A few months ago I posted a quick write-up of a design I did many years ago, not too long after DRM first started appearing on the airwaves. It was a pretty straightforward JFET + 2 BJT design, nothing special - but it did the job. I've never been entirely happy with it though and, since that amp has graced a friend's set for years, I figured I'd build something better for mine ;)
There's basically 2 ways of going about it. The first is simple brute force: build an amp - using valves, transistors, op-amps, doesn't matter - and let it rip! Bandwidth? Pah! Who cares, as long as it's got gain at the target frequency! Feed it all into your downconverter and let the mixer sort it out!
That's pretty much the previous design in a nutshell. While certainly simple, it has several problems - happily amplifying noise as well as (or better than!) signal, and a propensity for distortion (particularly IM), to name just two. And, let's be honest, it lacks a certain elegance…
The other way is to do what thousands of designers have done for years - build an IF strip with gain. It's more complex, but (provided you build it right!) you'll get much better results.
So let's start by looking at an IF strip:
That's a classic transistorised IF strip, shamelessly stolen from (I think) somewhere on www.vintage-radio.com. Typically, T1 would be fed from the output of the IF mixer; T1 then drives the first OC45 transistor, which in turn feeds T2, which in turn drives the second OC45, which in turn feeds T3, which finally runs off to the detector. (As an aside: you can tell it's an old circuit by the transistors used - not only is the OC45 a germanium transistor, not silicon, but it was one of the first commercially-available RF transistors, made by Philips in 1956!).
But take another look at T1-T3. They're adustable; their inductance can be tweaked somewhat. And notice the 250pF capacitor across the primary? That's a resonant circuit; just what we want to form a bandpass filter…
(Another aside: at first glance, it's probably not a 455kHz IF strip - 250pF caps were common in 470kHz IF strips of the time. But the principle is the same…)
So all up we've got a bunch of resonant circuits that can be adjusted to pass only the frequency band we want - and, with transistors there, we've got (or can have) gain. That's a good start!
For comparison, let's have a look at the 455kHz IF stage of the FRG-7700…
There's similar transformers there (they're adjustable; it's just not shown) - T08 is the input from the 455kHz 2nd mixer, T09 & T10 are obviously in the signal path, and T13 is the output (while T11 & T12 are just chokes) - but there's a subtle difference: the majority of the IF filtering is done by the ceramic filters CF01 (±12.5kHz) at the input, and CF02, CF03, and CF04 in the middle.
The reason is pretty simple: Selectivity.
In a normal AM broadcast receiver, you want a relatively wide IF bandwidth for decent-quality speech, acceptable-quality music, etc. To allow that, channel spacing is 9 or 10kHz (and transmitters in range of one another are rarely allocated to adjacent channels unless it's impractical); not only can you get away with the (typical best-case minimum) 9kHz or so bandwidth of a (fairly gentle) 455kHz LC filter, you need that bandwidth for music.
Shortwave is different. Its traditional purpose was long-distance information and voice communication, not entertainment - the 9/10kHz bandwidths required for acceptable music were unnecessary, and so 5kHz spacing was settled on for voice (and even narrower for SSB, CW, etc). Additionally, transmitter power levels and propagation mean that stations on opposite sides of the world are frequently in-range of each other, so keeping adjacent channels free is difficult. So narrower, sharper cutoff filters - ceramic or crystal - are used.
(Even so, the stock filters used in the FRG-7700 - 2.7kHz for CW/SSB/AM Narrow, 6kHz for AM Medium, and 12kHz for AM Wide (at their 6dB points) - are all considered v-e-r-y wide for their purposes, and hams especially like to replace them with narrower filters.)
Another subtle difference: there's not much amplification in the FRG-7700's IF filter itself. On the diagram above, Q21 feeds the AGC of a later amplification stage, Q23 does noise blanking via D20/21, Q22 is there to keep the signal level optimum for the filters, and Q24/25 really only exist to overcome filter losses and drive T13. All the real IF amplification (and AGC) is done in the following stage. For good reason…
Tuned circuit + high gain = oscillator. Or, at least, a tendency to oscillate. Being ready to ring like St. Mary's on Christmas Eve is a good thing in a regen receiver, but a bad thing in a superhet. We don't want that. The FRG's designers got around that by keeping the tuned circuits well well well well well apart from the gain (like caramel sauce & citrus tart†). So attention to overall gain, gain distribution, and the basics (e.g. keeping the output away from the input!) will be important.
With that bit of background done, it's time to think: what do we want from an IF amp?
- First of all, we want a high impedance input. Most of the FRG's 455kHz IF stage isn't too high an impedance, but we don't want to load it down if we can help it - that would affect the radio's operation.
- Secondly, amplification. I forget exactly what the signal levels are around the 455kHz IF stage of the FRG-7700, but they're low - too low to drive a typical NE/SA602/612 mixer properly*, let alone a diode mixer (if I wanted to go that way). But we want enough in the right places, and not too much in the wrong…
- Thirdly, bandwidth. A DRM signal on HF is 10kHz wide, and ideally I'd like 20kHz (i.e. ±10kHz, or 445-465kHz) for some other things I want to try e.g. spectrum analyser..
Point 1 is pretty easy to manage - a simple FET input or buffer stage will do the trick. Maybe needs a dual-gate device, if only to help buffer the input C against the Miller Effect, although a single gate device is worth a try (and may in fact help stabilise the whole thing).
Point 2 is only slightly more difficult in theory - just a matter of measuring, doing the calculations, and dividing it up across the stage - but can be a problem in practice. Layout will be key, and isolation may be necessary.
Point 3 is a bit trickier. Above, I used the rule of thumb of "at best, 2% of frequency" to guesstimate the IF transformer + C bandwidth as 9kHz or so (i.e. 2% of 455kHz). I want 20kHz. There's a couple of ways of getting there, but it's probably easiest to detune each stage slightly and sacrifice some gain (and there's a couple of ways of doing that). It'll need a bit of tweaking to ensure it can be made fairly flat across 20kHz or so, but that's what sweep oscillators are for - I might finally find a use for those damned AD5932's!
So let's design one! Or, not…
A couple of years ago I stumbled across a blog by a guy in the UK describing his ongoing build of a DRM receiver - one that was pretty straightforward, almost classical, in many ways, but very well thought-out. When I went back there recently, I was a little surprised that there hadn't been any updates in months. I remembered he'd been ill, but it was only after a little digging that I found he'd passed away fairly suddenly…
So, as a little tribute to Steve - a guy I never met or communicated with, but who's ideas intrigued me - I'll build his IF stage as a buffer amp. I only wish I'd subscribed to his newsletter…
You can read the relevant blog posts here:
- The 455kHz IF strip
- A Revision to the 455kHz I.F. Amplifier
- Automatic Gain Control
- The Quest for the Elusive Missing Decibel
 |
| IF strip from "A Revision to the 455kHz I.F. Amplifier". |
As you can see it's an extension of the basic transistor IF strip with the modification of a high-impedance FET input, and long-tailed pair phase splitters replacing the single transistors. I have a (probably unreasonable!) liking for LTPs; they're a classic design (1930's, IIRC), and ideal for use in this style of amp.
The big problem I see with Steve's design is component selection. The CA3028's he used are basically unobtanium nowdays - so I'll need to build them from discretes. While long-tailed pairs are fairly immune to variations in transistor β, it's probably best to keep in the ballpark of the originals. Other transistor characteristics do need to be fairly well-matched - so, ideally, matched pairs should be used. Steve also mentioned that the noise figure of the amp with CA3028s is not so good - so I'll want low noise transistors.
Discrete, matched pair, low-noise RF transistors with a β~120? Uh, yeah, I'll need to investigate that…
(A quick look indicates the likey contenders are all SOT363 / SC70 / etc packages, which may present some layout problems…)
The other component problem is the transformers. They're not critical; basically, any set of AM radio IF transformers will do (traditionally they came & are used in sets - usually 3 transformers with different impedances + 1 oscillator choke; the circuit uses the 3 different transformers). The problem is that AM radios these days use "all-in-one" ICs; sometime in the last 20 years the major manufacturers (e.g. Toko, etc) stopped making 455kHz IF transformer sets. None of the major parts distributors here in Oz carry them; most of the online ham & radio hobby stores only carry one or two of the set; and while the fart-toy shop has them listed, the've been showing as out of stock in their warehouse for months. Mouser has them, but I didn't really want spend $45 on shipping for ~$10 of parts.
So be aware that they may be hard to get. If you like, feel free to grab an old transistor radio and rip the coils out of that. You want the yellow, white, and black ones…
In the end, I was on the verge of designing homebrew IF transformers with trimcap tuning - the only thing stopping me was I didn't have any suitable cores - when I rang the local fart-toy shop and asked. Their system showed 10 in stock, but it took half the staff half a day to find them. So I bought 4 sets. If you go looking at Jaycar Aspley they're down the back in one of the unmarked cardboard drawers on the bottom shelf under the resistors & such. I'm surprised it didn't have a sign nearby saying "Beware of the leopard"…
(Tony, and the guys from Jaycar Aspley: if you're reading this, how about GETTING THE COMMON BLOODY COMPONENTS UP OFF THE FLOOR and putting them at a reasonable height!?)
(Yet another aside: I noticed that Altronics have opened a store a suburb or two from Jaycar. I haven't been in there yet - only heard about it yesterday - but I know what used to be in that space:
I think I'll start calling Altronics "the sex-toy shop"…)
Anyway … I'll need to calculate how much gain I want in the IF amp stage, and select some suitable transistors. Maybe a shielded case, too - I'd rather not have stuff getting out into the rest of the radio. I might even have a go at building one with parts to hand - basically whatever single-gate JFET and unmatched transistors are in the box o' bits - and see how that goes. It'll be interesting to see how well it works…
( † It's a private joke…)
No comments:
Post a Comment