|
|
|
On 06/02/2024 15:07, Mikek wrote: Has anyone made a model of "common base transformer feedback (CBTF) amplifiers, and also known as common base noiseless feedback (CBNF) amplifiers"?
As described in Dallas Lankford's paper.
https://web.archive.org/web/20070226191335/http://www.kongsfjord.no/dl/Amplifiers/Ultralinear%202N5109%20And%202N3053%20Amplifiers.pdf
With special attention to the unknown characteristics detailed in the last half of page 2. Yes, I'm sure many people have. These are often used for ultra-low phase noise amplifiers. They were originally used in Anzac hybrid amplifiers in the 1970-80s. All the relevant patents expired long ago. Ulrich Rohde came up with something similar that worked in common emitter. Did you have a particular question? -- Regards, Tony
|
|
|
On Tue, Feb 6, 2024 at 07:46 AM, Tony Casey wrote:
Yes, I'm sure many people have. These are often used for ultra-low phase noise amplifiers. They were originally used in Anzac hybrid amplifiers in the 1970-80s. All the relevant patents expired long ago. Ulrich Rohde came up with something similar that worked in common emitter.
Did you have a particular question?
Curious about the feedback transformer ratio, that actually gives you a near-perfect two way impedance match. As Lankford queried, "It appears that the turns ratios for a near-perfect two way impedance match to Z0 have yet to be determined for 2N5109 and 2N3053 BJT’s" That was written 30 years ago, I wonder if there are answers? In the article he does say, the ideal feedback transformer ratios, have the input impedance 60% to 80% of the load impedance. Mikek
|
I have built several Norton amplifiers for HF using MRF-136 MOSFETs. I find the input impedance reflects the load impedance which can be an issue if the amplifier is feeding a filter. A write-up of my experiments is here: http://w6dsr.com/HF_LNA/HF_LNA.pdf. After seeing the Rohde HAM Radio article, I tried to model his Fig 4 schematic in LTspice, but this is the first spice simulation I attempted, so its implementation might be quite amateurish. I could upload that attempt if requested. I do not know how to use LTspice to measure the IP3 of a circuit, which was the original reason I modeled it. Rohde claims an IP3 of 70 dBm for a push-pull version of his Fig 4 topology, and I wish he had included a full schematic for that design as I don’t believe it. If he meant the IP3 sidebands were 70 dB down, then an IP3 of 35 dBm is more believable. My designs only achieved ~60 dBm IP3 with a push-pull topology. Feed-Forward designs can achieve higher IP3s, but with much added complexity. -Doug, W6DSR
From: LTspice@groups.io [mailto:LTspice@groups.io] On Behalf Of Mikek
Sent: Tuesday, February 6, 2024 09:41
To: LTspice@groups.io
Subject: Re: [LTspice] Norton Noiseless feedback amplifiers. Has anyone modeled them. On Tue, Feb 6, 2024 at 07:46 AM, Tony Casey wrote: Yes, I'm sure many people have. These are often used for ultra-low phase noise amplifiers. They were originally used in Anzac hybrid amplifiers in the 1970-80s. All the relevant patents expired long ago. Ulrich Rohde came up with something similar that worked in common emitter.
Did you have a particular question?
Curious about the feedback transformer ratio, that actually gives you a near-perfect two way impedance match.
As Lankford queried,
"It appears that the turns ratios for a near-perfect two way impedance match to Z0 have yet to be
determined for 2N5109 and 2N3053 BJT’s"
That was written 30 years ago, I wonder if there are answers?
In the article he does say, the ideal feedback transformer ratios, have the input impedance 60% to 80%
of the load impedance.
Mikek
|
Doug wrote, "I do not know how to use LTspice to measure the IP3 of a circuit, which was the original reason I modeled it."
The most straight-forward way is to run a .TRAN simulation, then plot the FFT and pick out the frequencies you wish to measure. There are warnings about getting good FFT data. There may be a substantial learning curve for that.
As a start, plot a waveform, then right-click in the plot window and choose View > FFT, select the waveform (if not already selected), and click OK.
Andy
|
Thank you, I shall try that. It will be interesting to see how the modeled Norton feedback amplifier compares with the as built amplifier I use every day.
From: LTspice@groups.io [mailto:LTspice@groups.io] On Behalf Of Andy I
Sent: Tuesday, February 6, 2024 12:24
To: LTspice@groups.io
Subject: Re: [LTspice] Norton Noiseless feedback amplifiers. Has anyone modeled them. Doug wrote, "I do not know how to use LTspice to measure the IP3 of a circuit, which was the original reason I modeled it."
The most straight-forward way is to run a .TRAN simulation, then plot the FFT and pick out the frequencies you wish to measure. There are warnings about getting good FFT data. There may be a substantial learning curve for that.
As a start, plot a waveform, then right-click in the plot window and choose View > FFT, select the waveform (if not already selected), and click OK.
Andy
|
On Tue, Feb 6, 2024 at 11:11 AM, Epignathus wrote:
I find the input impedance reflects the load impedance which can be an issue if the amplifier is feeding a filter.
Hi Epignathus, From you paper, "The input impedance is totally reflected by the output impedance, and therefore it is imperative that the output of the amplifiers be terminated in 50 Ohms." This gets at another point I'm looking at. In Dallas Lankford's paper, https://web.archive.org/web/20070226191335/http://www.kongsfjord.no/dl/Amplifiers/Ultralinear%202N5109%20And%202N3053%20Amplifiers.pdf (Note: I fixed the link.) He wrote,
“The two way impedance match to Z0 is an ideal characteristic of BJT CBTF amps based on a mathematical
derivation which uses simplifying assumptions that are not true for actual common base amps”
He then concludes with,
Based on measurements with several BJT CBTF amps, I have found that if such an amp works into a
load of Z0 ohms real, then the input impedance tends to be about 60% to 80% of Z0.
I'm working with a 100Ω system, in and out. So, I'm contemplating a 1 to 1 ratio on the output, then per Lankford's statement, that would reflect back as 60Ω to 80Ω to the input, so I'm thinking a 1.2 to 1 ratio ( 100Ω to 70Ω) transformer on the input.
|
I found with a winding ratio of 1:n:m of 1:1:2 (which only yields a gain of 6 dB, but the best IMD), that if I terminated the amplifier with 50 Ohms resistive, my HP VNA measured S11 (as SWR) 1.06:1 at 1.5 MHz and 1.05:1 at 50 MHz. I did not see that 60% to 80% of Z0, but note that my TLTs were all made with RG-178 50 Ohm coax, where Lankford (and others) have used regular enameled copper wire. If you are shooting for 100 Ohms, then you should wind your TLT with 100 Ohm impedance twisted transmission line. I found that wire-wrap wire with a tight twist would yield a Z0 of around 100 Ohms. So you might first try your TLTs with 100 Ohm Z0, and maybe avoid another passive element. By-the-way, I’m still looking for a circuit topology with an input, an output, unity gain in-to-out, an input Z of (50 +j0), a decent S22, good isolation from input to output, and with an IP3 of >60 dBm. Not sure I said that right, but essentially some circuit that will present a 50 Ohm match to the amplifiers without degrading the IP3. Presently I use a resistive attenuator between the Norton amplifiers and on their output to try and achieve a 50 Ohm match before entering my bandpass filters. A diplexer would be suitable, but I would need many to cover the HF bandwidth. -Doug Ronald, W6DSR
From: LTspice@groups.io [mailto:LTspice@groups.io] On Behalf Of Mikek
Sent: Tuesday, February 6, 2024 12:54
To: LTspice@groups.io
Subject: Re: [LTspice] Norton Noiseless feedback amplifiers. Has anyone modeled them. On Tue, Feb 6, 2024 at 11:11 AM, Epignathus wrote: I find the input impedance reflects the load impedance which can be an issue if the amplifier is feeding a filter.
Hi Epignathus,
From you paper,
"The input impedance is totally reflected by the output impedance, and therefore it is imperative that the
output of the amplifiers be terminated in 50 Ohms."
This gets at another point I'm looking at.
In Dallas Lankford's paper,
https://web.archive.org/web/20070226191335/http://www.kongsfjord.no/dl/Amplifiers/Ultralinear%202N5109%20And%202N3053%20Amplifiers.pdf
(Note: I fixed the link.)
He wrote, “The two way impedance match to Z0 is an ideal characteristic of BJT CBTF amps based on a mathematical derivation which uses simplifying assumptions that are not true for actual common base amps” He then concludes with, Based on measurements with several BJT CBTF amps, I have found that if such an amp works into a load of Z0 ohms real, then the input impedance tends to be about 60% to 80% of Z0.
I'm working with a 100Ω system, in and out. So, I'm contemplating a 1 to 1 ratio on the output, then per Lankford's statement, that would reflect back as 60Ω to 80Ω to the input, so I'm thinking a 1.2 to 1 ratio ( 100Ω to 70Ω) transformer on the input.
|
On Tue, Feb 6, 2024 at 01:54 PM, Epignathus wrote:
I did not see that 60% to 80% of Z0, but note that my TLTs were all made with RG-178 50 Ohm coax, where Lankford (and others) have used regular enameled copper wire.
I wonder if it is the coax vs twisted pair OR, transistors vs your FETs, or bias difference with FETs that made output reflect 100% back to the input? Or some of all? Mikek
|
I’ve built many of these Norton style amplifiers originally with 2N5109s before graduating to the high-power MOSFETs, and have never seen that input mismatch. Note that all my TLTs have been on binocular cores, maybe that matters… -Doug
From: LTspice@groups.io [mailto:LTspice@groups.io] On Behalf Of Mikek
Sent: Wednesday, February 7, 2024 11:39
To: LTspice@groups.io
Subject: Re: [LTspice] Norton Noiseless feedback amplifiers. Has anyone modeled them. On Tue, Feb 6, 2024 at 01:54 PM, Epignathus wrote: I did not see that 60% to 80% of Z0, but note that my TLTs were all made with RG-178 50 Ohm coax, where Lankford (and others) have used regular enameled copper wire.
I wonder if it is the coax vs twisted pair OR, transistors vs your FETs, or bias difference with FETs that made output reflect 100% back to the input?
Or some of all?
Mikek
|
I found another Norton amplifier from Clifton Labs, link here: https://www.okdxf.eu/files/Z10040B%20Manual%201r7.pdf on page 43 he plots VSWR where the plot indicates a VSWR of 1.3 at 1 MHz. There is no plot of |Z|, but his design may be seeing that 60% to 80% of the termination Z0.
From: LTspice@groups.io [mailto:LTspice@groups.io] On Behalf Of Mikek
Sent: Wednesday, February 7, 2024 11:39
To: LTspice@groups.io
Subject: Re: [LTspice] Norton Noiseless feedback amplifiers. Has anyone modeled them. On Tue, Feb 6, 2024 at 01:54 PM, Epignathus wrote: I did not see that 60% to 80% of Z0, but note that my TLTs were all made with RG-178 50 Ohm coax, where Lankford (and others) have used regular enameled copper wire.
I wonder if it is the coax vs twisted pair OR, transistors vs your FETs, or bias difference with FETs that made output reflect 100% back to the input?
Or some of all?
Mikek
|
On Wed, Feb 7, 2024 at 12:59 PM, Epignathus wrote:
and have never seen that input mismatch
OK, I'll drop this area of questioning until I can measure it. I have two Norton amps, one is on my antenna, the other, which I had in a phasing unit for 28 years, then, when I removed it from the phaser and tried to use it on the antenna, suddenly it doesn't work properly. So, I'm troubleshooting the bad push/pull Norton amp, I have removed the output transformer and put a 50Ω load on each amp so I can compare one amp to the other. One amp has about 2 times the the output of the other. The transistors measure correctly, ALL DC voltage are the same between amps. The signals are ~ 180* out of phase?? Looks to be about 170*-- seems odd. I'm getting pretty far off the LTspice topic, so before I get moved off, I'm moving over to https://ham-antennas.groups.io/g/mainI'll work today, to see if I can get my amp working properly, I hope to have more to say after about the ratio of the output load reflected back as the input impedance after I get it repaired. If I do, I'll post it here. Thanks, Mikek Private emails Welcomed. I'm in the directory.
|
On Thu, Feb 8, 2024 at 05:06 AM, Mikek wrote:
I hope to have more to say after about the ratio of the output load reflected back as the input impedance after I get it repaired.
If I do, I'll post it here.
Thanks, Mikek
Private emails Welcomed. I'm in the directory.
Ok, I got my second Norton amp working. I did some simple testing to see what the input resistance looks like compared to what you load the output with. i.e. if you put 50Ω on the output, does it make the input look like 50Ω. They way I measured: Connect my scope to the (50Ω) signal generator output lead, set signal generator so the scope reads 8mV. Now, attach the SG output lead to the input of the amp. Now read the scope, to see how much it dropped, then calculate what the Rin is. It doesn't act anything like described in Lankford's paper. 50Ω load on the output calculates to 54Ω, OK. Now, if I put a 100Ω load on the amp, repeat as above, now the input looks like 68Ω. Hmm, this seems in line with Lankford. Now, since I will be driving the amp with a 100Ω source, I adjust the SG to 100Ω output impedance, and did the test again. Now, with 100Ω load on the output and 100Ω input resistance the measurement says 139Ω input impedance. So changing the resistance of the source, changes the input resistance! Here is a drawing with all the measurements and the math. I hope this makes it clearer. https://www.dropbox.com/scl/fi/gj3mjc8iciyp1gc969z7p/Norton-Input-Output-impedance.jpg?rlkey=2mxbl4f8421fpk9x1tg59iqx2&dl=0 Mikek
|
Mikek wrote, "So changing the resistance of the source, changes the input resistance!"
As I'm sure you know, that can't happen. If it does, it means the measurement is wrong.
Please don't use dropbox or any other third-party websiite for showing us photos or files. Photos should be uploaded to Photos. Non-photo files should be uploaded to Files > Temp.
Your method for estimating the input impedance is not good. It may be OK if the circuit under test and the signal generator are 100% linear and if the input impedance is purely real (resistive). Otherwise it is no good. Have you mentioned what frequencies are involved? Did you account for the input impedance (resistance and capacitance) of the scope and its probe? Was it an active probe, or a cable?
This would be a really good thing to simulate! Have you attempted to do LTspice simulations of the circuit? They might have limited accuracy, of course; but it could be better than nothing and might give you a clue about what's going on.
Andy
|
On Sun, Feb 11, 2024 at 10:56 PM, Andy I wrote:
Please don't use dropbox or any other third-party website for showing us photos or files. Photos should be uploaded to Photos. Non-photo files should be uploaded to Files > Temp.
Most of the groups I use, have a link to post in post pictures. When I couldn't find it, I just used dropbox. Now that I see how it can be done, I'll use the upload feature.
Your method for estimating the input impedance is not good. It may be OK if the circuit under test and the signal generator are 100% linear and if the input impedance is purely real (resistive). Otherwise it is no good. Have you mentioned what frequencies are involved? Did you account for the input impedance (resistance and capacitance) of the scope and its probe? Was it an active probe, or a cable?
I would think non-linearity is not the problem, but I did assume there was not a lot of reactance, i.e. it was a mostly resistive input. I tested at 1MHz, the capacitance reactance of the 15pf probe is so much higher than 50Ω at 1MHz, I didn't think it would make much difference. >This would be a really good thing to simulate! Have you attempted to do LTspice simulations of the circuit? They might have limited accuracy, of course; but it could be better than >nothing and might give you a clue about what's going on. No, I don't have Ltspice and I don't want to go through the learning curve for this project. I'll find another way to do the testing. I did know after I ask the initial question I got off in the weeds as to the intent of this group. So I will move to a different group, If you know of a group where this topic would be a better fit, please advise. I have posted this to https://groups.io/g/loopantennas/topic/104190619#18570so if other still have interest please follow their, although I found more interest in the subject here. Thanks for all the help, Mikek Private emails Welcomed. I'm in the directory.
|
Mikek wrote, " Most of the groups I use, have a link to post in post pictures." Yes, that is what you have in this group too. Did you read the instructions on the group's main webpage? It tells you what to do: Pictures should go here - https://groups.io/g/LTspice/photos . Create a New Album, then Add Photo to upload your photo into it.and: Please do not point us to other ("third party") file storage websites, even if it is one where you have an account. re: " I did assume there was not a lot of reactance, i.e. it was a mostly resistive input." That is quite possibly not true. If the SWR was, let's say, about 1.3:1, that could be a significant reactive (non-resistive) component. Unless the circuit includes a lossy pad at the input terminals, it is wise to assume that the impedance may have a significant phase angle. The conflicting results with your measurements might be evidence of that. re: " I tested at 1MHz, the capacitance reactance of the 15pf probe is so much higher than 50Ω at 1MHz, I didn't think it would make much difference." How reliable is that "15pF" number for the scope probe? Is that the input impedance of the scope itself, which would be measured at its BNC jack with zero cable? Did you use an active scope probe with 15 pF as its input capacitance spec? Was it a passive probe with a meter of cable? Was it a X10 probe? Was it a X1/X10 probe, and in which setting? I have little to go on here, but it is quite possible that the effect of the scope was much greater than the equivalent of a 15 pF capacitor. One of the things EEs learn when they first go into the lab, is that scope probes never look like infinitely high impedance. Probes always affect the circuit to which they are connected, often significantly. Some of it is or should be expected (e.g., the capacitance of the probe's cable, or the large capacitance of a X1/X10 probe in the X1 setting), but some might not be (e.g., the loss component of a passive probe at RF frequencies, or the loop inductance with the ground lead). re: " No, I don't have Ltspice and I don't want to go through the learning curve for this project." Hmm. Then I need to suggest that this is not the place. I don't want to make you leave this forum, but this is a group about LTspice so it has to come down to LTspice eventually. On the other hand, I feel that discussions about these amps, WITH SIMULATIONS, are important here and should not be turned away. You don't have LTspice? Why not? It is really darned easy to get it and usually quite easy to install it. Well, for the majority of users. (A few users do seem to have installation problems, but that is because their PC has unusual features, or they are installing LTspice on top of a previous version and things didn't go the way they expected.) For the majority of users, it is a breeze to install and start simulating. There is some learning curve with every new tool, but I want to say that LTspice's learning curve to get going, is really not bad. You have a couple of oddities in your circuit (e.g., the transformer, and maybe the particular transistors you used), but they are definitely not overwhelming, and many people in this group can help. That's why we are here. More people, like you, should learn how to use LTspice. This group is somewhat weighted more heavily with Amateur Radio operators, so lots of hams are comfortable using LTspice. " I have posted this to https://groups.io/g/loopantennas/topic/104190619#18570" I did see your question last week in that and I think another group too. (I do not visit them as often as I do this one.) The loopantennas group is mainly about antennas, not amps, but there are one or two preamp experimenters there too so maybe you will find help there. I am a member of so many groups I can't keep up with them all. I keep cutting back, only to find that I am in well over 100 groups and it's still growing. (So much to do, not enough time!) Andy
|
On 12/02/2024 14:15, Mikek wrote:
No, I don't have
Ltspice and I don't want to go through the learning curve for this
project.
I'll find another way to do the testing.
I did know after I ask the initial question I got off in the
weeds as to the intent of this group. So I will move to a
different group,
If you know of a group where this topic would be a better fit, please
advise.
I have posted this to https://groups.io/g/loopantennas/topic/104190619#18570
so if other still have interest please follow their, although I
found more interest in the subject here.
Sad you couldn't find the motivation to learn and use LTspice.
But, if this has raised the curiosity of anyone else following this
thread, I have uploaded an Example
Schematic of a Norton feedback amplifier for anyone to play
with. I have experience of using these from years ago, and they are
capable of amazing performance. The design I have uploaded should be
capable of 1MHz to at least 300MHz bandwidth with the transistor
used. To be fair, I have not really added "reasonable" parasitics to
the transformer - much will depend on the construction. I usually
used to use twin-holed ferrite beads (we called them pig's nose
beads). Toroids should be good to, but use the smallest one you can
get away with.
--
Regards,
Tony
|
Andy, I noted, accepted and said I how I would change to post picture correctly.
Why the need to chastise me for the second time in the very next post after I did this.
Mikek
|
Hi, Trying to find this schematic in the files library. Can't seem to find, have had a quick look in the archive too. Do you still have a copy?
Many thanks,
James
|
Tony Casey
Epignathus