Keyboard Shortcuts
ctrl + shift + ? :
Show all keyboard shortcuts
ctrl + g :
Navigate to a group
ctrl + shift + f :
Find
ctrl + / :
Quick actions
esc to dismiss
Likes
Search
S11 and S22 comparisons on 8753
|
Merry Christmas all!
My friends and I have been comparing what we see from the S Parameter Test Set on the two ports. These tests are with the VNA uncalibrated and with proper opens or shorts (even from APC7 cal kits). Typically we get completely different shaped curves ranging from -35 to -55 dB. I get very similar results using the same S Parameter Test Set on an HP8753B and an HP8753C so the curve shape is caused by the test set not the VNA. Typically we see one port 10 dB better and flatter than the other. Any comments? p.s. the cal kit discussion was very informative and timely. I still don't understand the comment (that was later corrected) about the open usually missing a part. My friend's Type N cal kit does not have extra parts for the open. Also I have seen an open/short combo advertised and wonder if this comes apart to use the open??? I have not used one and find the concept of one end a short (OK) and the other an open (how?) confusing. |
|
Correction! We are seeing this with a good load attached, no calibration. Same load gives different curve for each port.
toggle quoted message
Show quoted text
I would expect the reflected power to be much lower than -30 dBm. Of course with the open or short the reflected is close to 100%. Too much egg nog. --- In hp_agilent_equipment@..., "Peter Bunge" <bunge@...> wrote:
|
|
J. Forster
A 30 dB return loss is nothing to sneeze at.
toggle quoted message
Show quoted text
-John ===================== Correction! We are seeing this with a good load attached, no calibration. |
|
David Kirkby
On 23 December 2012 15:11, Peter Bunge <bunge@...> wrote:
Merry Christmas all!Merry Christmas My friends and I have been comparing what we see from the S Parameter Test Set on the two ports. These tests are with the VNA uncalibrated and with proper opens or shorts (even from APC7 cal kits).If you have not calibrated the VNA, then you are just seeing the uncorrect performance, and the fact that varies is no surprise. In fact, I'm a bit surprised you managed to get 35 dB uncorrected. There are a large number of systematic errors in a VNA. It does not matter, as the whole idea of the error correction is to remove those. I get very similar results using the same S Parameter Test Set on an HP8753B and an HP8753C so the curve shape is caused by the test set not the VNA.Don't worry about it. Typically we see one port 10 dB better and flatter than the other.It is only important once calibrated. The calibration removes the systematic errors. p.s. the cal kit discussion was very informative and timely. I still don't understand the comment (that was later corrected) about the open usually missing a part. My friend's Type N cal kit does not have extra parts for the open.It depends on what cal kit he has, but on for example and 85032B you will often see something like this: The small bit of gold played meter you see, is designed to increase the diamteer of the last few mm of the pin, so it is 50 Ohms, and not higher. If you think about it, to get as near as possible to an open, you want to have a 50 Ohm line that is cut off sharp at the end. Now consider what an open N male looks like. The male pin is thicker at the bottom. That bits makes 50 Ohm transmission line. Then the diameter is reduced for about 5 mm, so it can go in the female. That will not make a 50 Ohm transmission line, but a higher impedance line, due to the formula Zo=60 log(d_outer/d_inner) So an open N plug is not as close as possible to an open. Does that make sense? Hence the female part has an extender. I would add it is possible to calibrate a VNA with an open N connector without that. Accuracy suffers a bit, but that technique is used in Agilent's portable VNA's. A better bet is an shielded N. But again you need to know the paramters. Also I have seen an open/short combo advertised and wonder if this comes apart to use the open??? I have not used one and find the concept of one end a short (OK) and the other an open (how?) confusing.If you see an open/short advertised you MUST make sure you have the paramters on it. It will have an offset in ps, and three capacitance coefficents C0, C1, C2 and C3. You should then enter them in your VNA. If you don't have that information, the kit is effectivly useless. The short will have an offset too. The short might also have inductance paramters L0, L1, L2, and L3, but you can't enter them in the 8753 series, so don't worry about them. They are not important for the the frequency range of an 8753. I hope that makes a bit more sence. Dave |
|
On "missing parts": some opens(mostly early, and I'm specifically
toggle quoted message
Show quoted text
thinking of APC 3.5 connectors) consisted of a body having no center conductor attached, plus an insertable center conductor on a plastic rod. Without the inserted center conductor, the characteristics of the "open" would not match the values assumed in the calibration calculations. On "open / short combo": Neither standard has much length, so the two can easily be packaged into one assembly, strictly for convenience (less likely to be misplaced because it is larger). There's no coupling whatsoever between the two ends. Dave On 12/23/2012 7:11 AM, Peter Bunge wrote:
--
Clear Stream Technologies |
|
David Kirkby
On 23 December 2012 18:43, David McQuate <mcquate@...> wrote:
On "missing parts": some opens(mostly early, and I'm specificallyWas sort of like this though for 3.5 mm, not N as in this diagram? My own 3.5 mm kit does not have anything like this - the standards are self contained - no extra bits needed, and nowhere to fit anything. On "open / short combo": Neither standard has much length, so the twoTrue - they are convenient. There are 3 types of these 1) With just the open and short. 2) With open, short and load 3) With open, short, load and thru. Eiether way, one needs to have the data on the bits. If you read the 8753 manual, you will find the calibration kit it is deisgned to work with is the HP 85032B (as in the photo I showed). Any other cal kit has to be entered as a user calibration. Dave |
|
On the 8752/8753/8720 etc., and pretty much all newer VNAs, even what you call "uncalibrated" is actually calibrated at the factory or at the service center when you send the analyzer for calibration. The problem is that you don't know what cables were used for the factory calibration. Hence, two analyzers that were calibrated with different cables will show different results giving the impression that the hardware is vastly different. I think, the 8510 was the last VNA which showed true uncorrected hardware performance when the instrument wasn't calibrated. And it was impressive!
This factory calibration is in a way similar to the user calibration, but the correction constants are saved in non volatile memory and protected with a jumper on the processor board. This then is the default state of the VNA when it powers up without a user calibration. A VNA with 35dB return loss is perfectly usable for most work. Realistically, the accuracy improvements for most work are very small as you get past that range. The log scale of the return loss measurement makes it easy to forget that going from 40 to 50 dB return loss requires heroic efforts for an improvement that is tiny in absolute terms. Vladan --- In hp_agilent_equipment@..., David Kirkby <david.kirkby@...> wrote: If you have not calibrated the VNA, then you are just seeing the...... Dave |
|
David Kirkby
On 23 December 2012 22:09, pianovt <pianovt@...> wrote:
On the 8752/8753/8720 etc., and pretty much all newer VNAs, even what you call "uncalibrated" is actually calibrated at the factory or at the service center when you send the analyzer for calibration.Is that the same data you are supposed to back up to a floppy? The reason I ask, is that the manuals seem to imply this will make restoring the data much quicker if you have it on a floppy in the event of a failure. But you are suggesting it would be changed when the unit went for cal, suggesting to me there's not much point having it on a floppy. Or perhaps I'm wrong. The problem is that you don't know what cables were used for the factory calibration. Hence, two analyzers that were calibrated with different cables will show different results giving the impression that the hardware is vastly different.In the case of the 8720D, there is a set of cables available (Agilent t 85131F) for this, with the NMD connectors, to ensure a rugged connection to the VNA. I assume it would be calibrated with them. I know the length of them is nominaly 24.5", but at 20 GHz, with a wavelength of 15 mm, slightly differnt cables would give quite different phases. I think, the 8510 was the last VNA which showed true uncorrected hardware performance when the instrument wasn't calibrated. And it was impressive!I assume several of the internal cables were selected to give it this performance. Did the 8510 have a computer inside to do automatic error correction, or was it a manual process? I've used a few VNAs, but never and 8510. I own an 8720D and 8753A. This factory calibration is in a way similar to the user calibration, but the correction constants are saved in non volatile memory and protected with a jumper on the processor board. This then is the default state of the VNA when it powers up without a user calibration.At least on my 8720D, I'm sure it is nowwhere near 55 dB as one of the VNAs reported in this thread. I'm not sure it is even 35 dB. But I always do a user cal. I'm sitting in bed at this minute with the laptop - it has just gone midlight local time. Hence I can't be bothered to go downstairs and switch the VNA on to see what it shows. VladanDave |
|
David Kirkby
On 23 December 2012 15:11, Peter Bunge <bunge@...> wrote:
p.s. the cal kit discussion was very informative and timely. I still don't understand the comment (that was later corrected) about the open usually missing a part. My friend's Type N cal kit does not have extra parts for the open.The cal kit the 8753 is supposed to work with is the 85032B - see the 8753 manual. The 85032B needs the extender, as I showed. So unless your friend has a different cal kit, and has entered the data as a user calibration, he is not getting the best accuracy. If he has an 85032B, without the extender for the female N, all the data from that device will be wrong. Dave |
|
Dave,
Is that the same data you are supposed to back up to a floppy? TheYes, backing up on floppy will save those calibration constants (and other stuff as well). That way, if you have to replace the processor board, you don't have to recalibrate everything. You can replace the board in a few minutes and restore the constants. No need to send the analyzer in. In the case of the 8720D, there is a set of cables available (AgilentYou can do the internal calibration with those cables, but also with other cables or omit one of the cables altogether. Whatever you pick as your default cable setup will then become the calibrated state when you power up the analyzer. I assume several of the internal cables were selected to give it thisThe 8510 had top notch hardware and relied on it for pretty good raw performance at the plane of the test set ports on the front panel. So, if you connect something to either port directly, you can get a decent measurement of return loss even without a cal. For S21 and S12, you need a test cable, and that cable is an unknown until you do a calibration which is similar to the user cal on your 8720. In the back of the test set, you can attach a pair of delay lines that are hopefully of similar length to your test port cables. That's as close as you can get on the 8510 as a default state. The 8753 and 8720 are newer designs, and they made more use of digital correction and non-volatile memory, thereby allowing for less expensive hardware. The main provision is that you need stable hardware, and that is the case. At least on my 8720D, I'm sure it is nowwhere near 55 dB as one of theIf you were to send the analyzer for a "Internal Calibration" (this is the factory/service center job) and you asked them to use your supplied cables and calibration standards, in theory you would get the analyzer to have perfect calibration when you turn it on and test it with your cal standards. You would not need to do the manual calibration. In practice, connectors are not perfectly repeatable, and you might see something like 40-50 dB of return loss or even better, depending on repeatability. However, remember that your measurement is only comparing the unknown against your calibration standard. Hence, seeing 60dB of return loss only means that the unknown is almost identical to the calibration standard you used. In practice, you could have calibrated the analyzer with a poor load that is 45 Ohms instead of 50 and the analyzer would "think" that a 45 Ohm unknown is a perfect 50 Ohms. The "internal calibration" procedure is in the service manual for the 8753, and I assume that it's also in the manual for the 8720. I don't have an 8720, so I am speculating. If you find it, you will see that a portion of the procedure is quite similar to the regular cal that you do all the time. Vladan |
|
Hi:
toggle quoted message
Show quoted text
The most accurate way of characterizing S11 or S22 is to use a scalar network analyzer with a bridge and precision air line. The cal load is a precision load that's just slightly off from 50.0 Ohms but how much off doesn't matter. It's far more accurate than a VNA. The key idea is that the air line is exactly 50.0 Ohms and the mismatch causes ripple above and below that so the center of the ripples is where the location of 50.0 Ohms. Have Fun, Brooke Clarke hp_agilent_equipment@... wrote: Re: S11 and S22 comparisons on 8753 |
|
David Kirkby
On 24 December 2012 19:23, Brooke Clarke <brooke@...> wrote:
Hi:I don't claim to know much about scaler network analyzers, so what I say might be wrong. I don't see how you can consider a *scaler* network analyzer able to characterise S11 and S22, since both S11 and S22 require both magnitude and phase information. I believe the things you refer which have an impedance close to, but not exactly 50 Ohms are generally known as sliding loads, though I suppose they might have other names. I do have a few precision airlines which are machined very accurately. I've got a couple of 50 Ohm ones, and a couple which are stepped from 50 Ohms to 25 Ohms. These are part of VNA *verification* kits, so they need to be very accurate. I understand the weak point of calibration kits is the loads - hence sliding loads are used at higher frequencies. Dave |
|
laurens_db
A well calibrated VNA is the best way of measuring S11/ S22.
toggle quoted message
Show quoted text
A broadband cal kit load is as close to 50Ohm as HP can make it. A slight mismatch at the end of an air line will allow you to normalise the return loss measurements, but not a full correction (the port match and directivities are not resolvable with this method IIRC) --- In hp_agilent_equipment@..., Brooke Clarke <brooke@...> wrote:
|
|
David Kirkby
On 24 December 2012 21:11, laurens_db <laurens101@...> wrote:
A well calibrated VNA is the best way of measuring S11/ S22.That was my belief too - at least in the context of typical test equipment. I would not be surprised if there other techniques applicable to standards labs, which are not too practical for commerical or amateur use. I just opened this book: and looked up scalar network analyzers. By the very title of the book, you can see it is mainly devoted to VNAs, though there is a bit of discussion about scalar network analyzers. To quote: "Scalar network analyzers has the attribute of being very simple to use, with almost no calibration or setup required. The scalar network analyzers were designed to be quite flat in frequency response, and a typical system consisted of one and the input and one at the output of the DUT. However, for measurements of input or output match, or impedance, the scalar network analyzer relied on a very high quality coupler or directional bridge. If there was any cabling, switching or other test system between the bridge and the DUT, the composite matches of ALL were measured. There was no additinonal calibration possible, to remove the effects of the mismatch. As test systems became more complex and integrated, scalar network analyzers started to fall from favor and there are virtually none sold today by commercial instrument manufacturers" As far as I can tell, the error correction of a VNA offers the ability to compensate for errors than the scalar network analyzer simply does not. A broadband cal kit load is as close to 50Ohm as HP can make it.Yes, although because of the limitations of broadband loads, sliding loads are often used at higher frequencies. To once again quote from Joel Dunsmore's book: "The load standard is usually the most difficult to produce. <snip lots> The sliding load, which should more properly be called a sliding mismatch, is constructed from lengths of precision airline, The centre conductor of the airline is typically created in such a way that it can slide into place while the outer conductor is not yet mated, to allow a beadless connection. The load element is typically not resistive element, but is more commonly a tapered bead of lossy material, that essentially makes the airline look like a lossy element. It is designed to have an impedance which is not quite 50 Ohms, normally in the range of 26-40 dB return loss. <snip> "As the sliding load is moved, so its apparent impedance rotates around the Smith Chart" There's a diagram showing a full circle on the Smith chart, almost, but not quite in the centre.It slightly spirals in, as the frequency (and so loss) is incresed. "The difference between the computed centre and the actual centre of the Smith Chart determines the directivity error term" I think is should be obvious what I stated earlier, that sliding loads are impractical at lower frequencies. Looking at the manual for the 85054A 'N' cal kit, the minimum frequency of the sliding load is 1.999 GHz. At 100 MHz (lambda = 3m), the load would need to be 1.5 m long to get a complete half-wave on a Smith Chart. I suspect you can get away with less than a full circle, as it only needs 3 points to make a circle, but I think accuracy would suffer a lot if you could not get an appreciable part of a circle. Dave |
|
J. Forster
You can use a SNA for quick and dirty tests, comparing a known, high
toggle quoted message
Show quoted text
quality, component to the DUT, especially if you have a Storage Normalizer, but in no way is an SNA a substitute for a VNA. The only regime where they are used today, AFAIK, ia at frequencies above those practical for VNAs, over 40 GHz or more. FWIW, -John ================= On 24 December 2012 21:11, laurens_db <laurens101@...> wrote:A well calibrated VNA is the best way of measuring S11/ S22.That was my belief too - at least in the context of typical test |
|
David Kirkby
On 25 December 2012 00:24, J. Forster <jfor@...> wrote:
You can use a SNA for quick and dirty tests, comparing a known, highAgreed. That is why I was surprised when someone said a scalar one was more accurate. The only regime where they are used today, AFAIK, ia at frequencies aboveAgilent sell a VNA for 110 GHz. At $691,437 for a 4-port model, I don't expect Agilent sell too many of them! If you buy all the options, it will be over $800,00. Dave |
|
J. Forster
Agreed.
toggle quoted message
Show quoted text
Note that the 8754A(?) computer based VNA that went to 18 GHz cost prehaps $80,000 in 1970 (a guess). Probably more w/ options. That's $450,000 in today's deflated dollars, so $700,000 is not such a bad deal. I don't have quite $800k to blow this week after buying Christmas presents. LOL. -John ==================== On 25 December 2012 00:24, J. Forster <jfor@...> wrote:You can use a SNA for quick and dirty tests, comparing a known, highAgreed. That is why I was surprised when someone said a scalar one was |
|
Peter Gottlieb
And I'm sure the thing is completely unrepairable except by Agilent, schematics are not available, and once it goes out of support that is basically the end of that.
toggle quoted message
Show quoted text
On 12/24/2012 8:38 PM, J. Forster wrote:
|
|
J. Forster
Worse. It's likely chock full of custom MMIC's, sole sourced by Agilent.
toggle quoted message
Show quoted text
The stuff ceased to be repairable probably more than 20 years ago. -John ================== And I'm sure the thing is completely unrepairable except by Agilent, |
to navigate to use esc to dismiss