2997	2009-05-08 21:21:51	gt25psi2002	How to purify the LO 's harmonics and sub-harmonics?
Message	Date	From	Subject
Dear Sirs, Last week, I ordered my "Experimental Methods in Radio Frequency Design" from my local bookstore and they promised 6-8 weeks of delivery (with no extra charges of delivery and insurance). For the time being, I have a few questions to disturb the PEACE OF MIND of your goodselves. For a good practice, some mixers need the LO signals to be as clean as possible as well as all ports (RF, LO and IF) to be terminated with 50 ohms impedance for the best performance. For a fixed frequency LO, passband or T-Bridge type diplexer can do the job (I think). But what about LO that must span the entire HF range? What options do we have? Best regards, Amin
2998	2009-05-09 06:26:36	Kenneth E. String...	Re: How to purify the LO 's harmonics and sub-harmonics?

2999	2009-05-09 07:38:30	ehydra	Re: How to purify the LO 's harmonics and sub-harmonics?
The trick is to go with a Gilbert Mixer like the SA602 or others cheap to come on. In a Gilber Mixer the upper resistors act both as current sources _and_ termination for the standing waves backreflected from the ports. Also you must decide between driving it as +/-1 switch or as a 'true' multiplier (with reduced LO drive). It depends fully on your design-decisions what is the better mixer topology! http://en.wikipedia.org/wiki/Gilbert_cell For further help it may be of benefit if you desribe more what you want to do or achieve! - Henry -- www.ehydra.dyndns.info gt25psi2002 schrieb: > Dear Sirs, > > Last week, I ordered my "Experimental Methods in Radio Frequency Design" > from my local bookstore and they promised 6-8 weeks of delivery (with no > extra charges of delivery and insurance). > > For the time being, I have a few questions to disturb the PEACE OF MIND > of your goodselves. > > For a good practice, some mixers need the LO signals to be as clean as > possible as well as all ports (RF, LO and IF) to be terminated with 50 > ohms impedance for the best performance. > > For a fixed frequency LO, passband or T-Bridge type diplexer can do the > job (I think). But what about LO that must span the entire HF range? > What options do we have? > > Best regards, > Amin >
3000	2009-05-09 18:47:58	gt25psi2002	Re: How to purify the LO 's harmonics and sub-harmonics?
Dear Ken, Thanks very much, but I'm still have a half doubt about it. In my opinion, we can match the required 50 ohm impedance of the mixer with these circuits, but how is it going to reduce unwanted signals i.e. when we attenuate incoming signals, all signals will be attenuated including the wanted one . When we customize the circuit for certain range of frequencies, then it will become narrow band which will defeat our original purpose. Best regards, Amin
3001	2009-05-09 18:52:01	gt25psi2002	Re: How to purify the LO 's harmonics and sub-harmonics?
Dear Henry, Thank very much. I plan to use a high level mixer of either a H-Mode Mixer or a Doubly Double Balanced Mixer (DDBM). Is it applicable? Best regards, Amin
3002	2009-05-10 07:53:39	ehydra	Re: How to purify the LO 's harmonics and sub-harmonics?
gt25psi2002 schrieb: > Dear Henry, > > Thank very much. I plan to use a high level mixer of either a H-Mode > Mixer or a Doubly Double Balanced Mixer (DDBM). Is it applicable? > The diode-ring mixer (DRM) is the oldest variant. Still used in amateur radio and measurement equipment or specialized things like satellites. The main benefit is it's great capability to resolve a small signal in the near to a big signal. Bad is: price, LO driving level H-mode is possible if the device works at the wanted frequency. Mostly till 30-50MHz. There are other variants of active switching exists. The Gilbert Mixer is easy to integrate in a IC, needs less LO power and has the two benefits of amplification and auto-termination. Complicated to make of separate parts because of needed symmetric building. A SA602 works up to 1.3GHz I was told. The DRM goes higher. Some Gilbert Mixer variants too. For DRM termination of ports for small bandwidth you can use a diplexer of concentrated L and C. Otherwise a resistive variant is better. And yes, this will reduce the signal! Sorry, I'm not an expert in the field. Look here: http://home.earthlink.net/~christrask/ http://www.mydarc.de/dc4ku/ (german, but Google translation should help) There are many more links available. Just Google! Hope that help - Henry -- www.ehydra.dyndns.info
3004	2009-05-10 18:25:00	Chris Trask	Re: How to purify the LO 's harmonics and sub-harmonics?
> > Last week, I ordered my "Experimental Methods in Radio Frequency Design" > from my local bookstore and they promised 6-8 weeks of delivery (with no > extra charges of delivery and insurance). > > For the time being, I have a few questions to disturb the PEACE OF MIND > of your goodselves. > > For a good practice, some mixers need the LO signals to be as clean as > possible as well as all ports (RF, LO and IF) to be terminated with 50 > ohms impedance for the best performance. > > For a fixed frequency LO, passband or T-Bridge type diplexer can do the > job (I think). But what about LO that must span the entire HF range? > What options do we have? > Basically, in order to keep the harmonics and IMD products low, the mixer needs to have very good linearity on both the LO and RF ports. This means that any form of mixer that requires saturated diode (diode ring) or transistor base-emitter juncions (most active mixers) is not going to be an option. What is required in this case is a mixer in the form of a linear analogue multiplier. Making such a circuit cheaply is not easily accomplished, and normally some sort of compromise must be brought to bear. For instance, the familiar Gilbert Cell (actually first disclosed in an Australian professional journal) can be made into a linear mixer by applying the LO to the lower differential pair which has been linearised by way of a degeneration resistor between the emitters. Next, the RF signal is applied to the upper transistor quad at a level that is well below the saturation point. This way, the LO-IF and RF-IF isolation is retained and the mixer is operating much like a somewhat decent analogue multiplier. The circuit can realised with the fairly common LM1496 (aka MC1496). If RF-IF isolation is not a serious concern, the circuit can be realised with a long-tail pair of transistors or even tubes. The LO is applied to the long-tail device with a suitable amount of emitter (or cathode) degeneration, and the RF is applied to the differential pair bases (or grids). The differential pair colectors (or plates) then go to a transformer with a centre-tapped primary which will provide the LO-IF isolation and which will then pass the IF and amplified RF signals. Chris ,----------------------. High Performance Mixers and / What's all this \ Amplifiers for RF Communications / extinct stuff, anyhow? / \ _______,--------------' Chris Trask / N7ZWY _
3006	2009-05-10 20:01:09	gt25psi2002	Re: How to purify the LO 's harmonics and sub-harmonics?
Dear Chris, Very interesting idea - coming from a person who has acquired a high degree of authority in this electronic field. If it doesn't trouble you so much, please advise me the address (or articles) that I can further investigate and study the subject matter and figures. Best regards, Amin > Basically, in order to keep the harmonics and IMD products low, the > mixer needs to have very good linearity on both the LO and RF ports. This > means that any form of mixer that requires saturated diode (diode ring) or > transistor base-emitter juncions (most active mixers) is not going to be an > option. What is required in this case is a mixer in the form of a linear > analogue multiplier. > > Making such a circuit cheaply is not easily accomplished, and normally > some sort of compromise must be brought to bear. For instance, the familiar > Gilbert Cell (actually first disclosed in an Australian professional > journal) can be made into a linear mixer by applying the LO to the lower > differential pair which has been linearised by way of a degeneration > resistor between the emitters. Next, the RF signal is applied to the upper > transistor quad at a level that is well below the saturation point. This > way, the LO-IF and RF-IF isolation is retained and the mixer is operating > much like a somewhat decent analogue multiplier. The circuit can realised > with the fairly common LM1496 (aka MC1496). > > If RF-IF isolation is not a serious concern, the circuit can be realised > with a long-tail pair of transistors or even tubes. The LO is applied to > the long-tail device with a suitable amount of emitter (or cathode) > degeneration, and the RF is applied to the differential pair bases (or > grids). The differential pair colectors (or plates) then go to a > transformer with a centre-tapped primary which will provide the LO-IF > isolation and which will then pass the IF and amplified RF signals. > > > Chris > > ,----------------------. High Performance Mixers and > / What's all this \ Amplifiers for RF Communications > / extinct stuff, anyhow? / > \ _______,--------------' Chris Trask / N7ZWY > _
3008	2009-05-11 08:37:35	chris3trask	Re: How to purify the LO 's harmonics and sub-harmonics?

3011	2009-05-11 12:28:06	ehydra	Re: How to purify the LO 's harmonics and sub-harmonics?
Interesting insight, Cris! I found the Gilbert Mixer ('Cell', or what else it is common named. I remember first doc was as a phase comparator and Gilbert saw it and added the pre-distorti
3012	2009-05-11 13:38:27	Chris Trask	Re: How to purify the LO 's harmonics and sub-harmonics?
> > I found the Gilbert Mixer ('Cell', or what else it is common named. I > remember first doc was as a phase comparator and Gilbert saw it and > added the pre-distortion diodes to form a mixer?) a good mix between > price, performance, size, power consumption. > > There is only one project I made in RF ;-) There a digital clock > generator CY22393 output is shaped to have a 'sinusoid' drive to the > mixer. I cannot quantize the performance but was satisfied by the > achieved system performance. > How did you measure the performance? > > The prob with higher LO drive is that the mixer degrades to a +/-1 > switch and that means: mirror frequency AND ALL harmonics of (mirror > freq AND harmonics of mirror freq) will mix in! > The mirror (image) frequency will mix in regarless of the LO signal shape. > > I'm not sure what is better. What do you think about? > I prefer linear (multiplier) mixers for the lower noise. > > Is is useful the reduce LO drive to a DRM as in the case of the Gilbert > Mixer? > No. A diode ring mixer requires that the switching diodes be saturated by the LO. > > Seems the mixer is an endless theme. > Almost as bad as antennas. Chris ,----------------------. High Performance Mixers and / What's all this \ Amplifiers for RF Communications / extinct stuff, anyhow? / \ _______,--------------' Chris Trask / N7ZWY _
3014	2009-05-11 14:59:34	ehydra	Re: How to purify the LO 's harmonics and sub-harmonics?
Chris Trask schrieb: >> I found the Gilbert Mixer ('Cell', or what else it is common named. I >> remember first doc was as a phase comparator and Gilbert saw it and >> added the pre-distortion diodes to form a mixer?) a good mix between >> price, performance, size, power consumption. >> >> There is only one project I made in RF ;-) There a digital clock >> generator CY22393 output is shaped to have a 'sinusoid' drive to the >> mixer. I cannot quantize the performance but was satisfied by the >> achieved system performance. >> > > How did you measure the performance? In working 'range of link', no values sorry. Demodulated DBPSK with a relative low bit-error rate. Again no absolute value. Totally non-scientific, even non-engineering ;-) but hell, it is a fast quality value. > >> The prob with higher LO drive is that the mixer degrades to a +/-1 >> switch and that means: mirror frequency AND ALL harmonics of (mirror >> freq AND harmonics of mirror freq) will mix in! >> > > The mirror (image) frequency will mix in regarless of the LO signal > shape. Yes, but WILL include it's harmonics too if not a sinusoid LO. > >> I'm not sure what is better. What do you think about? >> > > I prefer linear (multiplier) mixers for the lower noise. > But there must be a hidden benefit if the LO is driven very large. Will this change the third-order intercept? >> Is is useful the reduce LO drive to a DRM as in the case of the Gilbert >> Mixer? >> > > No. A diode ring mixer requires that the switching diodes be saturated > by the LO. But why? I think there is no big difference to the Gilbert Mixer. Both work by shifting a first signal by a second signal on the BE-junction diode. For low quantities that is equal to a multiplication approximation. > >> Seems the mixer is an endless theme. >> > > Almost as bad as antennas. > Hm. A very shorted antenna will reduce bandwidth and can work as a bandpass to suppress image frequency and other transmitters too. regards - Henry -- www.ehydra.dyndns.info
3015	2009-05-11 15:27:43	Chris Trask	Re: How to purify the LO 's harmonics and sub-harmonics?
> > > > > > I'm not sure what is better. What do you think about? > > > > > > > I prefer linear (multiplier) mixers for the lower noise. > > > > But there must be a hidden benefit if the LO is driven very large. Will > this change the third-order intercept? > For diode ring and communtative mixers it will. Below saturation the conversion gain will be dependent on the LO level. Above the saturation threhold the IMD level will improve with increased LO level, but only to a point. The IMD products are primarily generated during the polarity transitions in the LO signal. Tucker and others investigated this back in the late 1940's and 1950's. > > >> Is is useful the reduce LO drive to a DRM as in the case of the Gilbert > >> Mixer? > >> > > > > No. A diode ring mixer requires that the switching diodes be saturated > > by the LO. > > But why? I think there is no big difference to the Gilbert Mixer. Both > work by shifting a first signal by a second signal on the BE-junction > diode. For low quantities that is equal to a multiplication approximation. > In a Gilbert mixer a reduced LO level will reduce the mixer to a poor analog multiplier, poor because it has no predistortion to overcome the BE junction nonlinearities. As long as we keep the switching quad in saturation and the driver pair somewhat linear it works reasonably well, but is very noisy due to the spectral folding. To make it work as a somewhat linear analogue multiplier/mixr we need to apply the LO to the driver pair at a level much higher than the RF applied to the quad, and the LO level must produce a current change that is much less than the bias current so as to keep that linear. It's not an easy problem to solve without some form of feedback. In a diode ring mixer, dropping the diodes out of saturation will greatly reduce the gain and drive the IMD levels up, as you will now have a finite logarithmic (very nonlinear) resistance in series between the source and the load. Chris ,----------------------. High Performance Mixers and / What's all this \ Amplifiers for RF Communications / extinct stuff, anyhow? / \ _______,--------------' Chris Trask / N7ZWY _
3016	2009-05-11 15:40:32	ehydra	Re: How to purify the LO 's harmonics and sub-harmonics?
Chris Trask schrieb: >> But there must be a hidden benefit if the LO is driven very large. Will >> this change the third-order intercept? >> > > For diode ring and communtative mixers it will. Below saturation the > conversion gain will be dependent on the LO level. Above the saturation > threhold the IMD level will improve with increased LO level, but only to a > point. The IMD products are primarily generated during the polarity > transitions in the LO signal. Tucker and others investigated this back in > the late 1940's and 1950's. Interesting. So I postulate there exits an optimum driving shape, like the RRCF for digital baseband signals to avoid ISI. Maybe it is sinusoid, I don't know. >> But why? I think there is no big difference to the Gilbert Mixer. Both >> work by shifting a first signal by a second signal on the BE-junction >> diode. For low quantities that is equal to a multiplication approximation. >> > > In a Gilbert mixer a reduced LO level will reduce the mixer to a poor > analog multiplier, poor because it has no predistortion to overcome the BE > junction nonlinearities. As long as we keep the switching quad in > saturation and the driver pair somewhat linear it works reasonably well, but > is very noisy due to the spectral folding. To make it work as a somewhat > linear analogue multiplier/mixr we need to apply the LO to the driver pair > at a level much higher than the RF applied to the quad, and the LO level > must produce a current change that is much less than the bias current so as > to keep that linear. It's not an easy problem to solve without some form of > feedback. Hm. I miss a scheme of feedback for mixers. It is just to simple to drive the beast with sinusoid or digital level. One such 'sophisticated' circuit is the image rejection mixer. But maybe there are other possible too?? Here I remember a patent by Oliver Bartels I will look for it. > > In a diode ring mixer, dropping the diodes out of saturation will > greatly reduce the gain and drive the IMD levels up, as you will now have a > finite logarithmic (very nonlinear) resistance in series between the source > and the load. > changing LO drive: If the LO drive is temporary blocked, you have a noise blanker. Maybe a opportunity to combine mixer and if-amp in one stage? Hm. That is a point I'd like to discuss with myself going now to bed. Thank you! regards - Henry -- www.ehydra.dyndns.info
3017	2009-05-11 16:37:51	Russell Shaw	Re: How to purify the LO 's harmonics and sub-harmonics?
ehydra wrote: > Interesting insight, Cris! > > I found the Gilbert Mixer ('Cell', or what else it is common named. I > remember first doc was as a phase comparator and Gilbert saw it and > added the pre-distortion diodes to form a mixer?) a good mix between > price, performance, size, power consumption. > > There is only one project I made in RF ;-) There a digital clock > generator CY22393 output is shaped to have a 'sinusoid' drive to the > mixer. I cannot quantize the performance but was satisfied by the > achieved system performance. > > The prob with higher LO drive is that the mixer degrades to a +/-1 > switch and that means: mirror frequency AND ALL harmonics of (mirror > freq AND harmonics of mirror freq) will mix in! Even with a "multiplier" (non-linearly compensated Gilber cell), the 3rd harmonic LO gain is unlikely to be more than 20dB less than that of a square-wave LO drive. This is still a very high level spur for a receiver, and must be designed around as normal with any square- wave mixer, but of course the level being dealt with is a bit less. > If drive is reduced, the mixer will read and produce less harmonics > interference, so filtering effort at in-/and output is reduced. > > I'm not sure what is better. What do you think about? I increase the LO whilst the 3rd-order intercept increases too, until the spurious-free 3rd-order dynamic range is equal to the rest of the receiver (or that of the next stage). This usually means a balanced mixer will end up being a "squarish" +/- 1 gate. > The AD834 is pricey and needs big current. Is there any real rf project > using it? BTW: I thought the Gilber Mixer IS a multiplier if both LO and > RF is low. Somewhere I read it but cannot recover where. > > > Another point is what is the behaviour if the LO drive tracks the RF > level. Any benefit of such or similar scheme? Not really. > @third-order intercept: > All cry for more more more! But what is a really practicable magnitude? > If the antenna is electrically 'small', the NF should go down and > third-order intercept can be smaller. The attenuator before the LNA should always be adjusted so that the atmospheric noise is just above the receiver added noise. Then the system dynamic range will always be maximum, regardless of antenna size. > @noise figure: > Adding a LNA will reduce dynamic range!? So I think there is a need for > a implied low NF in the mixer. The LNA should simply have a dynamic range equal to or higher than the following stages. The system dynamic range won't be reduced much then. > Is is useful the reduce LO drive to a DRM as in the case of the Gilbert > Mixer? No. Gilber cell mixers should have enough LO drive to become more of a switching action too imo. > I'm always only interested in elegant designs, not to find the ultimate > for high-priced measurement equipment aka R&S, Agilent etc. > > Seems the mixer is an endless theme. > > Sorry, my english is bad and the rf knowledge is worse.