EMRFD Message Archive 11110
Message Date From Subject 11110 2015-06-02 18:25:56 jwolczanski Discrete component version of Gilbert Cell Mixer EMRFD has a nice piece on this in Chapter 5 and the specifications look reasonable - infinitely better than the integrated circuit form (which ought to be banished!). I'm curious - has anybody used this discrete component version of the Gilbert Cell mixer?
11111 2015-06-03 10:09:32 Jon Wanzer Re: Discrete component version of Gilbert Cell Mixer Greetings Jerry,To which layout are you referring? Can you post the figure number? If you are looking at fig 5.28 that would certainly be doable and at low cost.I just finished a diode ring mixer similar to the one in figure 5.19(c) on my Beach 40 project, no input transformer on mine.The schematic is here http://www.kk6gxg.com/wp-content/uploads/2015/05/Bal-Mod-Prod-Det.jpg and the overall build page is here http://www.kk6gxg.com/projects/beach-40-txcvr/ I need to post some updates on the build page today or tomorrow.While I agree with using discrete parts whenever possible, I haven't been able to find a better solution in some applications. Like bandwidth for an oscillator. You can only push a crystal so far. Adding banks of crystals adds to cost, parts count, and cost.I am researching VFO possibilities and my first experiment will likely use a 602, but only the oscillator section, pins 6 &7, along with 3 & 8 for power. I am trying to design something I can drop in and replace the VXO on my Beach 40 project without making any other changes. Plug-and-play. I'm also trying to design it with commonly available parts that are inexpensive.IMHO I would much rather go with discrete components, even with a small performance hit, but I leave all options on the table.If you end up building the fig 5.28 Gilbert cell, please post about! It looks like fun.73,
~Jon Wanzer KK6GXG
11112 2015-06-03 10:15:57 Dana Myers Re: Discrete component version of Gilbert Cell Mixer 11113 2015-06-03 12:06:30 David Snowdon Re: Discrete component version of Gilbert Cell Mixer But where can you find diode-array ICs these days? I recall that the
was quite popular back in the 70's and 80's, but are hard to locate now.
Dana Myers email@example.com [emrfd] wrote:
> Keep in mind, in a balanced mixer, more balance is always better. A
> monolithic Gilbert
> cell mixer is likely to always be far superior to a discrete
> implementation. Even passive
> diode mixers benefit from using monolithic multi-diode chips.
> Dana K6JQ
This email is free from viruses and malware because avast! Antivirus protection is active.
11114 2015-06-03 12:12:46 Dana Myers Re: Discrete component version of Gilbert Cell Mixer 11115 2015-06-03 13:11:24 Lasse Moell Re: Discrete component version of Gilbert Cell Mixer I seem to recall a mixer made out of 4 pcs 2N5109 designed by Ulrich Rhode back in late -70's
Tried to find any reference
11116 2015-06-03 13:28:52 jwolczanski Re: Discrete component version of Gilbert Cell Mixer Yes the gilbert cell mixer shown on figure 5.28 is the one I was talking about.
I've made diode ring mixers - I was just noodling through the pages of EMRFD and found the description of this mixer rather interesting, given the fact that it's integrated cousin is rather awful....at least when it's used as a front-end mixer. I live 3km from a 22kW AM BC station. Nuff said.
VA3DKS mentioned diode arrays - darned if I didn't find a CA3039 in my junk box - a very well organized junk box at that - it was sitting all by itself in its own little compartment. I trust that would do well in a conventional diode ring mixer?
This thread started out with mixers, but Jon Wanzer mentioned a VFO using an NE602. I just made a permeability tuned VFO a few days agol - got it running and very stable, but unlike a capacitor which works with a conventional vernier, 10 turns on a brass screw doesn't lend itself to an analog display of frequency. I've seen a few mechanical solutions. Maybe the "FREQ-mite" would work - heck it's only $22.
11117 2015-06-03 13:35:18 Dana Myers Re: Discrete component version of Gilbert Cell Mixer 11120 2015-06-04 17:13:13 K5ESS Re: Discrete component version of Gilbert Cell Mixer
These are on sale at Newark for $1.01 each.
If you plan on using them better get your spectacles up to date. SOT363 pakageJ.
11121 2015-06-05 02:24:03 g3wie Re: Discrete component version of Gilbert Cell Mixer The Motorola (now OnSemi) MC1496 is an integrated Gilbert mixer where you get to choose the bias current etc with external resistors. They are still available, but maybe only in SMT packages.
11123 2015-06-05 09:08:48 K5ESS Re: Discrete component version of Gilbert Cell Mixer
Jan 2015 QST had a WSPR transmitter that I was interested in building but wanted to adapt it to other bands. It uses an SA612 as a balanced modulator (single ended output). I did an LTSpice simulation on the complete circuit. The output of the SA612 had a lot of distortion. This improved if I added a transformer so that the balanced output was used. I then substituted an MC1496 (single ended) for the SA612 and the output was very clean. The SA612 datasheet calls it a balanced mixer where the MC1496 is called a balanced modulator. Maybe the difference in my LTSpice results are because the two devices are optimized for different uses or perhaps the Spice models are not up to snuff. Haven’t had time to breadboard the two to see if the LTSpice analysis is correct. I can post my analysis if anyone is interested.
11124 2015-06-05 16:39:02 vasilyivanenko Re: Discrete component version of Gilbert Cell Mixer Please post your SPICE run. Thanks 11125 2015-06-05 19:47:54 Andy Re: Discrete component version of Gilbert Cell Mixer Jim Thompson (http://www.analog-innovations.com) has a device-level SPICE model of the MC1496. He claims it is accurate. It is on his "Device Models & Subcircuits" page. While at Motorola, Jim contributed to the design of the MC1494 which is also a Gilbert cell 4-quadrant multiplier from the same family (1494/1495/1496).I am wondering why this statement: "... the integrated circuit form (which ought to be banished!)" ? Does that refer to the 1496, and why?Andy 11126 2015-06-05 20:22:09 K5ESS Re: Discrete component version of Gilbert Cell Mixer
Will do. It make take me a day or two to pull together an organized package that has all the necessary files.
11127 2015-06-06 07:47:41 jwolczanski Re: Discrete component version of Gilbert Cell Mixer That was me, making that rash statement and I was not referring to the MC1496.
I was referring to the use of devices like the NE602 used in the front-end of a receiver. The dynamic range of this (and similar) devices is poor. For those of us who live next to large AM BC stations (that would be me), this family of devices is unusable as a front-end mixer. I note that a few successful designs, such as the Rock-Mite, actually use a crystal as the front end selectivity element.
Used as a general mixer, other than the front-end, I think these devices are fine. They certainly keep the parts count down!
This difference between the discreet component version and the integrated version, in terms of dynamic range, was what prompted my original post.
11128 2015-06-06 08:43:11 kb1gmx Re: Discrete component version of Gilbert Cell Mixer >>>I was referring to the use of devices like the NE602 used in the front-end of a receiver. The dynamic range of this (and similar) devices is poor. For those of us who live next to large AM BC stations (that would be me), this family of devices is unusable as a front-end mixer. I note that a few successful designs, such as the Rock-Mite, actually use a crystal as the front end selectivity element. <<<<
I'll grant for a moment that the statement is in frustration. The dynamic rage of the 602is about the same as many mixers if you allow for the RF gain.If you are having AM BC interference that's not the mixers fault, its the frontend before it. Simply put why should AM BC be getting into the mixer at all?If there is insufficient selectivity between the mixer and the antenna you will haveintermodulation.The rockmite attacked that with selectivity based on a simple crystal filter.At my location also close to three AM BC stations that was insufficient. Imust use a simple hIgh pass ahead of the rockmite, also my Tentec Tritonon 80M, and many other HF radios is required. I see a level of about0 dbm (1milliwatt) of 980, 1200, and 1430khz on the 80/75/40M invertedL from them. That crushes DBMs, 602m and FST switch based mixers.Being close to Boston I suffer the problems of major market broadcastersbeing near and strong for AMBC, FMBC, and TV.So lets attack the problem at the source. Undesired AM BC gettingto the mixer.The solution is for any radio below 40M is to insert a high-pass filter with acutoff at 1700khz. The circuit is a 7 pole LC. If you have only one or twooffending signals a tuned trap may do well.I'd suggest reading the section in EMRFD about high dynamic range as filtersare an important part of the subject area. Intermodulation is always an issueif the re are strong out of band signals. The solution is where possibleremove the offending signal(s).Allison
11130 2015-06-06 08:47:11 kb1gmx Re: Discrete component version of Gilbert Cell Mixer RCA also had transistor arrays and their linear IC data books had a four quardant mixer using them with good performance.Allison 11131 2015-06-06 11:47:57 roelofndb Re: Discrete component version of Gilbert Cell Mixer Hello Allison,
I have had excellent results on 40 m by using a top coupled double
tuned band pass filter. Sacrificing 6 dB gain by loose coupling
provides a bandwidth of 70 kHz and yet maintains ample sensitivity.
Two tone dynamic range is about 85 dB, which is in the same ball
park as a stock Drake R4C!
Roelof Bakker, pa0rdt
11132 2015-06-06 13:13:19 iq_rx Re: Discrete component version of Gilbert Cell Mixer Agreed that we've adopted the bad habit of expecting the mixer and roofing filter to do all the work. Serious RF selectivity will nearly always improve receiver performance, something well known in the 1950s that faded from amateur practice as magazine and handbook pages featured expensive boxes that claimed to do everything well in a single elegant-looking package.A longish dipole, center-fed with open wire line and a Johnson Matchbox, solves a number of receiver problems, and serves reasonably well for transmitting as well.Another mixer issue rears its inconvenient head for those of us who want receivers to sound as good as our home audio systems. That requires more than 70 dB in-channel dynamic range all the way from the antenna to the amplifier driving the speaker. The mixers need to get Stronger as you proceed through the receiver gain distribution, with the product detector the strongest mixer of all. That has never been standard practice in communication receivers, amateur and professional.Some of the characteristic sound of direct conversion receivers is that they have only one intentionally non-linear component, right at the front-end where signals are weakest and any distortion products are well below the band noise.Good discussion,Rick 11138 2015-06-06 18:54:21 jwolczanski Re: Discrete component version of Gilbert Cell Mixer Hi Allison:
I get your point. My argument about these devices are the number of beginning builders who build with these devices and end up with a very unsatisfactory receiver. For the cost and complexity of a few more parts, these beginners, be they old folk or youngsters, could have a higher performing radio. Not all of them have graduated to the pages of EMRFD. That's really the essence of my "dislike" of this family of mixers.
My BC station is 3km away with a daytime power of 22kw. I need a tunable-notch filter for my Palstar R30, and none of my homebrew 40 and 30 meter gear need anything, except a well-matched antenna (they have diode ring mixers and plenty of RF selectivity).
My best crunch proof radio is my WWII TRF regenerative radio, a Torn.e.b. See:
It does not know there is a nearby BC station - a rather amazing radio which speaks to the issue you mentioned of attacking the source.
I've recently gone back to an end-fed wire and I built a new L-network, but this time I used the inductor as the shunt element, which gives me a high-pass function.
I like to listen to the NDB band and built Rick's "High Performance Direct Conversion Receiver" for that band....BUT, LO harmonics mixed with strong AM band stations and I had to build a brute of low-pass filter to eliminate those responses. A DC receiver was probably the wrong architecture - I never thought THAT through! But it's a great radio now with a ton of front-end filtering. Picture of my NDB station is on my QRZ page.
I very much enjoy your posts!
11142 2015-06-07 09:42:16 kb1gmx Re: Discrete component version of Gilbert Cell Mixer First, don't blame the cow for soggy cereal. ;)Your absolutely right. New builders don't understand and build simple becausethey have no idea what simple can imply by omission. The rash of minimal artradios while cute and fun are generally horrid to use in less than ideal places.They are harder to understand why they fail when they are so simple, thatsimplicity makes for interactions that are often far from easily seen.I have the (dis) advantage of growing up near (under a mile) AM BC stationwhere it even came over the phones back in the day when phones werecarbon mic and mechanical dial. With that in mind the first radios werecrystal sets, WABC 770khz (rock and roll) was the preferred target alas thelocal was on 740khz and not rock. Through a succession of crystal setsnever did it, selectivity and sensitivity fail. Regens, more fail. Superhetsalmost, more front end selectivity success. As a result while I understandregens they are always last place in the simple radio pick list for me. I caneasily say regens are bad based on that, but I have over 50 years ofexperience to say, no they are not, but the circumstances made themseem inadequate. So with that we now have the 602. More of same.So we can blame something or understand it. I've been on for the latter.Why is what drove me then and still.>>> My BC station is 3km away with a daytime power of 22kw. I need a tunable-notch filter for my Palstar R30, and none of my homebrew 40 and 30 meter gear need anything, except a well-matched antenna (they have diode ring mixers and plenty of RF selectivity).<<<
I understand. I also have not less than 10 hams within a mile and one 800ft away.I've done more to develop crunchproof radios than most for cause. My most successful6M radio is a box, in a box, in a box, with no shortage of shielded wiring decouplingof DC and audio. There was substantial gains had in rejecting the unwantedsignals by alternate and unexpected paths. He even came thought he audio viathe headset leads. The result tolerates a signal of more than 15dbm 20khz awaywithout IMD. It has to as a ham with a KW 800 feet away is the source of that.The problem with your case and mine is often is both strong signal nearby andalternate paths. The BCRFI I found tends to sneak in on power and audio leads.Recently I had a wall wart (analog, not switching) that was taking the 1430 and 1200khzand mixing them to reradiate a spur at 2630 and every 230khz either side. They used diodes that were still rectifying at MW frequencies. A few caps fixed it. Still a rustyor corroded connection of the power lines can do worse.Diode mixers are often difficult as they seem simple but the subtle things thathappen are not obvious. Offsets and poor termination can make for many unwanted signals.My favorite is a balanced mixer using two(single balaced) or four(double balanced)jfets. IP3s in the +5 or better range are easy with them and that beats DBMs.>>> My best crunch proof radio is my WWII TRF regenerative radio, a Torn.e.b. See:
Torn. E. b, Radio receiver, German WWII, WW2
It does not know there is a nearby BC station - a rather amazing radio which speaks to the issue you mentioned of attacking the source.<<<<Cool radio!You have reinforced what I'd said. Selectivity, and massive shielding is a great aid.Going back the only successful regen I'd done had the same complexity as superhet.
Two transistors (2n170s) common base in a very loosely coupled TRF with a loosecoupled regen. Major pain to tune as each stage had a tuning cap plus the regen.it did have the requite selectivity. But a superhet with a single stage IF did as well ifthe front end was also protected from strong out of channel signals
I've said to many, look at any old radio deemed great and look at the construction.The HROs, R390 and Collins all weight a ton for a reason. Circuits are well shielded,both for stability and protection against unwanted signals. Today we can takecopper clad board and do same to achieve the desired effect and at far lower costwith limited tools. The advantage remains but its need and explanation of whymust be outlined.>> I've recently gone back to an end-fed wire and I built a new L-network, but this time I used the inductor as the shunt element, which gives me a high-pass function.<<
I have done same to revise my inverted L feed while adding ability to tune 160M ithelped at the cost of using it to listen to LF and VLF.>>> I like to listen to the NDB band and built Rick's "High Performance Direct Conversion Receiver" for that band....BUT, LO harmonics mixed with strong AM band stations and I had to build a brute of low-pass filter to eliminate those responses. A DC receiver was probably the wrong architecture - I never thought THAT through! But it's a great radio now with a ton of front-end filtering. Picture of my NDB station is on my QRZ page.<<
Again shielding is what is needed for that. I later went to an upconverting system.I did most of my Part-15 playing on 1750M in the late 60s so I got to learn a lot aboutRFI from all manor of sources all of which are worse now. Still listen down there.
As a pilot I morn the loss of NDBs, GPS is loosing all the backup systems (NDB,VOR, and Loran-C) still fly by IFR (I follow roads, rivers, recognized objects).
11144 2015-06-08 08:45:20 AD7ZU Re: Discrete component version of Gilbert Cell Mixer engineers learn early that simple does not mean easy.there are elegant design examples in the world but i suspect these designs were the product of a more complex beginnings.It is a more direct process to understand the function at each stage and the interfaces to the previous and following stages, get each part working well (meaning measure the performance and identify the deficiencies), then proceed to the next.. that method does not create a minimal implementation but a design that will work across broad conditions and that can be modified as new features are needed without a total redesign.just a few thoughtsRandyAD7ZU
11145 2015-06-08 09:30:41 jwolczanski Re: Discrete component version of Gilbert Cell Mixer Golly - 10 hams within a mile and BC stations galore. Makes me wonder what/where your building experience would have taken you had you lived in the wilds of North Dakota.....
Gone are tracking RF stages that were found in a few of the examples you quoted (I have an R-390). I note that my most recent ARRL handbook makes no mention of tracking RF stages and I don't think there are any modern transceivers with a front-panel pre-selector.
My little Ten-Tec 509 has a tunable pre-selector in front of a 40673 mixer and it does just fine in my high RF environment, unlike my Argosy which is just awful. Somewhere along the line I read a paper about modern receivers and the additional "stress" imposed on mixers because of these broad-band filters (and our collective unwillingness to have a front-panel pre-selector).
As for shielding, I'm a neophyte. I did build a nice receiver for 40m that has a crystal filter composed of 3 color-burst crystals. It's tightly shielded and the input cannot see the output. It's performance far exceeds what one might expect from such a simple filter. There is no trace of the opposite sideband - tho that might be due in part to the BFO/mixer that followed - it was a diode ring product detector followed by an 88mH filter from page 77 of "Solid State Design". What a great book!
Enjoy your posts!
11146 2015-06-08 11:29:15 Tayloe, Dan (Noki... Re: Discrete component version of Gilbert Cell Mixer
The “Tayloe detector” with the integrating detector capacitor on the output does form a tracking selector function for a DC receiver. The R of the input system impedance and the “C” of the detector capacitor form a low pass filter that tracks the center frequency of the detector. “Out of band” components (i.e., signals far off frequency) get shorted to ground by the R/C roll off of the detector. The effective “Q” of such a “tracking” detector can be pretty high.
You could emulate this band pass effect on a normal DC receiver diode double balanced mixer simply by hanging a 0.1 uf to 1 uf ceramic cap to ground on the DC output side. I have not tried this, but the theory is the same as was used with the bus switch version. This should decrease the mixer loss as half the signal power will no longer be thrown away in the form of a “sum” image (RF+VFO). This is a “difference only” (RF-VFO) baseband detection configuration. All the signal will be integrated onto the detection capacitor. Ideally, this should reduce the conversion loss to 3 dB (the integral of half a sine wave => 0.707x the RF peak) in addition to providing a band pass RF detection response.
You could probably apply this to active mixers like the NE602 as well, but the RC baseband roll off would be the “R” of the NE602 output pull up resistors.
- Dan, N7VE
11147 2015-06-08 12:44:02 Dana Myers Front-end preselectors (was Re: [emrfd] Discrete component version o 11148 2015-06-08 18:59:43 Ashhar Farhan Re: Front-end preselectors (was Re: [emrfd] Discrete component versi Dana,I have treasured a few multiple section capacitors from old sets for precisely this reason. However, I am surprised that the designers didn't think of using multiple resonators together instead of cascaded stages of tuned RF amplifiers. I suppose that in those days the emphasis was on sensitivity and image rejection. strong signal handling and low IMD were lower on the desirability list.-f 11153 2015-06-09 06:13:52 jorschei Re: Front-end preselectors (was Re: [emrfd] Discrete component versi
I follow this mixer discursion and with interest. I have designed a SDR receiver with the LINEAT LT5517 is a good mixer IP3 + 22 dB bud found a half octave RF filter necessary on the SW bands. I also designed a experimental pocked receiver with the Si4835 chip thad one work excilent. In the PE1KTH files you see the schematic. With a small wide band loop there was no problems with power stations on the SW band.
Received stations in the Netherland from the far east with de small wide band loop + preamplifier TR1. I wish Silic
11155 2015-06-09 09:34:46 Dana Myers Re: Front-end preselectors (was Re: [emrfd] Discrete component versi