek's p-p detector also correct for VSWR. But I wonder whether flatness specs for other detectors (for instance the detector mentioned by Raymond) also include VSWR effect. In other words, does such a spec assume constant open voltage amplitude of an ideal 50R signal generator, or assume a constant voltage amplitude at the input of the detector? A VSWR of 1.10 at some frequency accounts for 5% difference between these approaches, isn't it?
Albert
toggle quoted message
Show quoted text
Albert, I wouldn't use a sampling scope or even an analog one pushing its BW to calibrate RF generators, except maybe for comparing one against another. A power meter like HP848X heads with HP43X meter is a good way to go, but expensive.
You can also roll your own with modern power level detector ICs - they can also be used directly in the generator leveling loops to improve performance. I'm familiar with some from Analog Devices, but I'm sure the usual companies have plenty to choose from too. These are very common ICs used in huge quantities for all the wireless gear out there.
Here's a place to start looking:
It would take quite a bit of investigation and experimenting, but the trick is mostly to get the desired broadband flatness in the coupling of the signal to the IC, and of course, the right dynamic range. Logarithmic and linear responses are available. With linear, you can expand the measurement for better resolution over a small range for calibration purposes, but have less overall dynamic range.
These devices usually measure true RMS voltage over the entire bandwidth, so if your waves have lots of distortion, the harmonic power will be counted too, up to the BW, then diminished. This also could give a quite different result than looking at the p-p envelope even on a perfectly flat scope - another reason to not use a scope, unless p-p voltage is what's wanted.
Ed
|
Hi David,
In my response to Ed I said something about VSWR; that crossed your message.
Your calculations (indeed needing the famous sin(x)/x) assume a perfect rectangular windowing function. Do you have any reference for how good or bad this approach might be in practice? The overshoot in step response of my S-4 is in conflict with this theoretical approach.
Albert
toggle quoted message
Show quoted text
--- In TekScopes@..., David <davidwhess@...> wrote:
I think using a sampling oscilloscope for flatness calibration is a
great idea. The sampling heads are both very high bandwidth and have
a very predictable frequency response. The weakest links will be the
SWR match and termination but that applies to any system. You can do
away with cable losses by using a sampling head extender.
If you take the 3db bandwidth numbers I posted earlier and divide by
4, that is the point where the sample head output will be down by 2%.
The second number shown is where they will be down by 1%:
S-1 260 MHz 190 MHz
S-2 1.18 GHz 869 MHz
S-4 3.60 GHz 2.61 GHz
S-6 2.99 GHz 2.17 GHz
I think this is the first time I have had a need to do math involving
a sin(x)/x function.
|
Agreed.... --- In TekScopes@..., David <davidwhess@...> wrote:
I think using a sampling oscilloscope for flatness calibration is a
great idea. The sampling heads are both very high bandwidth and have
a very predictable frequency response.
<text removed>
|
I do not have any in dependant confirmation but I suspect given the relatively low frequencies involved for the sampling heads up to 5% loss that they will be very close to ideal. I do not know of a better way to calibrate for a constant level that does not require something else already calibrated to a better standard except for a thermal RMS based design which itself can be calibrated at DC. If you have even an unleveled microwave signal source, you could use it to find the first null in the sampler frequency response which would tell exactly what the sample gate time is. Do you mean overshoot or blow-by? I know my S-4 sampling heads have a lot of blow-by aberration or whatever that is at about 10ns but show ideal pulse response as far as I can test. My working S-1 shows no blow-by with the same input pulse but my best flat level pulse generator while clean is not fast enough for the S-4. On Fri, 15 Feb 2013 22:35:37 -0000, "Albert" <aodiversen@...> wrote: Hi David,
In my response to Ed I said something about VSWR; that crossed your message.
Your calculations (indeed needing the famous sin(x)/x) assume a perfect rectangular windowing function. Do you have any reference for how good or bad this approach might be in practice? The overshoot in step response of my S-4 is in conflict with this theoretical approach.
Albert
--- In TekScopes@..., David <davidwhess@...> wrote:
I think using a sampling oscilloscope for flatness calibration is a
great idea. The sampling heads are both very high bandwidth and have
a very predictable frequency response. The weakest links will be the
SWR match and termination but that applies to any system. You can do
away with cable losses by using a sampling head extender.
If you take the 3db bandwidth numbers I posted earlier and divide by
4, that is the point where the sample head output will be down by 2%.
The second number shown is where they will be down by 1%:
S-1 260 MHz 190 MHz
S-2 1.18 GHz 869 MHz
S-4 3.60 GHz 2.61 GHz
S-6 2.99 GHz 2.17 GHz
I think this is the first time I have had a need to do math involving
a sin(x)/x function.
|
Hi David,
The step response of my S-4 is very similar to the pictures shown in the cal proc of the manual. Overshoot from 400 ps to 25 ns is specified as 10% or less. Blow-by compensation adjusts the level or trend after 25 ns. Both overshoot and blow-by might well affect the frequency response (far) below 1 GHz I think.
The awkward thing is that we always need something that's again better than what we already have. I have to trust pulse flatness of my S-52 and 264 (also Square Wave for S-2 blow-by) but in reality I have no means to check that. In S-1 and S-2 the blow-by adjustment has considerable effect, so if your S-1 shows no blow-by then you or a previous owner did a good calibration job. I'm not sure I ever re-adjusted my S-4.
I did many amplitude measurements on my 067-0532-00, 3 MHz and 60-500 MHz, using 7T11A/7S11/S-2. At 1 Vpp the amplitude changes are within 1% (that is max - min) over the whole frequency range, including the separate 3 MHz. At 3 Vpp (connected via GR 2X attenuator) still within 2%.
That looks very nice, but in theory a non-flat response of the S-2 could compensate for a non-flat output of the signal generator. I used an S-2 to eliminate a GR to SMA adapter and also because the S-2 allows for internal triggering.
Albert
toggle quoted message
Show quoted text
--- In TekScopes@..., David <davidwhess@...> wrote:
I do not have any in dependant confirmation but I suspect given the
relatively low frequencies involved for the sampling heads up to 5%
loss that they will be very close to ideal. I do not know of a better
way to calibrate for a constant level that does not require something
else already calibrated to a better standard except for a thermal RMS
based design which itself can be calibrated at DC.
If you have even an unleveled microwave signal source, you could use
it to find the first null in the sampler frequency response which
would tell exactly what the sample gate time is.
Do you mean overshoot or blow-by? I know my S-4 sampling heads have a
lot of blow-by aberration or whatever that is at about 10ns but show
ideal pulse response as far as I can test. My working S-1 shows no
blow-by with the same input pulse but my best flat level pulse
generator while clean is not fast enough for the S-4.
On Fri, 15 Feb 2013 22:35:37 -0000, "Albert" <aodiversen@...>
wrote:
Hi David,
In my response to Ed I said something about VSWR; that crossed your message.
Your calculations (indeed needing the famous sin(x)/x) assume a perfect rectangular windowing function. Do you have any reference for how good or bad this approach might be in practice? The overshoot in step response of my S-4 is in conflict with this theoretical approach.
Albert
--- In TekScopes@..., David <davidwhess@> wrote:
I think using a sampling oscilloscope for flatness calibration is a
great idea. The sampling heads are both very high bandwidth and have
a very predictable frequency response. The weakest links will be the
SWR match and termination but that applies to any system. You can do
away with cable losses by using a sampling head extender.
If you take the 3db bandwidth numbers I posted earlier and divide by
4, that is the point where the sample head output will be down by 2%.
The second number shown is where they will be down by 1%:
S-1 260 MHz 190 MHz
S-2 1.18 GHz 869 MHz
S-4 3.60 GHz 2.61 GHz
S-6 2.99 GHz 2.17 GHz
I think this is the first time I have had a need to do math involving
a sin(x)/x function.
|
The blow-by effect on frequency response could be tested. Grossly misadjust it deliberately and see if the frequency response changes. I found it to be the easiest thing to calibrate on my S-4 sampling heads so I would not worry about temporarily misadjusting it. I would also compare two different types of sampling heads like an S-2 and S-4 which have significantly different transient response characteristics in the 10ns range do to design and see if they agree. I may try the above with an S-1 and S-4 and my SG503 just to see what kind of results I get. On Sat, 16 Feb 2013 10:48:35 -0000, "Albert" <aodiversen@...> wrote: Hi David,
The step response of my S-4 is very similar to the pictures shown in the cal proc of the manual. Overshoot from 400 ps to 25 ns is specified as 10% or less. Blow-by compensation adjusts the level or trend after 25 ns. Both overshoot and blow-by might well affect the frequency response (far) below 1 GHz I think.
The awkward thing is that we always need something that's again better than what we already have. I have to trust pulse flatness of my S-52 and 264 (also Square Wave for S-2 blow-by) but in reality I have no means to check that. In S-1 and S-2 the blow-by adjustment has considerable effect, so if your S-1 shows no blow-by then you or a previous owner did a good calibration job. I'm not sure I ever re-adjusted my S-4.
I did many amplitude measurements on my 067-0532-00, 3 MHz and 60-500 MHz, using 7T11A/7S11/S-2. At 1 Vpp the amplitude changes are within 1% (that is max - min) over the whole frequency range, including the separate 3 MHz. At 3 Vpp (connected via GR 2X attenuator) still within 2%.
That looks very nice, but in theory a non-flat response of the S-2 could compensate for a non-flat output of the signal generator. I used an S-2 to eliminate a GR to SMA adapter and also because the S-2 allows for internal triggering.
Albert
--- In TekScopes@..., David <davidwhess@...> wrote:
I do not have any in dependant confirmation but I suspect given the
relatively low frequencies involved for the sampling heads up to 5%
loss that they will be very close to ideal. I do not know of a better
way to calibrate for a constant level that does not require something
else already calibrated to a better standard except for a thermal RMS
based design which itself can be calibrated at DC.
If you have even an unleveled microwave signal source, you could use
it to find the first null in the sampler frequency response which
would tell exactly what the sample gate time is.
Do you mean overshoot or blow-by? I know my S-4 sampling heads have a
lot of blow-by aberration or whatever that is at about 10ns but show
ideal pulse response as far as I can test. My working S-1 shows no
blow-by with the same input pulse but my best flat level pulse
generator while clean is not fast enough for the S-4.
On Fri, 15 Feb 2013 22:35:37 -0000, "Albert" <aodiversen@...>
wrote:
Hi David,
In my response to Ed I said something about VSWR; that crossed your message.
Your calculations (indeed needing the famous sin(x)/x) assume a perfect rectangular windowing function. Do you have any reference for how good or bad this approach might be in practice? The overshoot in step response of my S-4 is in conflict with this theoretical approach.
Albert
--- In TekScopes@..., David <davidwhess@> wrote:
I think using a sampling oscilloscope for flatness calibration is a
great idea. The sampling heads are both very high bandwidth and have
a very predictable frequency response. The weakest links will be the
SWR match and termination but that applies to any system. You can do
away with cable losses by using a sampling head extender.
If you take the 3db bandwidth numbers I posted earlier and divide by
4, that is the point where the sample head output will be down by 2%.
The second number shown is where they will be down by 1%:
S-1 260 MHz 190 MHz
S-2 1.18 GHz 869 MHz
S-4 3.60 GHz 2.61 GHz
S-6 2.99 GHz 2.17 GHz
I think this is the first time I have had a need to do math involving
a sin(x)/x function.
|
Hi David,
So far I didn't find effect of S-6 blow-by adjustment on sine wave amplitudes. I probably did something wrong, since the effect on square waves (tilt) is considerable. A change of C20 in the S-6 only causes a vertical shift of the sine wave. I tested some frequencies from 100 kHz to 1 GHz.
It's very annoying that the SG503 and my GR-type oscillators don't have a trigger output. Since the S-6 doesn't have a trigger output either I'm forced to use a CT-3 trigger pick-off and external triggering. With the S-2 I can compare internal triggering (CT-3 omitted) with such external triggering. The CT-3 has considerable effect on frequency response, about 6% to 8% downward trend when frequency is increased to 900 MHz.
Albert
toggle quoted message
Show quoted text
--- In TekScopes@..., David <davidwhess@...> wrote:
The blow-by effect on frequency response could be tested. Grossly
misadjust it deliberately and see if the frequency response changes. I
found it to be the easiest thing to calibrate on my S-4 sampling heads
so I would not worry about temporarily misadjusting it.
I would also compare two different types of sampling heads like an S-2
and S-4 which have significantly different transient response
characteristics in the 10ns range do to design and see if they agree.
I may try the above with an S-1 and S-4 and my SG503 just to see what
kind of results I get.
|
Sorry for taking a while to get back to you on this. It seems it is time for me to add an S-6 manual to my binder of sampling stuff. Here is why I do not believe the blow-by matters in this application: When you are looking at the aberrations from a clean pulse measured by a sampling head, instead of the impulse response which correlates with the frequency and phase response of a linear network, you see charge coupled through the sampling bridge from the fast edge and level shift. When measuring an AC signal, there are no fast edges and no DC level shift to couple through the sampling bridge. In addition, the memory gate is too slow to see any AC leakage. There is a simulation showing the S-6 AC feedthrough here: I like discussing this subject because it yields ideas for correcting the annoying blow-by response in my S-4 sampling heads. Unfortunately if it was easy to correct, I suspect Tektronix would have done so. Do you actually need triggering to see just the peak to peak values? I think you could add probe a different part of the generator circuit to find a trigger so the leveled output could go only to the sampling head. On Mon, 18 Feb 2013 17:02:21 -0000, "Albert" <aodiversen@...> wrote: Hi David,
So far I didn't find effect of S-6 blow-by adjustment on sine wave amplitudes. I probably did something wrong, since the effect on square waves (tilt) is considerable. A change of C20 in the S-6 only causes a vertical shift of the sine wave. I tested some frequencies from 100 kHz to 1 GHz.
It's very annoying that the SG503 and my GR-type oscillators don't have a trigger output. Since the S-6 doesn't have a trigger output either I'm forced to use a CT-3 trigger pick-off and external triggering. With the S-2 I can compare internal triggering (CT-3 omitted) with such external triggering. The CT-3 has considerable effect on frequency response, about 6% to 8% downward trend when frequency is increased to 900 MHz.
Albert
--- In TekScopes@..., David <davidwhess@...> wrote:
The blow-by effect on frequency response could be tested. Grossly
misadjust it deliberately and see if the frequency response changes. I
found it to be the easiest thing to calibrate on my S-4 sampling heads
so I would not worry about temporarily misadjusting it.
I would also compare two different types of sampling heads like an S-2
and S-4 which have significantly different transient response
characteristics in the 10ns range do to design and see if they agree.
I may try the above with an S-1 and S-4 and my SG503 just to see what
kind of results I get.
|
Hi David,
Your reasoning might be all right for the blow-by effect itself. What still remains is the effect of blow-by compensation. Not at high frequencies, but at frequencies in the bandpass (S-4: 0.5 to 3 MHz) of the preamplifier. The 65-500 MHz oscillator has a 3 MHz reference.
I forgot that the S-4 has a trigger output. I used my S-4 now and found that the 3 MHz amplitude was 10% low compared to the 65 MHz amplitude, while the S-2 gave almost identical amplitudes. Above 65 MHz the S-2 and the S-4 give nearly identical results except that the S-4 shows a superimposed decrease of amplitude when frequency increases (about 1% per 100 MHz). Terrible enough of course!
Because of the far deviating results at 3 MHz I did some low frequency measurements at the S-4 while comparing R18 fully cw to fully ccw. At 3 MHz the difference was +1%, increasing to around +3% around 1.5 Mhz, then decreasing to -3% around 0.4 MHz. No difference at 50 kHz.
A severe amplitude drop of the S-4 at and around 3 MHz was present regardless the signal generator used (SG503, Type 191, GR-type). The effect of blow-by compensation is far too small to account for this.
I didn't check my S-4 for faults like diode leakage.
The SG504 (which I don't have) takes the "monitoring" signal from a point somewhere before the leveling amplifier. A similar pick-off would be possible in the GR-type oscillators indeed.
BTW I think there has been some mechanical change in the 65-500 MHz oscillator after I checked flatness some time ago. The frequency response with S-2 is very nice flat up to 440 MHz, but then increases very sharp to a +10% peak at 465 MHz. Then drops again. The oscillator is very sensitive to the (mechanical) position of an adjustable coupling loop.
Albert
toggle quoted message
Show quoted text
--- In TekScopes@..., David <davidwhess@...> wrote:
Sorry for taking a while to get back to you on this. It seems it is
time for me to add an S-6 manual to my binder of sampling stuff.
Here is why I do not believe the blow-by matters in this application:
When you are looking at the aberrations from a clean pulse measured by
a sampling head, instead of the impulse response which correlates with
the frequency and phase response of a linear network, you see charge
coupled through the sampling bridge from the fast edge and level
shift.
When measuring an AC signal, there are no fast edges and no DC level
shift to couple through the sampling bridge. In addition, the memory
gate is too slow to see any AC leakage. There is a simulation showing
the S-6 AC feedthrough here:
I like discussing this subject because it yields ideas for correcting
the annoying blow-by response in my S-4 sampling heads. Unfortunately
if it was easy to correct, I suspect Tektronix would have done so.
Do you actually need triggering to see just the peak to peak values? I
think you could add probe a different part of the generator circuit to
find a trigger so the leveled output could go only to the sampling
head.
On Mon, 18 Feb 2013 17:02:21 -0000, "Albert" <aodiversen@...>
wrote:
Hi David,
So far I didn't find effect of S-6 blow-by adjustment on sine wave amplitudes. I probably did something wrong, since the effect on square waves (tilt) is considerable. A change of C20 in the S-6 only causes a vertical shift of the sine wave. I tested some frequencies from 100 kHz to 1 GHz.
It's very annoying that the SG503 and my GR-type oscillators don't have a trigger output. Since the S-6 doesn't have a trigger output either I'm forced to use a CT-3 trigger pick-off and external triggering. With the S-2 I can compare internal triggering (CT-3 omitted) with such external triggering. The CT-3 has considerable effect on frequency response, about 6% to 8% downward trend when frequency is increased to 900 MHz.
Albert
--- In TekScopes@..., David <davidwhess@> wrote:
The blow-by effect on frequency response could be tested. Grossly
misadjust it deliberately and see if the frequency response changes. I
found it to be the easiest thing to calibrate on my S-4 sampling heads
so I would not worry about temporarily misadjusting it.
I would also compare two different types of sampling heads like an S-2
and S-4 which have significantly different transient response
characteristics in the 10ns range do to design and see if they agree.
I may try the above with an S-1 and S-4 and my SG503 just to see what
kind of results I get.
|
David C. Partridge
David Hess