Butterworth bandpass filter question
I am trying to scale an existing front end receiver (butterworth
bandpass) filter to a different frequency range. Unfortunately, it has a transformer in the original design, so I'm stuck. I also don't know how to handle the load presented by the active front end component other than it's probably not significantly reactive. The existing filter is for a 7 Mhz receiver, I'd like to have a similar filter design for 50 to 200 Khz. The filter components and transformer winding details are in the document at: http://www.amqrp.org/kits/softrock40...0Assy%20v1.pdf The input chip is an FST3126, spec sheet is at: http://www.fairchildsemi.com/ds/FS/FST3126.pdf The spec for the T30-2 transformer core is at: http://partsandkits.com/T30-2.htm I have aade filter design software, but it isn't allowing me to plug in the transformer into the design page of the software....so, I need to know it's equivalent circuit I think. The transformer winding details are on page 4 of the document and the schematic of the front end is page 9. Ultimately, I need new values for L1, L2, C20, C21 and C22. If someone can give me a reasonable guess as to the equivalent circuit of the transformer and the IC (U5), I can do the rest of the job myself using the aade filter software. Thanks, T |
In article , TRABEM
says... If someone can give me a reasonable guess as to the equivalent circuit of the transformer and the IC (U5), I can do the rest of the job myself using the aade filter software. Thanks, T You might try FilDes: ftp://ftp.lehigh.edu/pub/listserv/qr...ols/fds201.zip It will let you specify different source and load impedances for the filter being designed, so you can probably get away without the transformer altogether. The AADE software has a nicer interface but I've always liked the functionality in FilDes better. -- jm ------------------------------------------------------ http://www.qsl.net/ke5fx Note: My E-mail address has been altered to avoid spam ------------------------------------------------------ |
It looks like the filter is designed for 50 ohms in and out. The transformer
inductance is also a part of the filter. It's value should be about the same as L1 so to use it at 50k to 200K, you'll need to rewind for a lot more turns on the primary and both secondaries if you keep the same core material. Scaling it all and then optimizing for Coil Q of 40 and regular cap and coil values, I get new values like: L1 680nH - 100uH C1 470pF - .022uF C21 180pF - .027uF L2 1.8uH - 100uH C22 470pF - .022uF T1 680nH (13:6:6 Turns) - 100uH (1911:882:882 Turns) I strongly recommend changing T1 to a much higher mU core material. This should result in -3dB points at 50 and 200 kHz, flat pass-band and -30dB at 20 and 500 kHz. Good luck finding a 50 Ohm antenna at these frequncies. An active antenna may be the way to go, otherwise you'll need many acres for all the wire. On Sun, 11 Sep 2005 12:09:16 -0400, TRABEM wrote: I am trying to scale an existing front end receiver (butterworth bandpass) filter to a different frequency range. Unfortunately, it has a transformer in the original design, so I'm stuck. I also don't know how to handle the load presented by the active front end component other than it's probably not significantly reactive. The existing filter is for a 7 Mhz receiver, I'd like to have a similar filter design for 50 to 200 Khz. The filter components and transformer winding details are in the document at: http://www.amqrp.org/kits/softrock40...0Assy%20v1.pdf The input chip is an FST3126, spec sheet is at: http://www.fairchildsemi.com/ds/FS/FST3126.pdf The spec for the T30-2 transformer core is at: http://partsandkits.com/T30-2.htm I have aade filter design software, but it isn't allowing me to plug in the transformer into the design page of the software....so, I need to know it's equivalent circuit I think. The transformer winding details are on page 4 of the document and the schematic of the front end is page 9. Ultimately, I need new values for L1, L2, C20, C21 and C22. If someone can give me a reasonable guess as to the equivalent circuit of the transformer and the IC (U5), I can do the rest of the job myself using the aade filter software. Thanks, T |
I am trying to scale an existing front end receiver (butterworth
bandpass) filter to a different frequency range. Unfortunately, it has a transformer in the original design, so I'm stuck. I also don't know how to handle the load presented by the active front end component other than it's probably not significantly reactive. The existing filter is for a 7 Mhz receiver, I'd like to have a similar filter design for 50 to 200 Khz. =========================== If you are happy with the HF receiver as is, would it not be easier and more effective to build an LF (50 - 200 kHz) to HF converter. This should be an easy project and you can select a quiet "HF band" for the conversion. I have seen designs with a SBL-1 mixer but also a number with the NE612 osc/mixer. Perhaps others frequenting this NG have built such a converter. Frank GMØCSZ / KN6WH |
Hi,
Hi, I am trying to scale an existing front end receiver (butterworth bandpass) filter to a different frequency range. Unfortunately, it has a transformer in the original design, so I'm stuck. I also don't know how to handle the load presented by the active front end component other than it's probably not significantly reactive. The existing filter is for a 7 Mhz receiver, I'd like to have a similar filter design for 50 to 200 Khz. Notice that the primary to secondary (total) turns ratio is almost 1:1, that both shunt capacitors in the band-pass filter are of the same value and that the path resistance on either side of the secondary, through the FST3126M to the low-pass filters is around 20-ohms. So what you are looking at (to a crude approximation) is a design impedance of 50-ohm all the way through. As for the toroid, for optimum Q you might want to look for a mix more suitable for the frequency range you quoted (maybe 15?) and increase the diameter to take the extra wire as you will need to scale the impedances for use at 50kHz. The FT series, I think, need less turns for the same inductance and so should be easier to wind. Finally, a simple low-pass filter would probably be all you need here as the transformer primary impedance will be a limit the low end. Filter response is dependant on source as well as load impedance and most antennas that you would use at those frequencies will be quite reactive and difficult to design a fixed input network for. You may be aware that a lot of HF receivers use a separate hi-Z antenna terminal for LF/MF. Having said that, if you do want to use a simple low-pass configuration, remember that a symmetrical 'Pi' filter with source and load impedances of 1-ohm and a cut-off of 1-rad/sec would use two 1-Farad capacitors and a 2-Henry inductor. To scale to new impedances and frequency, divide C (and multiply L) by the design impedance and then divide both C and L by the new cut-off frequency in rads/sec. Cheers - Joe |
Notice that the primary to secondary (total) turns ratio is almost 1:1, that both shunt capacitors in the band-pass filter are of the same value and that the path resistance on either side of the secondary, through the FST3126M to the low-pass filters is around 20-ohms. So what you are looking at (to a crude approximation) is a design impedance of 50-ohm all the way through. OK, that's about what I thought, but I was estimating closer to 15 ohms for ac resistance between the secondary winding and ground. So, I was having a hard time believing it could be 50 ohms in and out. It probably makes little difference though in the filter values or passband response. As for the toroid, for optimum Q you might want to look for a mix more suitable for the frequency range you quoted (maybe 15?) and increase the diameter to take the extra wire as you will need to scale the impedances for use at 50kHz. The FT series, I think, need less turns for the same inductance and so should be easier to wind. I was looking at the CT series, specifically a CT-50-57, which is a little bigger, but probably can contain all the wire easily. Finally, a simple low-pass filter would probably be all you need here as the transformer primary impedance will be a limit the low end. Filter response is dependant on source as well as load impedance and most antennas that you would use at those frequencies will be quite reactive and difficult to design a fixed input network for. You may be aware that a lot of HF receivers use a separate hi-Z antenna terminal for LF/MF. I'm aware of the separate antenna inputs some LF radios have. The lowfer group has quite a bit of information on antennas, and it seems that making a 50 ohm antenna or something approaching 50 ohms from a large single or multiturn loop isn't that difficult. Making a 15 or 20 ohm antenna is even easier, perhaps I should think about 15 ohm input impedance and a 15 ohm output impedance since it's actually easier to handle the antenna step up. Having said that, if you do want to use a simple low-pass configuration, remember that a symmetrical 'Pi' filter with source and load impedances of 1-ohm and a cut-off of 1-rad/sec would use two 1-Farad capacitors and a 2-Henry inductor. To scale to new impedances and frequency, divide C (and multiply L) by the design impedance and then divide both C and L by the new cut-off frequency in rads/sec. Most likely it's probably best to keep it at 50 ohms and not to worry about the slight mismatch in that frequency range. Thanks to you and to all who commented, it was just he sort of information I needed to put me back on track. Regards, T |
On Sun, 11 Sep 2005 23:43:36 +0100, "Highland Ham"
wrote: The existing filter is for a 7 Mhz receiver, I'd like to have a similar filter design for 50 to 200 Khz. =========================== If you are happy with the HF receiver as is, would it not be easier and more effective to build an LF (50 - 200 kHz) to HF converter. This should be an easy project and you can select a quiet "HF band" for the conversion. What would this solve ? You still need some selectivity in front of converter. I would also question the need for a bandpass filter, but a good low pass filter would definitively required in any case. I would suggest a low pass filter below 150 kHz in Europe, Africa and Middle-East and below 500 kHz in the rest of the world to get the very strong LW/MW broadcast band signals out of the mixer. If 455 kHz IF is used, the LPF would have to be below 400 kHz in the rest of the world. I have seen designs with a SBL-1 mixer SBL-1 is specified for 1-500 MHz on the RF and LO port, so not really suitable for this band. However, the SBL-3 goes from 25 kHz to 200 MHz. The SRA-6H goes from 10 kHz to 50 MHz and should be able to handle up to +10 dBm signals. but also a number with the NE612 osc/mixer. I have used the Datong LF converter, which uses the Siemens S042 mixer/osc IC similar to the NE602/612 and it definitively needs a preselector in front of it to get away with spurious responses all over the LF band from broadcast stations. Paul OH3LWR |
What would this solve ? You still need some selectivity in front of converter. I would also question the need for a bandpass filter, but a good low pass filter would definitively required in any case. I would suggest a low pass filter below 150 kHz in Europe, Africa and Middle-East and below 500 kHz in the rest of the world to get the very strong LW/MW broadcast band signals out of the mixer. If 455 kHz IF is used, the LPF would have to be below 400 kHz in the rest of the world. I have seen designs with a SBL-1 mixer SBL-1 is specified for 1-500 MHz on the RF and LO port, so not really suitable for this band. However, the SBL-3 goes from 25 kHz to 200 MHz. The SRA-6H goes from 10 kHz to 50 MHz and should be able to handle up to +10 dBm signals. but also a number with the NE612 osc/mixer. I have used the Datong LF converter, which uses the Siemens S042 mixer/osc IC similar to the NE602/612 and it definitively needs a preselector in front of it to get away with spurious responses all over the LF band from broadcast stations. Thanks Paul, and yes....you're correct. Building a conventional converter would still require a passband filter, so little is to be gained, except that perhaps someone else has already done the design:: Also, this receiver design has no mixer, it is simply a detector and a very linear one to boot. No mixing byproducts are present because there is no non-linear mixer. In effect, this design is already a converter....except that it converts to audio directly from the rf frequency input. The "spurious responses all over the LF band from broadcast stations" probably don't exist in this type of receiver, which is one of the attractions for VLF use of this technology. Take a look at the link to the design in the original message and you will learn how it operates without mixers and without non-linear detectors. It's WHY I so interested in this particular method of reception and WHY I want to make a front end for vlf for it. The concept is explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. Regards and again, Thanks, T |
The concept is
explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. =================================== How do you know the QEX article is not a load of of old-wives tales. |
Actually, the thing you call a detector IS a mixer. You can probably
find some references for "H-mode mixer." It's good, to be sure, but it's inaccurate to say that it's _perfectly_ linear. (I'd LOVE to find a practical sampler which had zero distortion...though then I'd need amplifiers with zero distortion, too...) As someone else pointed out, any practical antenna you have for LF is very unlikely to be a good match to 50 ohms, and is very likely to be quite reactive so that by the time you add components to tune it, the bandwidth will be pretty narrow. So if you have a tuned LF antenna, which is quite usual, the response will be quite narrow, and why would you care about a bandpass filter? I'd recommend a loop with a tuning arrangement at its feedpoint (variable capacitance), and an appropriate preamp to drive a feedline back to the receiver. With that, you won't need any filter, just a transformer going into the mixer (converter-detector-whatever). A while back, I did some work to modify a design you can find at http://www.cpinternet.com/~lyle/bal-pre/bal-pre.htm, so that the control was done as a DC current , which also fed the power to the preamp on the same line that signals come back on. It worked out well. But check out other antenna options from Lyle's website (http://www.cpinternet.com/~lyle/) or others devoted to LF, too. Cheers, Tom |
If you read Gerald Youngblood's first QEX article on the SDR-1000
software-defined radio, you will see on page 7 another embodiment of this type of mixer/detector. It was originally popularized and patented by Dan Tayloe, but has recently been reconfigured (for patent purposes as much as anything else I suspect), and renamed 'quadrature sampling detector' (QSD). The original embodiment shown in the QEX article has no transformers. As a passing comment, one of the writers here said a mixer IS a detector. He's absolutely correct. A detector is just a special case of frequency mixing where the RF is mixed directly down to DC baseband. As another writer said, ALL mixers/detectors, regardless of whether they're the modern switching type or the antique-type based on square-law nonlinearity, have an overload point beyond which they make unacceptable distortion. Switching mixers, which include most double-balanced diode mixer (DBM) modules and most Gilbert-cell IC mixers, just happen to be more linear than many square-law devices up to the overload point, then they go to hell in a handbasket. The QSD is a just special case of switching mixer that can produce quadrature baseband outputs very conveniently, but it is a bit better in the distortion department than many diode DBM's. Joe W3JDR TRABEM wrote in message ... What would this solve ? You still need some selectivity in front of converter. I would also question the need for a bandpass filter, but a good low pass filter would definitively required in any case. I would suggest a low pass filter below 150 kHz in Europe, Africa and Middle-East and below 500 kHz in the rest of the world to get the very strong LW/MW broadcast band signals out of the mixer. If 455 kHz IF is used, the LPF would have to be below 400 kHz in the rest of the world. I have seen designs with a SBL-1 mixer SBL-1 is specified for 1-500 MHz on the RF and LO port, so not really suitable for this band. However, the SBL-3 goes from 25 kHz to 200 MHz. The SRA-6H goes from 10 kHz to 50 MHz and should be able to handle up to +10 dBm signals. but also a number with the NE612 osc/mixer. I have used the Datong LF converter, which uses the Siemens S042 mixer/osc IC similar to the NE602/612 and it definitively needs a preselector in front of it to get away with spurious responses all over the LF band from broadcast stations. Thanks Paul, and yes....you're correct. Building a conventional converter would still require a passband filter, so little is to be gained, except that perhaps someone else has already done the design:: Also, this receiver design has no mixer, it is simply a detector and a very linear one to boot. No mixing byproducts are present because there is no non-linear mixer. In effect, this design is already a converter....except that it converts to audio directly from the rf frequency input. The "spurious responses all over the LF band from broadcast stations" probably don't exist in this type of receiver, which is one of the attractions for VLF use of this technology. Take a look at the link to the design in the original message and you will learn how it operates without mixers and without non-linear detectors. It's WHY I so interested in this particular method of reception and WHY I want to make a front end for vlf for it. The concept is explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. Regards and again, Thanks, T |
TRABEM wrote in message ... [...] Thanks, T Heresy maybe but I'd be inclined to just dump the transformer and bandpass coils and caps and 10ohms etc. Use opamps for a low pass and a high pass filter. Then add another inverting opamp to provide the antiphase for the mixer chip. High 'Atmospherics' at these low frequencies mean pretty much any opamp will be OK. Use a resistor to present any input Z to the Ant'. The FST3126 (or 74HC4066) works well with opamp drive. regards john |
I'd hardly call it heresy, John. In my search for really good op amps
to use up to 50 and 100MHz (very low distortion and low noise), I've come across some that would be really outstanding up to a couple hundred kHz. In fact, since you can get 24bit ADCs that cover up to that range (e.g., AD7760, AD7762) with very good linearity and low noise, you could make the whole LF receiver with just the tuned antenna, the preamp out at the antenna, _maybe_ a bit of gain, and the ADC feeding into a PC. Then the "quadrature mixing" would all be done digitally, with much better accuracy than you'll get with an analog mixer. Yeah, yeah, you have to write some software to get it to work...but a modern PC should have no trouble keeping up with doing all the signal processing. There may even be sound cards out there with response out to 100kHz--that's an area I don't keep up with. Cheers, Tom |
On Mon, 12 Sep 2005 17:39:22 +0000 (UTC), "Reg Edwards"
wrote: The concept is explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. =================================== How do you know the QEX article is not a load of of old-wives tales. Not sure if you're serious or not. But, there are quite a few users with SDR-1000 HF transceivers that use the technology....and all of them aren't full of BS. GL. T |
On 12 Sep 2005 14:01:59 -0700, "K7ITM" wrote:
Actually, the thing you call a detector IS a mixer. You can probably find some references for "H-mode mixer." It's good, to be sure, but it's inaccurate to say that it's _perfectly_ linear. (I'd LOVE to find a practical sampler which had zero distortion...though then I'd need amplifiers with zero distortion, too...) As someone else pointed out, any practical antenna you have for LF is very unlikely to be a good match to 50 ohms, and is very likely to be quite reactive so that by the time you add components to tune it, the bandwidth will be pretty narrow. So if you have a tuned LF antenna, which is quite usual, the response will be quite narrow, and why would you care about a bandpass filter? I'd recommend a loop with a tuning arrangement at its feedpoint (variable capacitance), and an appropriate preamp to drive a feedline back to the receiver. With that, you won't need any filter, just a transformer going into the mixer (converter-detector-whatever). A while back, I did some work to modify a design you can find at http://www.cpinternet.com/~lyle/bal-pre/bal-pre.htm, so that the control was done as a DC current , which also fed the power to the preamp on the same line that signals come back on. It worked out well. But check out other antenna options from Lyle's website (http://www.cpinternet.com/~lyle/) or others devoted to LF, too. Thanks Tom, I've followed lowfer technical discussions for some time now and thanks to some recent input from a few of them, I have a much better idea of what I need to do to get a decent antenna up. My plan is to have a single turn or 2 turn centertapped loop, each side being 10 to 12 feet long and the turns spaced 5 inches apart. The conductor will be 200 A aluminum service entrance cable which I have laying around. The impedance of this antenna will be low and the Q should be quite high, with lots of area, so it should drive the receiver well. It has occurred to me that the antenna itself has a great deal of selectivity, yet some loop users still report front end overload from AM broadcast band and other megawatt LF rf sources. The QSD is susceptible to harmonics also, so a very high attenuation low pass filter is the minimum filter necessary to keep these signals out. Whether the tuned antenna by itself is adequate, I don't know. But, I'm considering putting in a bandpass or low pass filter designed to match the lower impedance loop antenna directly, so the filter would have input and output impedances of 5 or 10 ohms. Anyway, that's a topic for another day I suspect. Regards and thanks for the input. T |
Hi Joe,
It appeared to me that the transformer was used as a convenient means to introduce 1/2 of the Vcc to provide DC bias equally to each of the 4 switch inputs. The 10 ohm series resistors look like a resistive impedance matching scheme to me, with a built in 6+ db loss associated with them. I'm thinking of redoing the entire input circuit to take out hte reistors and to better match the lower impedance of most VLF loop antnnas. Thanks for your comments. T On Mon, 12 Sep 2005 23:08:13 GMT, "W3JDR" wrote: If you read Gerald Youngblood's first QEX article on the SDR-1000 software-defined radio, you will see on page 7 another embodiment of this type of mixer/detector. It was originally popularized and patented by Dan Tayloe, but has recently been reconfigured (for patent purposes as much as anything else I suspect), and renamed 'quadrature sampling detector' (QSD). The original embodiment shown in the QEX article has no transformers. As a passing comment, one of the writers here said a mixer IS a detector. He's absolutely correct. A detector is just a special case of frequency mixing where the RF is mixed directly down to DC baseband. As another writer said, ALL mixers/detectors, regardless of whether they're the modern switching type or the antique-type based on square-law nonlinearity, have an overload point beyond which they make unacceptable distortion. Switching mixers, which include most double-balanced diode mixer (DBM) modules and most Gilbert-cell IC mixers, just happen to be more linear than many square-law devices up to the overload point, then they go to hell in a handbasket. The QSD is a just special case of switching mixer that can produce quadrature baseband outputs very conveniently, but it is a bit better in the distortion department than many diode DBM's. Joe W3JDR TRABEM wrote in message ... What would this solve ? You still need some selectivity in front of converter. I would also question the need for a bandpass filter, but a good low pass filter would definitively required in any case. I would suggest a low pass filter below 150 kHz in Europe, Africa and Middle-East and below 500 kHz in the rest of the world to get the very strong LW/MW broadcast band signals out of the mixer. If 455 kHz IF is used, the LPF would have to be below 400 kHz in the rest of the world. I have seen designs with a SBL-1 mixer SBL-1 is specified for 1-500 MHz on the RF and LO port, so not really suitable for this band. However, the SBL-3 goes from 25 kHz to 200 MHz. The SRA-6H goes from 10 kHz to 50 MHz and should be able to handle up to +10 dBm signals. but also a number with the NE612 osc/mixer. I have used the Datong LF converter, which uses the Siemens S042 mixer/osc IC similar to the NE602/612 and it definitively needs a preselector in front of it to get away with spurious responses all over the LF band from broadcast stations. Thanks Paul, and yes....you're correct. Building a conventional converter would still require a passband filter, so little is to be gained, except that perhaps someone else has already done the design:: Also, this receiver design has no mixer, it is simply a detector and a very linear one to boot. No mixing byproducts are present because there is no non-linear mixer. In effect, this design is already a converter....except that it converts to audio directly from the rf frequency input. The "spurious responses all over the LF band from broadcast stations" probably don't exist in this type of receiver, which is one of the attractions for VLF use of this technology. Take a look at the link to the design in the original message and you will learn how it operates without mixers and without non-linear detectors. It's WHY I so interested in this particular method of reception and WHY I want to make a front end for vlf for it. The concept is explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. Regards and again, Thanks, T |
Hi Tom,
I think the loop antennas are pretty quiet compared to wire antennas and you might need some gain. But, since the antenna is high Q and tuned, it might have enough voltage output to drive the soundcard more or less directly. I don't know for sure. But, if the switch is anywhere near linear, you would not want your gain stage before the switch, would you?? I can't see using an rf amp at the antenna that just creates non linearity when you could use a nice quiet audio amp op amp on the far side of the analog switch. An all software based receiver shoulnds like a neat idea until you realize you run out of dynamic range by trying to sample such a wide bandwidth directly. I think the analog switch (hardware) is here to stay, at least for ashile. GL. T On 12 Sep 2005 17:13:12 -0700, "K7ITM" wrote: I'd hardly call it heresy, John. In my search for really good op amps to use up to 50 and 100MHz (very low distortion and low noise), I've come across some that would be really outstanding up to a couple hundred kHz. In fact, since you can get 24bit ADCs that cover up to that range (e.g., AD7760, AD7762) with very good linearity and low noise, you could make the whole LF receiver with just the tuned antenna, the preamp out at the antenna, _maybe_ a bit of gain, and the ADC feeding into a PC. Then the "quadrature mixing" would all be done digitally, with much better accuracy than you'll get with an analog mixer. Yeah, yeah, you have to write some software to get it to work...but a modern PC should have no trouble keeping up with doing all the signal processing. There may even be sound cards out there with response out to 100kHz--that's an area I don't keep up with. Cheers, Tom |
There are easier ways to introduce the bias than to add transformers.
Personally, I hate transformers and go out of my way to avoid using them but sometimes they're a necessary evil. I'd venture to say that the transformer in this implementation of Tayloe's detector was introduced to: 1) Improve second-order distortion performance 2) Eliminate the need for differential amplifiers at the output 3) Attempt to skirt some patent issues. These are just my opinions...I'd like to hear different points of view. As to your redesign goals, be aware that the input source resistance is an important part of the inherent bandpass response that this detector has. It's based on commutating filter principles.....you can read up on this by doing a web search. The resistors were probably added to stabilize the selectivity performance in the face of variable antenna impedances. The resistors will add a little to the insertion loss and thus the system noise figure, but this will probably not be an issue at LF frequencies where atmospherics and man-made noise dominate. Overall, I think this is an outstanding detector for an LF receiver. You should be able to get very good quadrature LO phasing using the common & simple Johnson counter approach. Once you get to quadrature detector outputs, an audio frequency DSP should result in a very good receiver. As you probably know, there are several public-domain DSP software packages available for both Windows and Linux that will do a very good job for you. Joe TRABEM wrote in message ... Hi Joe, It appeared to me that the transformer was used as a convenient means to introduce 1/2 of the Vcc to provide DC bias equally to each of the 4 switch inputs. The 10 ohm series resistors look like a resistive impedance matching scheme to me, with a built in 6+ db loss associated with them. I'm thinking of redoing the entire input circuit to take out hte reistors and to better match the lower impedance of most VLF loop antnnas. Thanks for your comments. T On Mon, 12 Sep 2005 23:08:13 GMT, "W3JDR" wrote: If you read Gerald Youngblood's first QEX article on the SDR-1000 software-defined radio, you will see on page 7 another embodiment of this type of mixer/detector. It was originally popularized and patented by Dan Tayloe, but has recently been reconfigured (for patent purposes as much as anything else I suspect), and renamed 'quadrature sampling detector' (QSD). The original embodiment shown in the QEX article has no transformers. As a passing comment, one of the writers here said a mixer IS a detector. He's absolutely correct. A detector is just a special case of frequency mixing where the RF is mixed directly down to DC baseband. As another writer said, ALL mixers/detectors, regardless of whether they're the modern switching type or the antique-type based on square-law nonlinearity, have an overload point beyond which they make unacceptable distortion. Switching mixers, which include most double-balanced diode mixer (DBM) modules and most Gilbert-cell IC mixers, just happen to be more linear than many square-law devices up to the overload point, then they go to hell in a handbasket. The QSD is a just special case of switching mixer that can produce quadrature baseband outputs very conveniently, but it is a bit better in the distortion department than many diode DBM's. Joe W3JDR TRABEM wrote in message . .. What would this solve ? You still need some selectivity in front of converter. I would also question the need for a bandpass filter, but a good low pass filter would definitively required in any case. I would suggest a low pass filter below 150 kHz in Europe, Africa and Middle-East and below 500 kHz in the rest of the world to get the very strong LW/MW broadcast band signals out of the mixer. If 455 kHz IF is used, the LPF would have to be below 400 kHz in the rest of the world. I have seen designs with a SBL-1 mixer SBL-1 is specified for 1-500 MHz on the RF and LO port, so not really suitable for this band. However, the SBL-3 goes from 25 kHz to 200 MHz. The SRA-6H goes from 10 kHz to 50 MHz and should be able to handle up to +10 dBm signals. but also a number with the NE612 osc/mixer. I have used the Datong LF converter, which uses the Siemens S042 mixer/osc IC similar to the NE602/612 and it definitively needs a preselector in front of it to get away with spurious responses all over the LF band from broadcast stations. Thanks Paul, and yes....you're correct. Building a conventional converter would still require a passband filter, so little is to be gained, except that perhaps someone else has already done the design:: Also, this receiver design has no mixer, it is simply a detector and a very linear one to boot. No mixing byproducts are present because there is no non-linear mixer. In effect, this design is already a converter....except that it converts to audio directly from the rf frequency input. The "spurious responses all over the LF band from broadcast stations" probably don't exist in this type of receiver, which is one of the attractions for VLF use of this technology. Take a look at the link to the design in the original message and you will learn how it operates without mixers and without non-linear detectors. It's WHY I so interested in this particular method of reception and WHY I want to make a front end for vlf for it. The concept is explained in a QEX article in greater detail, Im happy to send the url to anyone who wants to learn more about these high performance direct conversion receivers. Regards and again, Thanks, T |
An all software based receiver shoulnds like a neat idea until you
realize you run out of dynamic range by trying to sample such a wide bandwidth directly. In another reply to this thread, I mentioned commutating filters. At VLF, one of these ahead of a good ADC would probably also yield very good performance without downconverting to audio baseband first. As to speed of the ADC, using undersampling you theoretically only need enough speed to sample at twice the modulation (information) bandwidth. This is just a few kilohertz sampling rate. I've heard of this technique, but don't recall ever seeing it implemented in a real receiver design. Can anyone comment and shed more light on this?? Joe W3JDR |
On Mon, 12 Sep 2005 21:31:52 -0400, TRABEM wrote:
It has occurred to me that the antenna itself has a great deal of selectivity, yet some loop users still report front end overload from AM broadcast band and other megawatt LF rf sources. Look for "antenna effect". Most likely the broadcast signals are connected directly to the pickup wiring, which works then as an ordinary "electric" antenna. One way would be to make the pickup loop of coaxial cable. Bend the end of the cable back to a convenient place on the cable to form a loop, solder the end (both centre connector and shield) to the shield at that point. At the opposite end of the loop cut the shield and now you have a nice electrostatic shield around the centre conductor. Paul OH3LWR |
On Mon, 12 Sep 2005 21:52:01 -0400, TRABEM wrote:
But, if the switch is anywhere near linear, you would not want your gain stage before the switch, would you?? I can't see using an rf amp at the antenna that just creates non linearity when you could use a nice quiet audio amp op amp on the far side of the analog switch. If you use a huge amplification on a single frequency, you can end up with stability problems, due to unexpected feedback paths, such as a direct conversion receiver with a lot of audio gain started to oscillate, when the loudspeaker sound vibration was connected back to some front end component that was microphonic. Putting a (possibly switchable) preamplifier between the selective loop and the mixer would allow some gain to be done at non-audio frequencies, thus, reducing the risk for stability problems. Paul OH3LWR |
Hi T,
I think you missed the point. With a tuned antenna, what comes out of the antenna+tuning is NOT broadband! Frequencies not so very far from the one you are tuned to will be greatly attenuated. With that sort of input, you should be able to get by with even a 16-bit converter, if it's linear (such as a decent delta-sigma is). For example, a square loop antenna one meter on a side, at 150kHz, tuned to resonance with a capacitor, should have a Q around 300. That means it's about 0.5kHz wide at the 3dB points, and will be down about 20dB at 2.5kHz away from center, and 40dB down at 25kHz away. -- I just read your other reply to my other posting in this thread, where you worried about broadcast band overload. I suspect that if you have that, it's because, as someone else said, the antenna is acting as something other than a loop for that frequency. It's important to keep the loop balanced with respect to ground. I'd strongly recommend against a "shielded" loop unless you understand just why you are doing that. The shield becomes the antenna, and as such, it must be symmetrical... Also, for the antenna you described, about 3 meters on a side and large wire, expect the Q to be even higher and the bandwidth narrower. I think you'll find the resonated impedance to be more like a few kohms for a single such turn. Then, use a good balanced FET amplifier to get to a low impedance to drive your transmission line. By the way, I would note that the switching detector/mixer/converter in the schematic you showed is not as good as the usual current implementation of the H-mode mixer, because the channels of the FETs doing the switching in the one you gave a link to operate at a voltage which depends on the instantaneous signal amplitude, if I read the schematic right, and since the channel resistance is a non-linear function of that voltage, the detector will not be strictly linear. The H-mode mixers operate one end of the switches at a constant voltage, and of course the other end when the switch is on must be very close to the same voltage. I expect (and I think the practical experience is) that the H-mode mixer will be more linear. Do you know what the third order intercept for your mixer is supposed to be? I'd be pretty surprised if it was better than about +45dBm. But even so, even if you DID have a broadband antenna, you can find op amps, and you can make amplifiers with discrete parts, that have distortion products more than 120dB below the level of signals in excess of a volt at the amplifier output, in the LF frequency range. In other words, the distortion products will be less than a microvolt, with one volt output signals. You don't need to run that preamp with any appreciable voltage gain, so you're handling some pretty big input signals. And the best of the 24-bit delta-sigma ADCs shouldn't be far behind that. (I wish I could do that well at 50MHz!--we do make a 23-bit ADC that samples at up to 20MHz, but it's a bit pricey for what you're trying to do.) As others have pointed out in this thread, the atmospheric noise is so bad at LF that the antenna doesn't have to be very efficient to capture enough signal to be useful for receiving. Unless you are practically next door to a transmitter operating on a frequency near the ones you care about listening to, dynamic range isn't likely to be a big issue at LF. [Your other posting suggests that folk DO have troubles with other signals. I'd go looking for answers about WHY before jumping to conclusions about what to do about them.] That's a far cry from the case at HF. Cheers, Tom |
On 13 Sep 2005 14:21:59 -0700, "K7ITM" wrote:
Hi T, I think you missed the point. With a tuned antenna, what comes out of the antenna+tuning is NOT broadband! Frequencies not so very far from the one you are tuned to will be greatly attenuated. With that sort of input, you should be able to get by with even a 16-bit converter, if it's linear (such as a decent delta-sigma is). For example, a square loop antenna one meter on a side, at 150kHz, tuned to resonance with a capacitor, should have a Q around 300. That means it's about 0.5kHz wide at the 3dB points, and will be down about 20dB at 2.5kHz away from center, and 40dB down at 25kHz away. -- I just read your other reply to my other posting in this thread, where you worried about broadcast band overload. I suspect that if you have that, it's because, as someone else said, the antenna is acting as something other than a loop for that frequency. It's important to keep the loop balanced with respect to ground. I'd strongly recommend against a "shielded" loop unless you understand just why you are doing that. The shield becomes the antenna, and as such, it must be symmetrical... Also, for the antenna you described, about 3 meters on a side and large wire, expect the Q to be even higher and the bandwidth narrower. I think you'll find the resonated impedance to be more like a few kohms for a single such turn. Then, use a good balanced FET amplifier to get to a low impedance to drive your transmission line. By the way, I would note that the switching detector/mixer/converter in the schematic you showed is not as good as the usual current implementation of the H-mode mixer, because the channels of the FETs doing the switching in the one you gave a link to operate at a voltage which depends on the instantaneous signal amplitude, if I read the schematic right, and since the channel resistance is a non-linear function of that voltage, the detector will not be strictly linear. The H-mode mixers operate one end of the switches at a constant voltage, and of course the other end when the switch is on must be very close to the same voltage. I expect (and I think the practical experience is) that the H-mode mixer will be more linear. Do you know what the third order intercept for your mixer is supposed to be? I'd be pretty surprised if it was better than about +45dBm. But even so, even if you DID have a broadband antenna, you can find op amps, and you can make amplifiers with discrete parts, that have distortion products more than 120dB below the level of signals in excess of a volt at the amplifier output, in the LF frequency range. In other words, the distortion products will be less than a microvolt, with one volt output signals. You don't need to run that preamp with any appreciable voltage gain, so you're handling some pretty big input signals. And the best of the 24-bit delta-sigma ADCs shouldn't be far behind that. (I wish I could do that well at 50MHz!--we do make a 23-bit ADC that samples at up to 20MHz, but it's a bit pricey for what you're trying to do.) As others have pointed out in this thread, the atmospheric noise is so bad at LF that the antenna doesn't have to be very efficient to capture enough signal to be useful for receiving. Unless you are practically next door to a transmitter operating on a frequency near the ones you care about listening to, dynamic range isn't likely to be a big issue at LF. [Your other posting suggests that folk DO have troubles with other signals. I'd go looking for answers about WHY before jumping to conclusions about what to do about them.] That's a far cry from the case at HF. Hi Tom, I read your message above about 3 times now AND read all the other posts in this thread over again. After all of this, I must say that I'm very much in agreement with you although I know little about the linearity of the particular analog switch used in this configuration. When I started this, I was paranoid about out of band signals mixing and creating problems. Read posts from almost anyone using an LF preamp or presenting a design for one and they will almost certainly contain warnings about overload and mixing byproducts. So, I wanted an almost unattainable filter on the front end, without realizing in fact how much attenuation the antenna I hope to build will have for out of band signals. You are exactly right, there is probably no need at all for a tuned input in the receiver since the antenna tuning will be so sharp. Maintaining HI-Q in the antenna should be the primary goal I think rather than worrying about the input filter parameters! I disagree about the rf preamp however, Bill Ashcock says I shouldn't need one at all as long as I keep the Q high in the antenna and feed it into a balanced line to get to the shack. Bill says some of the guys who have single turn loops 40 feet per side or larger have so much signal, they have to attenuate. I'd like to start out without a preamp unless it is really needed. By the way, the antenna might not be balanced.....but if it's fed through a balun on each, the feedline is balanced. And, the feedline can be simple twisted wire pairs....which goes a long way towards reducing stray rf pickup and makes the feedline cheap and the losses low. If I feed the antenna into a balun and run that to another balun at the receiver, isn't that the same as having a traditional center tapped loop in terms of 'balance'? I'll try to find some software for loop design and see what the loop looks like in terms of impedance and then decide how to couple that directly into the receiver without a front end filter (just simple impedance matching). Thanks so much to you and everyone who wrote regarding this, and for the nudge in the 'right direction'. I feel a lot better now regarding the plan of attack than I did just 2 or 3 days ago! Regards, T PS:I like your estimate for my loop impedance. If it's 1000 ohms as you think it might be, I can handle that step down just fine with a balun or 2. I hope it turns out to be true: |
On 13 Sep 2005 14:21:59 -0700, "K7ITM" wrote:
But even so, even if you DID have a broadband antenna, you can find op amps, and you can make amplifiers with discrete parts, that have distortion products more than 120dB below the level of signals in excess of a volt at the amplifier output, in the LF frequency range. In other words, the distortion products will be less than a microvolt, with one volt output signals. You don't need to run that preamp with any appreciable voltage gain, so you're handling some pretty big input signals. With some kind of vertical antenna (possibly with some capacitance) which is small compared to the wavelength on LF, it is going to have a large capacitive reactance. There will also be some antenna input capacitance from input to ground, so essentially there is a capacitive voltage divider formed by the antenna capacitance and input capacitance and the amplifier is trying to tap off that voltage. So the amplifier really needs current gain, not voltage gain i.e. a high input impedance and a manageable (50 ohm etc.) output impedance. Paul OH3LWR |
You can find a good program to estimate loop inductance and some other
parameters on Reg Edwards' web pages. He has a link in many of his postings on this group and the r.r.a.antenna one. The impedance comes from the Q and the fact that you are resonating it--or at least it's presumed that you are resonating it. So if it has a Q of 300 and the inductive reactance is 50 ohms, the resistance when resonated is 300*50=15000 ohms, for example. That's why folk like to use preamps at the antenna: transform that high impedance down to a low impedance that's easy to send along a transmission line. Seems to me that if they are having trouble with intermod distortion in the preamp, the preamp isn't designed properly. It's not terribly difficult to get very low distortion at LF these days. By the way, if you build a really big loop and have so much signal you can attenuate it, that gives you a chance to lower the Q and increase the bandwidth: if what you want to listen to occupies much bandwidth, you don't want your antenna to filter out the information you want to listen to! I'd suggest you read an antenna book like Johnson and Jasik, or the antennas chapter of King, Mimno and Wing's "Transmission Lines, Antennas and Waveguides." They will make it a lot clearer why you might want a balanced loop. You don't need a grounded center-tap to make it balanced--just make it very symmetrical. Cheers, Tom |
On 15 Sep 2005 01:14:37 -0700, "K7ITM" wrote:
You can find a good program to estimate loop inductance and some other parameters on Reg Edwards' web pages. He has a link in many of his postings on this group and the r.r.a.antenna one. Yes, I found it day before yesterday and was stunned to see the loop impedance change so much as the loop is moved off frequency. I want my antenna to cover 50 Khz to 200 Khz and the antennas impedance will vary from 6K ohms to 1 K ohm, quite a LARGE range. The software was very helpful and enlightening. I was hoping to feed the antenna to the house over twisted pair line laying on the ground. This requires a balun to make the low impedance line balanced. I think the antenna should be balanced as well, which helps in elimination of out of band signals that might overload the preamp. I think, but aren't positive that the balanced antenna is necessary to eliminate the 'antenna effect' which allows the antenna to pick up other signals that it wasn't designed for just because it's a piece of wire hanging in free space. So, my plan was to build a balanced loop and feed it to the house with a balanced feedline. The impedance comes from the Q and the fact that you are resonating it--or at least it's presumed that you are resonating it. So if it has a Q of 300 and the inductive reactance is 50 ohms, the resistance when resonated is 300*50=15000 ohms, for example. That's why folk like to use preamps at the antenna: transform that high impedance down to a low impedance that's easy to send along a transmission line. Seems to me that if they are having trouble with intermod distortion in the preamp, the preamp isn't designed properly. It's not terribly difficult to get very low distortion at LF these days. By the way, if you build a really big loop and have so much signal you can attenuate it, that gives you a chance to lower the Q and increase the bandwidth: if what you want to listen to occupies much bandwidth, you don't want your antenna to filter out the information you want to listen to! It's not likely that I will ever want to listen to SSB or any other wider band modes, but I did consider putting in a resistor to kill the Q if I ever wanted to do this. I'd suggest you read an antenna book like Johnson and Jasik, or the antennas chapter of King, Mimno and Wing's "Transmission Lines, Antennas and Waveguides." They will make it a lot clearer why you might want a balanced loop. You don't need a grounded center-tap to make it balanced--just make it very symmetrical. No, it's pretty clear that I want a balanced loop. Several lowfers made strong suggestions that I should not waste my time building anything that wasn't balanced, and I couldn't agree more. Cheers, Tom Tom, I hope to put up a high Q, but relatively large loop. I expect to have a very large signal output to the receiver. Signal strength of received signals will probably NOT be an issue. Because the signal will be relatively high level, I would like to resist using a preamp at all. While I was playing with filters last night, I tried to design a filter that would also convert impedance from 6 K down to 100 ohms. It became impractical with a balun BUT, a filter that transforms impedance seems to kill 2 birds with one stone. However, performance sucks real bad with filters of higher impedance and with any filter that attempts to make a large or moderate impedance transformation. Is there any other means of converting impedance with out an active amp (using passive components)? Since I have a big signal, I can sacrifice some signal strength as long as the losses are not to great. Maybe I should start another thread? Regards, T |
Yes, this might be good for another thread...
Given the limited LF bandwidth you're interested in, I can't imagine that making a transformer would be all that difficult. Not trivial, certainly, but far from impossible. A 5:1 turns ratio will give you a 25:1 impedance ratio. You'll want to use a core material that doesn't introduce distortion. A transformer like that also gives you a way to keep the loop loading balanced. But--are you going to put the tuning capacitor at the loop, or do you have in mind putting it, say, at the receiver end of the feedline? If it's at the loop, how will you adjust it? And just what size signals do you expect to get? One of the nice things about LF/VLF is the predictability of signal strengths. Also, beware of worrying a lot about feedline impedance. How long will your feedline be, in wavelengths? If it's, say, 0.05 wavelengths at 100kHz (and THAT's 150 meters long!), does it really make much difference that it's quite a different impedance than the antenna? And...what IS the impedance of the line, at that frequency? It may well be a bit different from what you calculate for the line at 10MHz. What would happen if you fed your one turn loop with 100 feet of "300 ohm twinlead" or "450 ohm ladder line" and just tuned it at the receiver with a capacitor across that line? Small transmission line wire size would ding the Q some, but would that be an issue? I'm just speculating here, and maybe someone with direct experience with that sort of feed will offer suggestions. One thing to keep in mind here is that the LOOP construction will almost certainly be the most challenging and expensive part, for a big loop. And--you may really not NEED THAT big a loop! More signal also means more atmospheric noise, and you won't improve signal:noise ratio just by getting more of both signal and noise. Anyway, once you have a well-constructed loop, it's relatively easy to play around with different feed systems--preamp and remote tuning, straight feedline, whatever. Cheers, Tom |
Hi Tom,
Given the limited LF bandwidth you're interested in, I can't imagine that making a transformer would be all that difficult. Not trivial, certainly, but far from impossible. A 5:1 turns ratio will give you a 25:1 impedance ratio. You'll want to use a core material that doesn't introduce distortion. A transformer like that also gives you a way to keep the loop loading balanced. But--are you going to put the tuning capacitor at the loop, or do you have in mind putting it, say, at the receiver end of the feedline? If it's at the loop, how will you adjust it? And just what size signals do you expect to get? One of the nice things about LF/VLF is the predictability of signal strengths. OK, I just ran some numbers for a 6000 ohm to 100 ohm toroid transformer at 190 Khz. I didn't have to deal with the secondary at all, because the primary has to be so large...I never made it past the primary! Perhaps I made an error in the calculations? The transformer has to present about 6000 ohms of inductive reactance, which is 16.6 millihenrys. Even on a large high mu core, I'd have to wind 400 turns!!!!!!!!!!! With that many turns, the losses would be big, and would still have to wind a secondary (although it would much much smaller). Did I make a mistake in the calculations? I don't mind going to the antenna to tune it-lowfer signals don't change frequency much. Also, beware of worrying a lot about feedline impedance. How long will your feedline be, in wavelengths? If it's, say, 0.05 wavelengths at 100kHz (and THAT's 150 meters long!), does it really make much difference that it's quite a different impedance than the antenna? And...what IS the impedance of the line, at that frequency? It may well be a bit different from what you calculate for the line at 10MHz. What would happen if you fed your one turn loop with 100 feet of "300 ohm twinlead" or "450 ohm ladder line" and just tuned it at the receiver with a capacitor across that line? Small transmission line wire size would ding the Q some, but would that be an issue? I'm just speculating here, and maybe someone with direct experience with that sort of feed will offer suggestions. Was hoping to use twisted wire which can be homebrewed or cat 5...it's cheap and available. I think the twisted wire runs 80 to 90 ohms impedance. I considered that I might just tolerate the mismatch since the run was so short....but it's such a big difference, I am not sure the input filter in the receiver will react as expected. One thing to keep in mind here is that the LOOP construction will almost certainly be the most challenging and expensive part, for a big loop. And--you may really not NEED THAT big a loop! More signal also means more atmospheric noise, and you won't improve signal:noise ratio just by getting more of both signal and noise. Anyway, once you have a well-constructed loop, it's relatively easy to play around with different feed systems--preamp and remote tuning, straight feedline, whatever. Are there any other ways to convert the impedance without big losses or without resorting to an active preamp? I'd like to avoid the preamp if possible, especially since I expect a good output voltage from the big and relatively high Q loop. T |
Well, the 8:1 turns ratio transformer at
http://www.jensen-transformers.com/datashts/110khpc.pdf would almost do it for you. It falls off a bit at the high end. But with it, you could go down to less than 10Hz. That might be fun. What core material were you using in your calcs? I really think you're being too cautious about preamps. But again, my point is that you can put the effort into the loop, put it up, and then play around with different feed methods. If you do use a feed line and leave the tuning capacitor at the antenna, you won't notice any significant loss in the line. Reg will try to "sell" you one of his programs for looking at how the line transforms the impedance, but I'd just use a Smith chart program myself. Much more educational, to me, than just seeing numbers. If you want to try tuning at the receiver end, use heavy wires in the line: their resistance adds to the resistance of the loop itself, lowering the Q. But with the capacitor at the loop, the circulating currents stay out there, for the most part. You could add a small variable at the receiver end, to do fine tuning. You may well discover that's an advantage...it's a pain to try to tune the loop without being able to listen to the signal... Cheers, Tom (about to bow out of this...time for you to go try some things.) |
All times are GMT +1. The time now is 09:16 AM. |
Powered by vBulletin® Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
RadioBanter.com