HF Transceiver front end

[Wendymott]

Hi Jeremy. I had already decided to remove the 220pf and 47pf caps.... as like you I realised I was putting another attenuator in the signal path. The signal sensitivity has greatly increased.
I really do appreciate your comments... I hope I did not come over wrongly..

I attach the photo's taken of the full filter response and it is what I was hoping for. However when connected to the RF amp, some of the responses changed, particularly the 20/15 and 10M.
This was with the caps removed, thus only a 100n in series with the gate and a 100k to ground.
There were some serious SSB signals on 20 and some on 40 tonight, thus it seems to be working much better.
I have just rebuilt the PIC Controller pcb for installation tuesday, and then I will look at the front end again....as a lashup, but with the filter block in situ.
I have to fit the PIC controller as that is what switches the relays.

[G0HZU_JMR]

Yes, removing the 47p and 220p divider will definitely improve the noise figure. I'd be tempted to also put something like a 1k resistor across the 100uH drain choke to tame the stage gain down and flatten the response. Otherwise I'd expect that amp to be a bit lively with lots of peaky gain.

By the way to prove that you can make a 28MHz filter using 470pF shunt caps I quickly designed one and built it. I used cheapo leaded ceramic caps, the kind of thing Maplin used to sell but I did use a powdered iron core for the inductors. This filter has a similar component count to yours with 5 caps and two coils.

However, this was just an exercise to show a 470pF cap 'could' be used. I wouldn't recommend this filter as the parasitic inductance of the 470pF caps and also the unloaded Q of the other components is starting to cause issues. It needed a tight layout for example. See the plot below.

If I assume your radio still has a 9MHz first IF then I guess you need decent image rejection 18MHz below the 10m band. The filter below only gives about 60-65dB rejection here at marker 3. By adding one more cap to add a zero it is possible to get the image rejection to be about 80dB. But I wouldn't use 470pF shunt caps because of the parasitic inductance 'baggage' they bring. I would redesign this filter for smaller shunt caps if I wanted to make it more tolerant of component layout and component parasitics.

But the 470pF based filter still looks good... just under 2dB loss across the whole of the 10m band

I really do appreciate your comments... I hope I did not come over wrongly..

No it was fine. I'm the one who always struggles to come across well. I tend to write stuff in engineer mode and I think it may appear a bit cold sometimes. I like to think it is efficient but I know some people have issues with it, especially on other forums.

[Wendymott]

Hi Jeremy. I attach the new figures in the schematic...
I changed the RF choke to 1 m/H but I am still getting am "odd" response....overall... I am going to re look at it today...
Yes I am still using a 9 Mhz I.f
I also attach some photo's of the physical filter block..construction is pretty tight.
If you think the filter is wrong.. please suggest alternatives.

[G0HZU_JMR]

The construction of the prototype looks very solid

I think a lot depends on what specs requirements you have for things like (9MHz) IF rejection and image rejection. On both receive and transmit these filters will be needed to reject these terms. Ideally you would want at least 50dB rejection of these terms on transmit so that kind of sets the minimum requirement. You will get some IF rejection for free in the mixer but not enough on its own.

Therefore, I think a different filter topology will be required on each ham band. This is because the capacitive tap with top C coupling that you are using on all ranges tends to have very good rejection on the low side stop band but quite poor on the high side. So I doubt you will achieve adequate IF and image rejection on some bands?

So if you were on 7MHz for example you may find you get leakage of the 9MHz IF on transmit and also some leakage at the image at 25MHz. Your filter type isn't the natural choice here. A simple swap across to a filter that rolls off better at the high side (maybe with a notch/zero at 9MHz) would give much better performance from a similar component count. The quick and dirty simulation below uses a shunt C coupled BPF topology with an added notch and this has 5 capacitors and three coils. But it produces really good IF rejection and over 70dB image rejection from just this basic filter. I can post up the circuit if you are interested in this filter? It will need coils with decent Qu at 7MHz. The alternative is to add another section to the filter and use lower Qu coils to get similar performance but from more components.

[Wendymott]

Hi Jeremy. Yes please..,.. I would be most interested... As an update... I rejigged my top 3 filter assys today...MUCH improved... I attach the new schematic and photo's of the responses. For some reason the 40M and 80M filters were not improved by reversing the coil connections........so for the moment I have left them. It was quite easy to turn the other coils around.. looks a bit iffy now.. but its "development".

[G6Tanuki]

For filters, I'd suggest looking at the work of Seymour Cohn: he published his results in various 'professional' magazines [ProcIRE etc] in the late-1950s.

In later years most of his stuff was microwave-related and that's what tends to show up most often in searches.

http://www.qsl.net/in3otd/bpfilt.html for example....

Also, "Single-Sideband Systems and Circuits" by Sabin and Schoenike has a discussion of getting the most out of such filters. Alas on retiring I had to surrender my university-libraries reader-card so can't offer to get copies of any articles for you....

[Wendymott]

Hi Tanuki. Thanks for your input....I have reverted to making a couple of test modules to test the "direct" coupling to the first mixer, rather than a fet RF amp, and then I Made a Norton amp module, Post Xtal filter.... I am also looking at the old Plessey RF amps..post filter again.
Its early days yet...but I am not looking at a professional RX/TX.... just something to keep me busy, with reasonable 5/9 reports at the end of it.
There are some members.. I suspect... grinding their teeth.. but its how I operate.

[G0HZU_JMR]

It's good to see you ditched that mosfet RF amplifier. It wasn't a great circuit, with (very) poor input match and gain flatness and signal handling. The noise figure would have been very high with the original capacitive tap at the input as well.

It's telly night for me tonight so I'll post up the 7MHz BPF details tomorrow. I really want to build it first as well. I'd rather post up the results of a real filter. I know some people are sceptical of RF simulators but they are capable of very accurate results as long as the user puts in decent models for the components.

Note that you can get a pretty good indication of overall receiver performance from a classic excel spreadsheet with inputs for gain, OIP3 and noise figure for each stage in the receiver front end. Eg TRX switch loss, bandpass filter loss, mixer conversion gain/loss, crystal IF filter loss, IF amplifier, mixer2 etc etc. This way you could predict your noise figure (and therefore sensititivity) with a few lines in a spreadsheet. That way it's easy to spot is something isn't behaving as expected. There are numerous spreadsheets like this available for free online. If you aren't interested in OIP3 you can just put big numbers in each stage entry for OIP3 or just ignore that column altogether. For the sake of maybe 20 seconds of typing a handful of numbers into a spreadsheet you can predict the performance of your receiver system quite accurately.

[Radio Wrangler]

Originally Posted by G0HZU_JMR

It's good to see you ditched that mosfet RF amplifier.

Seconded!

Dual-gate Mosfet amplifiers, when matched properly can give low noise figures, far lower than are needed on the HF bands. Their weak spot lies in limited ability to handle large signals, and that is a more pressing concern in an HF receiver than low noise figure.

In good HF RF amp designs, expect to see things like moderate power transmitting transistors, transformer-feedback, or large-area JFETs. Some good receivers had no RF amp at all. Most had a switchable one which was most usually switched out.

The highest performance thing on the go for measuring the noise figure of transistors/amplifiers/receivers is itself a highly precise receiver with carefully controlled noise contributions and dynamic ranges in its various stages. IT was designed on a spreadsheet. Spreadsheets can be useful tools and simulators, spreadsheets work at RF if you use good models. Garage in, garbage out if you don't.

David

[G0HZU_JMR]

I built a simple 7MHz BPF tonight and tested it and it worked pretty much as the simulation. However, I also connected up a Minicircuits SRA3 mixer to a couple of sig gens. I used a 16MHz LO and a filtered 9MHz RF signal and then looked at what came out of the mixer at 7MHz.

See the unfiltered plot to see what the raw (unfiltered) spectrum from this mixer looks like with about -15dBm RF drive at 9MHz. It shows all the mixer terms from DC to 30MHz.

Then see what the spectrum looks like when the BPF is added. You can see that it cleans it up a lot. Obviously this is just for one type of mixer and I don't know how well the MC1496 can reject the 9MHz IF and the other mixer terms. But I'd expect that the image at 25MHz will be about the same as the wanted term at 7MHz.

Realistically, you probably only need to aim for -50dBc spurious so the results in this case are way better than this. But it shows what is possible from a fairly simple filter design.

[G0HZU_JMR]

Here's the circuit for the BPF. It does need to be built with decent components or it will lose performance. I used powdered iron cores for this design and I used large SMD parts for the 1000pF caps. But decent leaded NPO/COG parts should be OK here. I wouldn't recommend class 2 ceramics here (eg X7R).

You can see that the insertion loss was about the same as the simulation and so is the image rejection at 25MHz. I used a powdered iron core for the 310nH 'notch' coil and this gave a deeper notch than the simulation.

If your mixer gives fairly good IF rejection at 9MHz then maybe you won't need the notch LC section in the middle. The 1000pF cap and the 310nH coil could be replaced by a single 2400pF capacitor made from a couple of 1200pF caps in parallel. This would give about 35dB rejection at 9MHz. Probably enough...
The alternative would be to add a notch/zero to your existing filter to get good image rejection at 25MHz. If your MC1496 gives 30dB IF rejection then maybe you can get away with a minor tweak to your existing BPF? It depends on how far down you want to reject any unwanted spurious terms. Only you can decide how much rejection is right for you

[Wendymott]

Hi Jeremy. Wow..... you have been busy... and VERY much appreciated... I was out most of the day, but last night I build a First mixer module with a MC1496, today I added the 9.0018 Mhz filter and a single transistor Norton amp.
This is a quick reply, to show my appreciation...the schematic will follow......
You say you built the filter, so can you give me the dimensions of your inductors please..

[Wendymott]

Just completed the schematic....
Should AGC be added.... and if so where ?? As a Pin diode attenuator, ?
The Norton amp has a 37/43 torroid with 10 + 10T Bi filar
If I need more IF gain, what are your suggestions .. I have a few Plessey SL612's with a AGC pin.
Tomorrow I will connect the new front end to the second mixer and see what sensitivity I get.

[G0HZU_JMR]

The inductors were all T37-2 powdered iron toroids because that's all I have here for this frequency range. This makes it a bit fiddly to tune because it means spreading/compressing the turns to get the right inductance. But the Q of these inductors is very good and this is why the insertion loss is low.

I used 30 turns of 0.25mm enamelled copper wire for the 3.9uH coils and 7 turns for the 310nH coil.

At the moment I'm struggling with my Tek analyser as I think I've found another bug in the user interface. It doesn't manage/save screenshots properly if you save to a file on a USB stick. So my before and after mixer plots in my earlier post aren't the right pair of plots. I guess it doesn't matter because the error/difference is subtle but I've had three attempts to save the two correct files and correct it all but the damn analyser keeps saving a previous screen image again and again. Quite how the Tek software engineers can mess this up is beyond me but 'save screen to file' seems to save an earlier screenshot to my newly typed filename every time. I've tried new filenames and it still saves a historical screenshot instead of the current screen. D'oh!

[G0HZU_JMR]

I had a quick look at your schematic and noted a couple of things. Your RF input to the mixer isn't terminated in 50 ohms. Is this termination done on another board? Ideally you would want your preselector filters to see a good termination somehow.

Also your IF amplifier looks like a classic BJT wideband amplifier with a 4:1 transformer and resistive feedback to set a 50 ohm input and output impedance. Is this impedance right for your crystal filter? I thought those 9MHz filters were 560 ohm?

[Wendymott]

Hi Jeremy. These souper douper techy boxes ... sometimes you wonder if they are too good.

Your RF input to the mixer isn't terminated in 50 ohms. No not terminated properly yet.. and the Xtal filter should be 560 ..... This is an area that I am not to familiar with "how to".

The components on the input to the filter were "lifted" from somewhere.... cant remember where ... so should I be looking at similar for the output to the MC1496 ?
Or would you suggest a torroid transformer.

[Radio Wrangler]

First place to look is for the spec sheet of your crystal filters. All sorts of different impedances have been used over the years and terminating impedance was a selectable option on many filters

Unless the terminating impedance (what the rest of the circuitry looks like to each end of the filter) you usually get a lot of passband ripple.

One way to determine what impedance to use for an unknown filter is to sweep the filter, connected with resistors for source and load Z, and then change values for minimum ripple, then try a bit of capacitance with the resistors. Datasheets are quicker if you can find 'em.

In the real radio you then have to concoct two networks to take whatever your mixer output and IF amp input happen to look like and transform them into the impedance of your filter's dreams.

With the MC1496 probably giving a fairly high Zout, most people just wodge a resistor across the output

David

[G0HZU_JMR]

You could try a simple L match between the filter and the BJT amplifier. This matches 50 ohm to 560 ohm at 9MHz. See the attached diagram with the changes shown in red.

However, with just 6 ohms in the emitter, that BJT amplifier looks to have a lot more gain (20dB fixed gain?) that would be needed there because it has no AGC capability.

I don't think that amp is critical in terms of noise figure and I don't think it needs much gain either. Assuming this is a single conversion superhet design, you really want IF amplifiers with lots of AGC range after this amplifier.

[Wendymott]

Hi Jeremy. Yes the amp has a 20db gain.. And thanks for the termination changes...
I had considered using the Plessey 612's as they have a built in AGC pin...If I copied the Plessey data book re the Transceiver, I could implement the IF amp and AGC chip.
I will experiment with both a 2 stage and 3 stage amp. I can "knock up" a pcb tonight and give it a try.
Thanks for your continued support...David too.
I have ordered some T37-2's from ebay.... should be with me next week.

[Wendymott]

Hi Peeps...... I am in a bit of a dilemma ..... What approach to take ref the Multiband filter.
It was intended that it be dual direction.. i.e Input on receive and output filter on transmit.
The filters were designed to be 50R "ish" impedance in and out.
However if I go directly into the MC1496, which has a typical input "Z" of 10K, obviously there is a mismatch there, and equally if I drive the Output filter with a MC1496, which is my intention, again the "Z" is 10K.
I found a filter calculator on the web, which is very helpful, however it assumes the input "Z" and output "Z" are the same. I can get values for a "Z" of 10k, but then there would be a serious mismatch at the aerial.
I am trying to observe the great assistance I have already received, but I am getting to a point where I may shelve it and do something else.
As a foot note, I got a reasonable sensitivity on all bands with a "post filter" amp using 2 x SL612's, but that then over loaded the 2nd MC1496 UP converter on TX.