9468	2013-12-01 21:37:58	iq_rx	Common Gate Amplifiers -- Long Post
Message	Date	From	Subject
Hi All, The common gate amplifier has been one of my favorite topologies since the mid-1970s. My designs have been at audio through microwaves, and I have gone through a bag of 100 U310s and approximately 200 J310s, mostly in different and new prototype designs. The J310 is one of the most stable and trusted components in my repertoire, and the common gate amplifier one of my most repeatable and trusted topologies. But there are some basic guidelines to follow. I've used the J310 for amplifiers and oscillators from audio through about 500 MHz, and the U310 from audio through 600 MHz. So here's the first bit of wisdom, straight from Barrie Gilbert: The transistor doesn't know what it's supposed to do. If you are building an audio amplifier with a transistor that can also work as an oscillator at 600 MHz, you can't use 12AX7 vacuum tube audio construction techniques with long leads to a single point ground. Examine every physical circuit (not simulation!) at the highest frequency at which the transistor has gain. Bypass capacitors with leads aren't particularly effective above a few hundred MHz, and tuned circuits in my 600 MHz oscillators are less than an inch of straight wire. If your PC board layout has an inch of wire between the J310 and the next components, it may well oscillate at 400 MHz or so. Common Gate means exactly that. If you can see the gate lead, it's not common gate--it's "inductor in series with the gate." The U310 is in a metal can, with the gate attached to the can, and was designed to be soldered into a hole in the circuit board ground. A strategy I've used for years with HF through UHF amplifiers is to drill a hole for the J310 plastic body, drop the transistor in the hole with the leads up, and then solder a tin shield across the hole, with the source lead on one side, the drain on the other, and the gate lead soldered to the shield. That works every time and is certainly not original with me. It was a standard technique in the 1970s when I first started using J310s and U310s in VHF and UHF projects. The voltage gain of a common gate transistor from source to drain is roughly equal to the impedance driven by the drain divided by the impedance that drives the source. So a common gate J310 driven by 50 ohms and connected to a 500 ohm load has a voltage gain across the device of about 10. But in order to connect the drain circuit to a 50 ohm load, you need to step the impedance down by a factor of ten, using the equivalent of a transformer. At audio, we use actual transformers. At RF we often use a Pi Network. In either case, the impedance transformation is 10:1 and the voltage step down is the square root of that, or 3.16:1. The gain in dB from 50 ohm input to the 50 ohm output is 20 log Vout/Vin, so the power gain of the above common gate amplifier (20 log 3.16) is 10 dB. When I design VHF and UHF common gate amplifiers, I usually do a quick back-of-the envelope calculation to come up with pi-network components for a gain of 10 dB. At HF sometimes I go as high as 13 dB. More precise analysis with all the transistor variables would yield a gain of about 9 or 12 for those two cases, but the math detail obscures the basic underlying circuit operation. It's more enlightening to design for a gain of 10 and expect a little less. The published component values in my designs are usually taken straight from the finished and working circuit on the bench. If the simulation shows a gain peak of 6.8 MHz using values that I found to work at 7.1 MHz on the bench, spread the turns a little on the toroids in your simulation. Oh wait, you can't do that--it's a simulation. Hence, a bit of wisdom from Wes Hayward: The simulation is the greater experiment. Engineering professors and their students place great faith in circuit simulators, but engineering professors who have actually designed, built, and measured anything are becoming rare. If you know some, treasure them. Because of the size of components such as variable capacitors and toroid inductors, the source and drain leads can't be as short as the gate lead. Keep them as short as is practical and as far apart as possible. The source circuit is low impedance, so electrostatic coupling is less significant than in the drain circuit. If the drain circuit is a tuned Pi-Network, divide the total needed capacitance into a fixed cap as close to the drain as possible and a variable a convenient distance away. In PC board layouts with modern components, it is worthwhile to use a chip capacitor. Remember the J310 is a UHF component, even when we're building a 40m rig. The transistor doesn't know what it's supposed to do. Now, what about that 22 ohm resistor in series with the drain lead? In the 500 ohm drain circuit it clearly doesn't do much at the signal frequency. But the drain lead, capacitor frame, and wire leading into the pi-network inductor may easily have some resonant behavior up in the UHF range. If so then it may oscillate. So the 22 ohm resistor is a simple VHF suppressor, just like is connected to the plate cap on top of virtually every tube PA ever commercially built. Those parasitic suppressors became common in vacuum tube circuitry as soon as tubes had gain above 30 MHz, and the same basic technique works with transistors. A 22 ohm chip resistor is ideal. Regarding circuit simulators and other video games. I admit, I spend a lot of enjoyable time designing circuits in a simulator. I also spent nearly a decade as a professional RF and Analog simulator driver--that was more work than fun. But a simulation is not a radio. It isn't even a circuit, or a single transistor. It's just a model. It may be a very expensive model, like on Project Runway, but it isn't reality. Use the simulator as a tool, like a soldering iron. You could use a soldering iron to draw your circuits on a breadboard--but that isn't the best use of the tool. Develop a sense of when to turn off the simulator and warm up the iron, and vice versa. The original R2, T2 and miniR2 were designed entirely using classic engineering circuit analysis, taped-by-hand layouts, and measurements of the resulting bench prototypes--no simulators. The toughest part of the design was cramming all that high gain circuity on a small double-sided PC board with stability and signal integrity, and the simulator doesn't yet exist to successfully do that. The first of my designs that really benefitted from simulation was the R2pro. But the first attempt to build that all on a single PC board failed, so I divided it up into 5 functional PC blocks. As soon as I did that, I was able to make a number of significant performance enhancements, and since then have resisted all pressure to do another single board high-performance direct conversion receiver. Modular construction works, and is both easier and almost alway higher performance than a single board approach. The microR2 was designed with help from a simulator after carefully studying everything in Chapter 8 and 9 of EMRFD. By far the toughest part was getting a stable layout, but I took that on as a design challenge and the resulting receiver came out nicely. The J310 RF amplifier and LO were never a problem. Newly manufactured 2N3904 transistors will oscillate above 300 MHz. That means you can't treat garden variety NPN transistors like HF and audio transistors anymore either. Finally, a word about ugly construction. Yes, of course it works--but remember that the origin of the term comes from a project Wes Hayward did at his famous bench with his son Roger. Wes understands more about parasitic inductance, capacitance and RF layout than most RF designers, and had read and absorbed just about everything written in the RF Design literature for two decades by then. If you examine his ugly circuits, you will find that the leads that need to be very short are very short, and the ones that can be--or need to be--long are long. Under it all is an unetched ground plane, and particular attention should be paid to how he treats the leads soldered to that. So ugly construction is quick and works exceptionally well--usually significantly better than the first attempt at a printed circuit board--but it is a "thinking person's" construction technique. This was and remains of no concern for Wes--he assumes that every designer has moving parts until proven otherwise. These days a lot of beginning designers expect to go straight from the simulator to a working circuit, with the brain switched off. But simulators are full of things that don't exist in nature: wire, ground, and voltage-source power supplies for example. So you need to think about which wires need to be zero length (they all are in the simulator), which grounds need to be zero inductance and resistance (they all are in the simulator) etc. My schematics are in the public domain, but I put a copyright symbol on most of my printed circuit board artwork. They are art and I recommend that you study them, with particular focus on what is short, what is long, how ground is treated, how power supply leads are treated, and how components with UHF gain are handled in the layout. Becoming a successful radio designer is a little science, a lot of engineering, and a significant amount of art, and there are different learning techniques for each. For art, the time honored method is continuing study of other people's work and practice on your own creations. You will get better with time. Your 20th project will look better and work better than your first or second, so it is critical that you get off the simulator and get to the bench and start building and measuring your designs. Have fun, Rick KK7B
9477	2013-12-02 11:21:44	Dana Myers	Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers -- Lon
At the risk of stepping into a religious discussion (I'm not religious). Simulation is a powerful tool made more powerful by thoughtful application. Of course, when it comes down to "my real circuit doesn't match the simulation", the real circuit "wins". Practically speaking, this happens when the simulation is not sufficiently specified, which is the default case for the typical SPICE simulation at any frequency greater than DC. What? Did I say that simulation is inherently wrong for RF? Not really. I said that the default case for RF is that circuits are insufficiently specified. Let's start with an easy circuit - a .1uF capacitor to ground (with a SPICE line like "C1 1 0 .1u"). What's the impedance of that capacitor at 16MHz? Uh, that's easy, it is -0.1 ohms. OK, now I'll build that circuit, a .1uF capacitor, and I measure 0 ohms at 16Mhz... huh? Right, we've found the self-resonant frequency of that capacitor, in reality, the capacitor has around 1nH of parasitic inductance in series, never mind the inductance of the leads themselves and PCB traces and so on. From looking at various manufacturer datasheets, it's a pretty safe bet that your ceramic caps have ~1.5nH of series inductance. Even leadless chip caps have 600-1000pH of series inductance. So, just for fun, I took that common-gate J310 pre-amp from EMRFD Fig 9.64 and added 1.5nH of series inductance to every capacitor in the input and output networks and re-ran the AC analysis. I see a broad peak in response at 7.0MHz, with -3dB points at 5.7MHz and 8MHz. Sure - it isn't peaked at 7.1MHz like the real circuit was on Rick's bench, but it's pretty darn close and I haven't even warmed a soldering iron up. Inductors have similar issues with core loss; I don't go overboard, I generally just simulate Q with a parallel resistor of Q * impedance at center frequency. This shows up mostly in broadened frequency response.
9483	2013-12-02 16:51:41	Ashhar Farhan	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
Dana, Another great, Donald Knuth of Computer Science fame, once wrote on his blog : "Be aware of the bugs in this code. I have only proved it correct, not tried it." Proofs are derived from typographical/mathematical operations on axioms. However, every system is inherently incomplete (Godel). This is why experimental methods are at the core of science. This is not to berate theoretical derivations, but that complete and perfect modelling is impossible. Perhaps Knuth's advice to his fellow practitioners is vaulable : If you find yourself spending too much time on theory, start giving attention to practice, it will improve your theory. if you find yourself spending too much time on practice, start giving attention to theory, it will improve your practice. - f
9485	2013-12-02 17:20:07	Dana Myers	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --

9486	2013-12-02 18:11:48	Ashhar Farhan	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
Dana, Thanks, yes, simulations can confirm a circuit's performance and predict behaviour. but the key thing is to get the circuit right. that only comes from insight. For instance, a few years ago, I built a relay bank switched front-end for a multiband transceiver. The out of band rejection was abysmal. I blamed the low-q toroids, I blamed the capacitor values. But no amount of playing around would cure it. Reassembling the BPFs on a separate board worked predictably though. It turned out that there were two problems : 1. the relays were picking RF energy and coupling it to the bpf. Bypassing the relay coils solved that. 2. In spite of the toroidal properties, there was coupling between the various bandpass filters. Shielding them cured it. (Actually, Wes pointed that out when I had shown him the pictures of my new build). So, if you really want to know whether a circuit will work or not, I guess both are required. I am not pulling the plug on the 465 yet. - f
9489	2013-12-03 05:10:01	Chris Trask	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
> >Thanks, yes, simulations can confirm a circuit's performance and predict >behaviour. but the key thing is to get the circuit right. that only comes >from insight. For instance, a few years ago, I built a relay bank switched >front-end for a multiband transceiver. The out of band rejection was >abysmal. I blamed the low-q toroids, I blamed the capacitor values. But no >amount of playing around would cure it. Reassembling the BPFs on a separate >board worked predictably though. It turned out that there were two problems >: >1. the relays were picking RF energy and coupling it to the bpf. Bypassing >the relay coils solved that. >2. In spite of the toroidal properties, there was coupling between the >various bandpass filters. Shielding them cured it. (Actually, Wes pointed >that out when I had shown him the pictures of my new build). > >So, if you really want to know whether a circuit will work or not, I guess >both are required. I am not pulling the plug on the 465 yet. > Yes, that's exactly right. I caught significant grief from other engineers and managers for not indulging in simulations, and I simply ignored them. I had learned electronics hands-on before the cyber-crutches became available, and surprisingly I wrote the very first PC compatible software for RF/µwave simulations. Chris Trask N7ZWY / WDX3HLB Senior Member IEEE http://www.home.earthlink.net/~christrask/
9497	2013-12-03 09:02:33	Dana Myers	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --

9498	2013-12-03 10:09:49	Chris Trask	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
> > > > > >So, if you really want to know whether a circuit will work or not, I guess > > >both are required. I am not pulling the plug on the 465 yet. > > > > > > > Yes, that's exactly right. I caught significant grief from other engineers and managers > > for not indulging in simulations, and I simply ignored them. I had learned electronics > > hands-on before the cyber-crutches became available, and surprisingly I wrote the very > > first PC compatible software for RF/µwave simulations. > > > > Note that I'm not suggesting that simulation is necessary for building a circuit, or > is in any way better than actually building a circuit. > > I am saying that simulations that include more of the real circuit will give better > results than simulations that don't. This means better simulations require better > understand of the real circuit - not "pulling the plug on the 465", but the contrary. > That was understood. Myself, I make use of simulations to evaluate the validity of any new ideas I come up with before committing to a prototype. Many bad ideas have been set aside that way. It took me years to come up with the transformer feedback amplifier with a single transistor and both negative- and positive-feedback loops to improve linearity and noise, which eventually appeared in the September 2012 issue of RadCom and for which I received the 2012 Price-Courtney award from RSGB.
9525	2013-12-05 06:24:47	Mike Mayer	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
Simulation is one of those topics where the answer is not “always” or “never” but “it depends”. I happen to work for a company that sells simulation software so I am biased, but many of our customers design systems where the prototype cost is in the hundreds of thousands of dollars, so it is economical to first build a test board, use measurements from the test board to dial in the simulations and then use the simulations when designing the real system. Obviously in other cases it is easier to just build and test the system. The other issue is that for digital systems like memory interfaces one or two prototypes don’t give you the full range of variation that you will see in volume production. Simulation lets you explore the limits of the variation and see what happens in the worst case. Again, if you are building one then if it works it works. ================================================================== Mike Mayer mwmayer@tds.net
9532	2013-12-05 08:22:53	William Carver	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
On Tue, 2013-12-03 at 11:20 -0600, Mike Mayer wrote: > > Simulation is one of those topics where the answer is not “always” or > “never” but “it depends”....... > .........prototypes don’t give you the full range of variation that > you will see in volume production. Simulation lets you explore the > limits of the variation and see what happens in the worst case. > ....if you are building one then if it works it works. Very well put, Mike. When you're duplicating a ham design "that works", yours may not if it was "designed" with trimpots and variable capacitors and an active device that was far from the "typical" numbers on a data sheet. W7AAZ
9533	2013-12-05 08:36:00	Dana Myers	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --

9540	2013-12-05 14:37:27	William Carver	Re: Simulation vs reality (was Re: [emrfd] Common Gate Amplifiers --
> I recall seeing exactly this where someone could not reproduce the > cascode > JFET IF amplifier design - it turns his JFETs had a relatively low > Vgs(thresh) and > the recommended diode chain biased them off :-) JFETs are like > that... > > 73, > Dana Yea, JFETs can be all over the map. You "should" design bias stabilization for min/max for the device. You can do that with modeling but a bag of 100 J310s may to have even one device that is anywhere near the published limits. The other area where modeling shines is variation of circuit performance over wide temperature ranges. Especially when you don't have an environmental chamber. Bill