|
Spectrum Analyzer - VCO Results
I've tried benching up a Minicircuits POS-900W VCO module.
With a Spectrum Analyzer (RBW 1 kHz, VBW 1kHz, peak hold, 100kHz span) I've been hoping to see a clean waveform that drops to -75dBc at 1,000Hz away, as specified in the Minicircuits spec. Instead, I have something that looks like a triangle spreading for MHz! I've got in such a mess, my brain is now totally confused (not difficult!). I just don't understand why the trace doesn't look good, compared to my own discrete design(lower frequency) VCOs. Even with the POS-900W out of circuit (ie. no PLL), well grounded, and run from batteries, the waveform is still more-or-less a triangle! What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Al. ---- mail[underscore]me[at]freenet[dot]co[dot]uk |
What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB |
What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB |
"Harold E. Johnson" wrote in message
news:_fwlc.22193$kh4.1279179@attbi_s52... What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB Thank you for the reply. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Even if the conections aren't ideal, I'd expect 10dB or so more noise, but it doesn't change however good/bad it's shielded. I'm probably doing something really stupid. Al. --- |
"Harold E. Johnson" wrote in message
news:_fwlc.22193$kh4.1279179@attbi_s52... What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB Thank you for the reply. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Even if the conections aren't ideal, I'd expect 10dB or so more noise, but it doesn't change however good/bad it's shielded. I'm probably doing something really stupid. Al. --- |
Alex Collins wrote:
"Harold E. Johnson" wrote in message Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Well, Harold is pointing out that noise on the tuning line will show up as broad phase noise. What are you using to drive the tuning voltage line? The old Ramsey FX440 UHF transceiver had a horrific noise floor on the transmitted audio, since the VCO had about 10MHz/volt sensitivity and the audio modulation was fed from an ordinary op-amp into the tuning line. You could be seeing a very similar problem. Dana |
Alex Collins wrote:
"Harold E. Johnson" wrote in message Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Well, Harold is pointing out that noise on the tuning line will show up as broad phase noise. What are you using to drive the tuning voltage line? The old Ramsey FX440 UHF transceiver had a horrific noise floor on the transmitted audio, since the VCO had about 10MHz/volt sensitivity and the audio modulation was fed from an ordinary op-amp into the tuning line. You could be seeing a very similar problem. Dana |
I'm not terribly familiar with the MiniCircuits VCOs, but I have been
involved with getting the phase noise down to low values on some of our designs. I can tell you that it's important to have power supplies be very clean, as well as to have the control voltage very clean (down into the microvolt region or below) as Harold says. But from what you wrote, I gather there's a problem even with things supposedly very clean. BTW, I don't 'zactly understand the "POS-900W out of circuit (ie. no PLL)" bit. What are you measuring if the oscillator is out of the circuit? Or are you saying you're measuring the oscillator "bare" and not in a feedback loop? Do you have any other signal you can measure in the same frequency range? Is your spectrum analyzer up to the task? I'd think your 1kHz res bandwidth is pretty big to be seeing the noise level at 1kHz offset...I'd want a res bandwidth at most 1/10th the offset. And it might be interesting to look with much lower res bandwidth if you can, to see if there is a particular structure to the sidebands, or if they are indeed random noise. Cheers, Tom "Alex Collins" wrote in message ... "Harold E. Johnson" wrote in message news:_fwlc.22193$kh4.1279179@attbi_s52... What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB Thank you for the reply. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Even if the conections aren't ideal, I'd expect 10dB or so more noise, but it doesn't change however good/bad it's shielded. I'm probably doing something really stupid. Al. --- |
I'm not terribly familiar with the MiniCircuits VCOs, but I have been
involved with getting the phase noise down to low values on some of our designs. I can tell you that it's important to have power supplies be very clean, as well as to have the control voltage very clean (down into the microvolt region or below) as Harold says. But from what you wrote, I gather there's a problem even with things supposedly very clean. BTW, I don't 'zactly understand the "POS-900W out of circuit (ie. no PLL)" bit. What are you measuring if the oscillator is out of the circuit? Or are you saying you're measuring the oscillator "bare" and not in a feedback loop? Do you have any other signal you can measure in the same frequency range? Is your spectrum analyzer up to the task? I'd think your 1kHz res bandwidth is pretty big to be seeing the noise level at 1kHz offset...I'd want a res bandwidth at most 1/10th the offset. And it might be interesting to look with much lower res bandwidth if you can, to see if there is a particular structure to the sidebands, or if they are indeed random noise. Cheers, Tom "Alex Collins" wrote in message ... "Harold E. Johnson" wrote in message news:_fwlc.22193$kh4.1279179@attbi_s52... What could I be doing wrong? (My other POS-900W's do the same). Incidenatlly, I have a screenshot (23Kb) of the display which I can e-mail. Well, you don't specify the frequency coverage of your other VCO's, MCL's have a very high MHz/volt tuning rate and unless the tuning voltage is absolutely quiet, they'll look pretty bad. (BOTH the positive AND the negative leads shielded to ground?) W4ZCB Thank you for the reply. But surely if the Minicircuits datasheet says the trace will fall to less than -75dBC at 1,000Hz away I should see that, or something close to it? Irrespective of the frequency (800MHz)? Even if the conections aren't ideal, I'd expect 10dB or so more noise, but it doesn't change however good/bad it's shielded. I'm probably doing something really stupid. Al. --- |
My experience with Minicircuits is that they have EXTREME sensitivity to
the AFC line. I'd use a resistor divider from the same battery and a bypass cap. Something like a pair of 100K resistors and a 10 uF cap. Naturallly, ground the cap at the oscillator ground pin. ---------- Also, the Vcc has to be well bypassed AT THE OSCILLATOR. These things are very sensitive to any change in the VCC voltage. A battery is often not all that low of an impedance and so some good bypassing will help. Finally, how are you taking the signal off to your analyzer? I haven't used the POS-900 but in general, these critters want a paasive 50 ohm attenuator right on their output to avoid signals sneaking back and screwing up the oscillator. Jim Pennell N6BIU |
My experience with Minicircuits is that they have EXTREME sensitivity to
the AFC line. I'd use a resistor divider from the same battery and a bypass cap. Something like a pair of 100K resistors and a 10 uF cap. Naturallly, ground the cap at the oscillator ground pin. ---------- Also, the Vcc has to be well bypassed AT THE OSCILLATOR. These things are very sensitive to any change in the VCC voltage. A battery is often not all that low of an impedance and so some good bypassing will help. Finally, how are you taking the signal off to your analyzer? I haven't used the POS-900 but in general, these critters want a paasive 50 ohm attenuator right on their output to avoid signals sneaking back and screwing up the oscillator. Jim Pennell N6BIU |
I've also observed................it you are not using the auto mode on your
spectrum analyzer, meaning if you have too fast of a sweep rate for your I.F. bandwidth, you will see the results that you are running into. I use an HP 8558B, and if I set the SA improperly, I will see the same results. Try slowing down the sweep rate if you haven't done so. Pete "Jim Pennell" wrote in message ink.net... My experience with Minicircuits is that they have EXTREME sensitivity to the AFC line. I'd use a resistor divider from the same battery and a bypass cap. Something like a pair of 100K resistors and a 10 uF cap. Naturallly, ground the cap at the oscillator ground pin. ---------- Also, the Vcc has to be well bypassed AT THE OSCILLATOR. These things are very sensitive to any change in the VCC voltage. A battery is often not all that low of an impedance and so some good bypassing will help. Finally, how are you taking the signal off to your analyzer? I haven't used the POS-900 but in general, these critters want a paasive 50 ohm attenuator right on their output to avoid signals sneaking back and screwing up the oscillator. Jim Pennell N6BIU |
I've also observed................it you are not using the auto mode on your
spectrum analyzer, meaning if you have too fast of a sweep rate for your I.F. bandwidth, you will see the results that you are running into. I use an HP 8558B, and if I set the SA improperly, I will see the same results. Try slowing down the sweep rate if you haven't done so. Pete "Jim Pennell" wrote in message ink.net... My experience with Minicircuits is that they have EXTREME sensitivity to the AFC line. I'd use a resistor divider from the same battery and a bypass cap. Something like a pair of 100K resistors and a 10 uF cap. Naturallly, ground the cap at the oscillator ground pin. ---------- Also, the Vcc has to be well bypassed AT THE OSCILLATOR. These things are very sensitive to any change in the VCC voltage. A battery is often not all that low of an impedance and so some good bypassing will help. Finally, how are you taking the signal off to your analyzer? I haven't used the POS-900 but in general, these critters want a paasive 50 ohm attenuator right on their output to avoid signals sneaking back and screwing up the oscillator. Jim Pennell N6BIU |
Thanks for all the help.
Yes, you are right in the assumption that I've had it as 'clean' as is realistically possible. Yes, it's random noise. To explain, when the output of my board showed a 'triangle' as an output on the Spectrum Analyser, I first suspected by PLL, so I took the VCO module off the board and 'benched it up'. I used separate batteries for Vtune and Vcc supply, with very short (2cm) connections. I even put 1UF, 1N and 100N caps across the Vcc ** and Vtune ** pins and took the RF out on RG316 directly to the Analyzer sheilded an grounded with extreme care. When I still observed hideous phase noise, I even tried decoupling the batteries with a resistors/caps in each supply line. I also tried all sorts of pads using SMD resistors cleverly soldered to the output pin and case... So I'm convinced I've given the VCO a better environment than anyone could possibly achieve in reality (ie. even on the best PCB layout). The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. Perhaps if one of you guys would mind me sending you a screenshot, direct (23kB), that may shed some light on it? Al. ---- mail[underscore]me[at]freenet[dot]co[dot]uk |
Thanks for all the help.
Yes, you are right in the assumption that I've had it as 'clean' as is realistically possible. Yes, it's random noise. To explain, when the output of my board showed a 'triangle' as an output on the Spectrum Analyser, I first suspected by PLL, so I took the VCO module off the board and 'benched it up'. I used separate batteries for Vtune and Vcc supply, with very short (2cm) connections. I even put 1UF, 1N and 100N caps across the Vcc ** and Vtune ** pins and took the RF out on RG316 directly to the Analyzer sheilded an grounded with extreme care. When I still observed hideous phase noise, I even tried decoupling the batteries with a resistors/caps in each supply line. I also tried all sorts of pads using SMD resistors cleverly soldered to the output pin and case... So I'm convinced I've given the VCO a better environment than anyone could possibly achieve in reality (ie. even on the best PCB layout). The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. Perhaps if one of you guys would mind me sending you a screenshot, direct (23kB), that may shed some light on it? Al. ---- mail[underscore]me[at]freenet[dot]co[dot]uk |
Alex Collins wrote:
Thanks for all the help. Yes, you are right in the assumption that I've had it as 'clean' as is realistically possible. Yes, it's random noise. To explain, when the output of my board showed a 'triangle' as an output on the Spectrum Analyser, I first suspected by PLL, so I took the VCO module off the board and 'benched it up'. I used separate batteries for Vtune and Vcc supply, with very short (2cm) connections. I even put 1UF, 1N and 100N caps across the Vcc ** and Vtune ** pins and took the RF out on RG316 directly to the Analyzer sheilded an grounded with extreme care. When I still observed hideous phase noise, I even tried decoupling the batteries with a resistors/caps in each supply line. I also tried all sorts of pads using SMD resistors cleverly soldered to the output pin and case... So I'm convinced I've given the VCO a better environment than anyone could possibly achieve in reality (ie. even on the best PCB layout). The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. Perhaps if one of you guys would mind me sending you a screenshot, direct (23kB), that may shed some light on it? Al. ---- mail[underscore]me[at]freenet[dot]co[dot]uk Perhaps that's -75dBc at a bandwidth of 0.001Hz? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com |
Alex Collins wrote:
Thanks for all the help. Yes, you are right in the assumption that I've had it as 'clean' as is realistically possible. Yes, it's random noise. To explain, when the output of my board showed a 'triangle' as an output on the Spectrum Analyser, I first suspected by PLL, so I took the VCO module off the board and 'benched it up'. I used separate batteries for Vtune and Vcc supply, with very short (2cm) connections. I even put 1UF, 1N and 100N caps across the Vcc ** and Vtune ** pins and took the RF out on RG316 directly to the Analyzer sheilded an grounded with extreme care. When I still observed hideous phase noise, I even tried decoupling the batteries with a resistors/caps in each supply line. I also tried all sorts of pads using SMD resistors cleverly soldered to the output pin and case... So I'm convinced I've given the VCO a better environment than anyone could possibly achieve in reality (ie. even on the best PCB layout). The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. Perhaps if one of you guys would mind me sending you a screenshot, direct (23kB), that may shed some light on it? Al. ---- mail[underscore]me[at]freenet[dot]co[dot]uk Perhaps that's -75dBc at a bandwidth of 0.001Hz? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com |
In article ,
says... Thanks for all the help. The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. You're subtracting 30 dB (i.e., 10*log10(RBW)) to account for the 1-khz RBW, right? Also, to look at noise at a 1 kHz offset, you need a much narrower RBW filter. Looking at 1 kHz from the carrier through a 1 kHz filter, you will be only -6 dB down (or -3, depending on how your analyzer specifies its RBW) with a perfect signal! Take a look at the parameters in the graphs at http://www.qsl.net/ke5fx/synth.html -- I use a 100 Hz RBW filter to make measurements at 1 kHz from the carrier. Also, what other VCOs are you comparing to? A rule of thumb is that equivalent-quality VCOs tend to get about 6 dB noisier per octave of frequency. A very good 900-MHz VCO will have about the same noise profile as a crappy 90-MHz VCO.... and that's without even getting into the tuning-sensitivity question that others have raised. Finally, simple temperature-related drift (both short term and long term) will make it tough to measure anything 1 kHz from the carrier of a 900 MHz VCO. You're expecting almost one part per million stability from an unlocked VCO, and that isn't reasonable. To characterize phase noise at 1 kHz from the carrier, you should: 1) Lock the VCO with a PLL whose loop bandwidth is at least an order of magnitude lower (100 Hz or less) 2) Use an analyzer RBW at least an order of magnitude lower (100 Hz or less) 3) Perform this test only with a stabilized analyzer (otherwise, its internal LO's noise performance may be even worse than the VCO under test... something that's especially likely with lower-frequency VCOs) 4) Don't forget the 10*log10(RBW) adjustment factor. Those four steps will get you close, but to do the job right, you need two mo 5) Distinguish between AM and PM noise by using the VCO under test to drive a mixer's LO port to saturation, with a significantly-cleaner reference driving the other port of the mixer 6) Compensate for the noise-equivalent bandwidth of your analyzer's RBW filters. Typically this will make the true noise values 2-3 dB worse than the observed values. All 6 of these steps may be necessary to reproduce Mini-Circuits' catalog specification for the VCO. -- john ------------------------------------------------------ http://www.qsl.net/ke5fx Note: My E-mail address has been altered to avoid spam ------------------------------------------------------ |
In article ,
says... Thanks for all the help. The phase noise results seem unachievable, but I hardly think Minicircuits are lying, and that it should be pretty easy to achieve -75dBc at 1,000Hz away. Even if it was -60dBc I'd be happy. But all I know is I'm getting the same results on all POS-900Ws I have. I'm using an Anritsu MS2661C, and basically just hoping to see the traditional 'needle spike' at Fc. (which is what I see from other VCOs with the Spectrum Analyzer on the same settings). But, instead, it's more like a triangle than a needle, and nothing seems to change that. You're subtracting 30 dB (i.e., 10*log10(RBW)) to account for the 1-khz RBW, right? Also, to look at noise at a 1 kHz offset, you need a much narrower RBW filter. Looking at 1 kHz from the carrier through a 1 kHz filter, you will be only -6 dB down (or -3, depending on how your analyzer specifies its RBW) with a perfect signal! Take a look at the parameters in the graphs at http://www.qsl.net/ke5fx/synth.html -- I use a 100 Hz RBW filter to make measurements at 1 kHz from the carrier. Also, what other VCOs are you comparing to? A rule of thumb is that equivalent-quality VCOs tend to get about 6 dB noisier per octave of frequency. A very good 900-MHz VCO will have about the same noise profile as a crappy 90-MHz VCO.... and that's without even getting into the tuning-sensitivity question that others have raised. Finally, simple temperature-related drift (both short term and long term) will make it tough to measure anything 1 kHz from the carrier of a 900 MHz VCO. You're expecting almost one part per million stability from an unlocked VCO, and that isn't reasonable. To characterize phase noise at 1 kHz from the carrier, you should: 1) Lock the VCO with a PLL whose loop bandwidth is at least an order of magnitude lower (100 Hz or less) 2) Use an analyzer RBW at least an order of magnitude lower (100 Hz or less) 3) Perform this test only with a stabilized analyzer (otherwise, its internal LO's noise performance may be even worse than the VCO under test... something that's especially likely with lower-frequency VCOs) 4) Don't forget the 10*log10(RBW) adjustment factor. Those four steps will get you close, but to do the job right, you need two mo 5) Distinguish between AM and PM noise by using the VCO under test to drive a mixer's LO port to saturation, with a significantly-cleaner reference driving the other port of the mixer 6) Compensate for the noise-equivalent bandwidth of your analyzer's RBW filters. Typically this will make the true noise values 2-3 dB worse than the observed values. All 6 of these steps may be necessary to reproduce Mini-Circuits' catalog specification for the VCO. -- john ------------------------------------------------------ http://www.qsl.net/ke5fx Note: My E-mail address has been altered to avoid spam ------------------------------------------------------ |
| All times are GMT +1. The time now is 10:00 AM. |
Powered by vBulletin® Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
RadioBanter.com