Sunday, October 18, 2015

Simpleceiver ~ Part 5

The Product Detector ~ Dual Gate MOSFETS

Plots added for the 2N3819 (10/19/2015)

Data Plots for Additional JFETs Added 10/20/2015

Having covered the audio amplifier stage for the Simpleceiver (again the choice is yours) we will now move on to the Product Detector stage. As the name product detector implies, the output of this stage is a product of the mixing action of two signals.
 
Think of the product detector as a "black box" where two signals are input to the box and the single output contains two products. One product is the sum of the two input signals and the other contains the difference. Filtering at the output port can remove one of the products.
 
In its simplest form the "black box" can be the basis of a direct conversion receiver where a variable frequency oscillator (known as a Local Oscillator, LO) is connected to one of the ports and signal from the Antenna (after passing through a stage of RF amplification) is connected to the second port. The sum frequency would contain the LO + Antenna signals but the difference would be in the audio range. Now one of the down sides of the Direct Conversion Receiver is that you get the same  audio signal for two values of the LO, where the LO is above AND below the incoming signal. Thus it is not single signal reception. But that does not detract from its capability as a simplistic receiver.
 
Here is the math part of what is being said. The antenna is tuned to 7.030 MHz and one supplies a LO at 7.0292 and the difference is 800 Hz (nice CW sound). Now if the same 7.030 MHz Antenna signal is mixed with 7.0308 LO signal then the difference is 800 Hz (again a nice CW Sound). In one case the LO signal is above the incoming and in the second case it is below the incoming. BUT it is the same audio signal so you will receive the same signal at two places above the dial. [This also is important in visualizing USB and LSB.] With direct conversion you will receive the same signal at two places on the dial but for a simple receiver this is only a slight inconvenience!
 
Another example of a product detector response is when a signal is input at 12.0945 MHz ( a BFO signal) and the RF signal at 12.096 MHz( coming from a crystal filter) the two outputs would be: 1) 24.1905 MHz (sum) and 2) 1500 Hz (difference). For the product detector, the  one we want is the 1500 Hz as this is then the audio output.

Typically we add a low pass filter after the product detector so only the difference (audio signal)  is passed. This filter is a Pi type comprised of a 10 NF at either end with a 1 mHy choke in the middle which now will only pass the difference frequencies. Note in this case because the input signal is coming from a Crystal Filter you will have single signal reception which is now governed by the placement of the BFO signal. To receive the opposite sideband you would need a BFO frequency of 12.0975 MHz
 
Our "black box" can take many forms including 1N4148 diode ring, packaged Double Balanced Mixer's like the SBL-1, Gilbert cells such as the SA602 or SA612, vacuum tubes like the 12AU7, or a Dual Gate MOSFET. Of course the most famous Dual Gate MOSFET is the 40673 which today are on the unobtainable list. Some of these devices have no gain, in fact have a loss while others provide a substantial amount of gain. 
 
Since the Simpleceiver is a minimalist approach we have chosen to use the Dual Gate MOSFET, which is one of the devices that has gain in the conversion process. There are a whole new crop of RF Dual Gate MOSFETs and one in particular is from NXP and is the model BF991. Most of these unfortunately are Surface Mount Devices that for many newbie homebrewer's is an anathema. The Simpleceiver shown in an early post video now has a BF991 installed --so if a homebrewer is not shy about SMD --just drop one of those into the circuit.
 

Why use the Dual Gate MOSFET?

I would like to take just a few lines to explore the why of our choice to employ the Dual Gate MOSFET in the Simpleceiver given that there are so many new technology "black boxes" at our disposal. Many would say use the SA602 or SA612 which are also gain devices which even have a "twofer" capability wherein you can have the detector and carrier oscillator in a single 8 pin device. Simply plug in a crystal and a few caps between pins 6 and 7 where you have an instant BFO.
 
The Dual Gate MOSFET is not a black box like the Sa602 or SA612. As my friend Bill, N2CQR would say about a DGM, "you can better visualize the signals being applied to the device and have greater in depth understanding of the signal conversion process." This also satisfies his term of "more homebrewedness" with discrete components.

But the SA602 or SA612 can be used equally as well with the choice is left to the builder. BUT we do have as a goal for the completer Trans-receiver project to use the Dual Gate MOSFET in many of the circuits and given we bought them for 20 cents a piece (delivered) that is far less expensive than the Gilbert Cell SA602's/SA612's which cost about $3 USD --each!
 
At this point I will leave it to the reader to further explore the pros and cons of the Dual Gate MOSFET (DGM). Some will argue noise figure issues, while others will argue tendency to overload and lest I forget phase noise issues. There are always better mousetraps; but on the continuum of choices, for a simple project, the DGM passes muster as a viable candidate.
 
But realizing that those new to homebrewing do not have 50 years experience soldering their fingers together on a routine basis,  we offer the alternative of making a Dual Gate MOSFET from two individual leaded type J310 JFET's. These devices are plentiful and recently a 50 piece quantity J310's was purchased with shipping for around $10 USD.

In one of the earlier posts we mentioned the use of LT Spice to simulate the circuit that will be used in the Simpleceiver and so it is with our "homebrewed" DGM. Initially I missed the selection for JFETS and used MOSFETS. You certainly can get some interesting results using two IRF510's in Cascode (Source of one device connected to the Drain of the second device -- Read up on Cascode circuits using a Google search). Actually I think it will work -- but a better choice is the U309 which is close to the J310. The 2N3819 resulted in less gain for the same set of circuit values. So below is my simulated DGM using two JFET's in a Cascode Circuit. In the audio range -- the simulation shows about a 19 to 20 dB of gain.


This is the circuit that was simulated for our homebrew DGM (U309's in Cascode) and below is the output plot. This works!! In our next post (or an addendum) we will build the circuit and mate it up with the audio amplifier stage. Stay tuned.

73's
Pete N6QW


 In this post we mentioned that for the same circuit the use of the 2N3819 had less gain for the same set of circuit values. It is about 6 dB less and the plot for that is shown below. Do we now rise up like those who bash the Si5351 phase noise claiming that it is 6 dB worse than the Si570. NO is the answer! The 2N3819 is a viable device --it just has less gain. This is the value of LT Spice --no soldering is involved in the evaluation and we now can expect less gain with the 2N3819. Our frequencies of choice between 300 Hz and 3 kHz for the 2N3819 vary by only about 1.5 dB.

 
10/20/2015
 
The added value of the LT Spice simulation aside from assuring that a device will work in the circuit is the cost factor. The U309 is about $6 USD ( remember we are using the J310 which is 20 cents) and another good candidate is the 2N4393 which is about 1/3 the cost. So how would these stack up in an apples to apples comparison. I re-ran the plots for these two devices and the results are shown below. The striking difference is about 1/2 of a dB (your ears will never know) improvement at 3X the cost.  Gain variation as well as maximum gain must also be considered and the 2N4393 has slightly better overall gain than the U309 in this circuit. So this makes it an easy choice to go with the 2N4393 for this application.

The 2N3819 would be "Good Enough" but has about 6 dB less gain. The J310 is less expensive and when used in this circuit works better. I am trying to locate the data factors for the J310 so I can run this same plot. What I have now is practical data of how the J310 works versus the 2N3819 -- but for some who read this blog -- that is not good enough. [Usually I get an email that someone read on the EMRFD or BITX reflector that the J310 was not a good device yet they have no practical experience with the specific device!]

Pete N6QW


 

6 comments:

  1. Looks like I need practice at biasing Fets. Other than that, your DGM schematic looks very straight forward. I love radio design! Thanks Pete!! My W1FB books are starting to look more relevant again...Lots of DGMs in them. Found these if you are interested: http://www.utsource.net/ic-datasheet/40673-1908933.html

    ReplyDelete
  2. Thanks for your post and that is the beauty of the LT Spice. I did try various values of source resistors and even added in a few 1N4148 diodes in series (ala W7ZOI). The current circuit seemed to give the best overall performance -- but then it was just a random add a few elements and take a quick look at the plot. A more rigorous, detailed, thoughtful and measured study may result in a different circuit with better performance. But this seems to do the job. Thanks for the link on the 40673 --I will check it out.

    Pete N6QW

    ReplyDelete
  3. Pete, Hoping you can post a tutorial on decoupling? Is including decoupling necessary in LT Spice?

    ReplyDelete
  4. I am not sure I understand enough to answer your question as the subject of decoupling could be all encompassing. To accurately simulate a circuit I think would pretty much entail including all the important elements.

    That said it is good design practice to include circuitry to prevent unwanted coupling (decoupling?) via the power source feeding the circuit block element. To accomplish that end you would find usually a small resistor and a capacitor in the drain (or collector) circuitry to prevent the AC signal component from "riding along" on the DC power source and being fed back into other parts of the circuit.

    This combination of a resistor and capacitor comprises a filter. In very extreme cases you may need to "decouple" at ALL frequencies. Thus you might find in combination a 100 microfarad capacitor in parallel with a 100 NF in parallel with a 10 NF in parallel with a 1NFall tied together. I do know that in one of the earlier simulations with the 2N3904 pre-amp stage that the value of the bypass capacitor at the junction of the 100 Ohm and 100K Ohm resistor markedly impacted the output plot. In fact the best plot was with a 1000 Microfarad capacitor. Usually I would stick a 47 Microfarad capacitor in there and call it happy. The value of bypass caps on the source (or emitter) resistor also impacts the shape of the output plot.

    Decoupling also may be a function of physical layout -- more than once I have found a circuit to oscillate because of the output being inadvertently coupled back into the input and it was caused purely by a physical arrangement of the circuit elements.

    Again I am not certain of your expectation of providing a tutorial on decoupling but a search through Google may provide you the information you are seeking. One of your questions asked if decoupling was necessary in LT Spice --given what I said out the impact to the output plots I would have to answer --YES!

    Pete N6QW

    ReplyDelete
  5. Pete

    Nice series of posts. Is it correct to assume the larger gain of certain JFETs also comes at a cost of lower IP3? I've built a few DGM receivers and found they were susceptible to overload (from a dipole as you suggest) unless there is double pole RF filtering and possibly attenuation before the mixer. Since there is considerable gain from any of the JFETs, does the gain matter much if we are trading gain for IP3 and thus may need attenuation when using a good antenna? Of course cost and availability still matter. Chuck - KE5HPY

    ReplyDelete
  6. Hi Chuck,

    There is never something for nothing and undoubtedly using the higher gain JFETS causes something to be lost in the mix. The Simpleceiver concept is to provide a vehicle for those who would like to homebrew a radio using commonly available parts (not too many) and that it can be done cost effectively. Twenty cents a piece delivered for the J310's is what I consider cost effective. This is not a competition grade nor a "professional" receiver but one that will enable the builder to hear signals and be usable as a station receiver.

    In this project there are some other factors that will affect overall performance including the use of the homebrew four pole filters. A much better receiver would use 6 or 8 pole filters.

    If you fast forward to part 15 you can hear how it sounds on CW during a contest on 7.030 MHz. Lots of signals --but move away a few kHz and you hear a crystal clear strong signal. So my test is not so much measuring IP3 as it is "how does it hear?" With so few devices I would say "not too bad". Now we have tried to add a few refinements such as being able to individually set the stage gain for the 1st and 2nd IF Amplifiers as well as the RF amp. We also have followed the post mixer amp with a small 2 dB pad.

    DGM's do have their limitations like many devices --but there also are some really good things and one I like is the ability to add AGC which is a further refinement. You might want to check my website http://n6qw.com and look for the link to the Tri-Band Transceiver. --there are quite a few DGM"s in that radio and you can listen to the transceiver during a contest. Not professional grade but nonetheless usable.

    Thanks for you comment.

    73's
    Pete N6QW

    ReplyDelete