Mystic crystals...

[dominicbeesley]

Hello all,

Thanks to everyone's help and advice on winding coils in another thread I'm now a bit further on at trying to make a Spectrum analyser. The next hurdle I've fallen at is the crystal filter. I hope this isn't too off topic for the forum but I hope it will be useful also when I come to try and rebuild my HRO's crystal filter...too

Anyway I knocked up a quick crystal 2 pole T filter with a series 620p and parallel 620p as described at http://www.giangrandi.ch/electronics...talladder.html

I picked a narrow bandwidth (145Hz) this (according to the calculator) gives a nice match into 50ohms!

I've built this on a little bit of copper clad and it works ok...so long as the frequency changes very slowly.

The first set of scans show the frequency being scanned over 10kHz around the centre frequency. At one scan per second all is more or less well but at anything higher than this there is an odd ringing. The second set of scans are with the oscillator being turned off and on at 1/2/5/10 times a second...as can be seen the crystals continue to resonate after this time. The last picture shows the wave form at the input to the filter (all the others are at the output of the oscillator).

What seems odd to me is that the oscillations don't decay off nicely but with these odd bumps if it was just resonance I'd expect it to die off as an inverse exponential...At first I thought this down to the frequency changing in steps (about 20000 per second) but looking again that gives 360 updates per 50/s sweep so that doesn't account for the bumps!.


So the questions are:
- is this normal?
- am I doing something wrong?
- Is this just a function of using a narrow bandwidth?

As ever any help would be appreciated. I really want to get this thing going so I can start using it!

Cheers

Dom

[jimmc101]

Getting into deep water here, practical things first...

Minimum usable sweep time (ST) is related to resolution bandwidth (RBW) and sweep width (Span)

ST = k * Span / (RBW^2) where k depends on the filter type (2 - 3 for typ commercial analyser)
So for your 10kHz sweep and 145Hz B/W the minumum sweep time is about 1sec.

see HP App note AN150 p22 http://cp.literature.agilent.com/lit.../5952-0292.pdf


Now for a bit of hand waving to give an idea of what's involved...

The IF is the difference frequency between the LO and Input signal (RF), it doesn't matter whether the LO or RF is modulated the result at RF is the same (reciprocal mixing). OK the phase wrt baseband may be reversed for FM.
Now what would you expect for the spectrum of a carrier modulated by a 50Hz sawtooth with a peak deviation 5kHz (10kHz pk-pk) ?
Further complicated by the sweep being too fast to respond to them properly, makes me head hurt just thinking about it.
Similarly for the pulsed signals, this is 100% AM with a square wave modulation, more sidebands!

The Tchebycheff filter design gives the fastest frequency roll-off but has a time delay that varies with frequency (Group delay) so it does ring badly, a Gaussian filter has a slower roll-off but a much lower group delay and less ringing.

Jim

[Andrew2]

Dom,
Funnily enough I've just constructed two xtal filters at 9 MHz for a h/b 80m transceiver. One has a b'w of 7KHz for AM, the other is 2.7 KHz for SSB. If I plot them slowly on the analyser, all is well, if I try to do a fast sweep I get similar results to yours, but less so (probably due to my filters being less Q'ey than yours. Hours of fun....
By the way, I found that 2-pole filters can be made to have nice 'on the nose' shapes, but the skirt response is pretty poor. I settled on four pole which gives a nice overall shape, especially when a spot of capacitance neutralisation is used, as here:

http://www.agder.net/la8ak/12345/n23.htm

It sure shows those slopy sides the door!

[dominicbeesley]

Thanks both,

Jim, I'd sort of thought it would be something like that but was surprised by the shape of the roll-off but now you mention FM sidebands it start to make sense as each of these passes through the filter it will make a blip and some will be in opposite phase...I really ought to sit down and get my maths books out!

Andrew, thanks for that link I'll try those neutralising coils when I get my ferrites through next week.

For now I'll just go and make some 1kHz filters....

I really want to resolve quite narrow signals (to set up filters etc) so I guess next project will have to be some form of sample and hold A/D -> memory -> D/A to look at these narrower signals...

Thanks both

Dom

[cmjones01]

Originally Posted by dominicbeesley

I really want to resolve quite narrow signals (to set up filters etc) so I guess next project will have to be some form of sample and hold A/D -> memory -> D/A to look at these narrower signals...

Or a storage scope to look at the output. Either analogue or digital would do the job. I've successfully used my Tek 2430A DSO to store slow sweeps from my spectrum analyser, as long as the scope's sweep is triggered from the start of the sweep of the analyser.

An analogue storage scope is more fun, of course. I've recently picked up a Tek 549 which needs quite a lot of work doing to it, but I'm looking forward to having it work.

Chris

[Andrew2]

My HP 8568 has a 'max hold' facility which is of great assistance in doing slow sweeps. Once I've got one done, I can then press 'view', which freezes the display on the screen. I can then use this as a reference while using channel 2 for a new sweep to test modifications.
Without this marvellous (but rather large!) bit of clobber from the early 80's, I would be lost.

AM & SSB filters:

[dominicbeesley]

Yes, a nice modern fangled scope would be nice but at the moment I'm on a mission to build as much of my own test equipment as I can...mainly because its more fun but also because I understand more....however if I saw a nice scope, SA or VNA going cheap I'd probably be tempted. I doubt my efforts will ever be as good as a proper one!

I spent the afternoon measuring parameters of a batch of crystals in the hope of making a 1kHz bandwidth filter to see how that works...

Cheers,

Dom

[glowinganode]

I'm following this thread with interest, just wondered about your setup. Are you using the sawtooth output of the scope to drive the sig gen or vice-versa?
I have a spectrumanalyser and want to trace response curves like you,but I'mnot sure how to keepthe sig-gen and S-A in sync. Have you though about using or played with a wide band noise generator,it might be what I need to try.
Excuse the shameless plug, storage scope for sale here; https://www.vintage-radio.net/forum/...ghlight=iwatsu
PM me if interested.
Rob.

[Andrew2]

I hope you don't mind my butting in here, Rob.
Some SA's have built-in tracking generators which do the whole job for you, others have a sample of their sweeping local oscillator brought out which can be fed to a seperate tracking unit.
A wideband noise source sounds attractive, but given that most SA's are looking for 0dBm (1 mW/50ohms) for a full-height display, that's an awful lot of noise-power to generate when you consider the bandwidth involved in a typical sweep! And if it all gets to the front end of the SA at once, there may be a price to pay.
I once made a diode noise gen, followed by a MAR6 MMIC and a MAR4 MMIC, but the noise was still too low to show very much on the analyser.
I just set my analyser to free-run at a suitable rate (with MAX HOLD on) and allow my sig gen to step up slowly and non-synchronously. Any gaps or lumps missing can be filled in by doing another pass. It can be slow, but it bothers me not.

[glowinganode]

Andy, thanks for the info. Unfortunately my SA has neither which is a shame. I was going to use the sweep output to control the sig gen, but obviously the sig gen needs to be running at exactly the same frequency as the SA is seeing (within say half of the filter bandwidth) otherwise I'll end up falling down the response curve. I need to get at the LO somehow so I can PLL the sig gen to the SA.
OK about noise source, excuse my ignorance, MAR4/6 MMIC - what are they?
Rob.

[Andrew2]

Rob, MMIC's are Monolithic Microwave IC's, small 4-lead packages with 50-ohm in/out impedances. The common ones such as the Mar6 will provide 10 to 15 dB of gain at up to 1 or 2 GHz. The bigger Mar4 will do about 60mW, easily enough to blow the brains out of the mixer in your nice shiny SA!

[dominicbeesley]

Hi,

I've not yet got a SA but I have built my own Sig-gen see here: http://vintage-radio.net/forum/showthread.php?t=50579

I'm using that as the sweep at the moment straight into the IF stages of my proto-SA. Afraid I've blown this months budget on coil formers and other bits otherwise I might have been tempted for that 'scope!

I'm not sure what I'm going to do for my front-end oscillators. I was going to go with two varicap VXOs and try and get them to align but getting them within a few hundred Hz over 0-36 MHz and 36-72MHz is probably not very realistic (unless theres some cunning trick to keep them in sync) so my plan for now is to first just get a simple SA going with what I've got and an analogue LO...

Later I'll build two DDS oscillators, one at the LO and one at the RF freq. I'll then have a simple scalar network analyser, if I do two sweeps with the LO 90 degrees different I might even manage a vector network analyser....

Now I've written most of the difficult code for controlling the DDS chips I should be able to make something clever...however I really need to get my head around more of the RF theory before I get too carried away!

Dom

[dominicbeesley]

Hello all,

I had a go at another filter this time 1kHz wide. Here's a few pictures of it and the original filter being scanned. The first picture shows the 0.145kHz (supposedly!) scanned at 1kHz per division second is the 1kHz filter also at 1kHz per div. The scan speeds are 5 seconds and 1 second respectively. This is with the sig gen (via attenuator) direct to the filter

The other scans show the setup pictures having a 27MHz..47MHz (roughly) sweep at different speeds. The 1kHz wide filter is in place. So it looks like with a 1kHz filter I should be in business. So long as I keep the sweeps below 1 second! This via the 36.4MHz...4.4MHz down mixer.

I'm still not sure I've got everything right... I'm losing a lot of signal in the crystal filter. I've built a little tuned amplifier with about 20dB gain that seems to bring it all back where it should be....anyway having great fun!

Cheers

Dom

[Andrew2]

Great stuff, Dom.
It's all jolly good fun, and despite being a radio & electronics nut since I could see, it's only over the last few years that I could afford to start collecting decent test gear. It makes a huge difference to home-brewing. Suddenly, a double-tuned bandpass filter is easily possible and quickly tuned up, no more over/under coupling! Crystal filters would have been unthinkable.
My first few attempts at a SA were based on TV tuners or satellite tuners, with up-converters where necessary, all tuning done by simple ramp generators directly feeding the varicaps. They worked a treat for looking at band activity and wider filters, but
looking at narrow stuff was almost impossible due to jitter on the sweep voltage. DDS is definitely a massive step in the right direction.
As a matter of interest, what IF amp and detector does your analyser use? I settled on the SA604, an FM amp/demodulator chip which has a 80dB log detector output. The trouble with this chip is that the detector goes off the boil above 10.7 Mhz, meaning another down-conversion from whatever 1st IF I was using. I think Analogue Devices did some nice stuff, though I never got to try it.

[PJL]

Amazing work, I am also interested in your detector solution for the the Polyskop I have just got going. Of course it will never acheive the resolution yours DDS does.

[dominicbeesley]

Thanks lads,

So far it's a bit of a funny set up I may rationalise it later.

The front end isn't built yet, first attempt will be a 602 fed by the DDS running at 36..72MHz. This with an output at 36MHZ 1st IF. The trouble at the moment is winding a set of 36MHz IF coils with sufficient Q to get completely rid of the image at 28MHz...as the next stage is a 36..4.443 MHz mixer (32 MHz crystal) this is built and is what is being scanned. This feeds the crystal filters with amplifier and buffers thence to another 602 to translate this up to 10.7MHz which is fed direct into a SA614 (like a 604) used as an RSSI.

I will eventually get rid of the 4.43 to 10.7MHz stage and rebuild the RSSI to work at 4.43 MHz. At the moment it's built around some 10.7MHz ceramic filters but I think it will work well enough if these are replaced with simple IFTs tuned to 4.43 MHz so I can get rid of a stage...

For testing I'll use the DDS sig-gen to generate the sweeps but not sure what I'll do long term. I'm going to try and build an analogue sweeper too...just for the hell of it but I do like the DDS as it's dead easy to set up the frequencies you want...just type in the numbers on the keypad!

Anyway my big project for today if I'm allowed to play is a first cur front end. For this I need to wind some good hi-Q 36MHz coils. Trouble is my scope won't pick anything up at such a high frequency and I have no other test gear to help...my LCR meter won't go small enough for the coil and I've not managed (yet) to build an oscillator that will work at such a high frequency. What I'm going to try next is sweeping 30-40Mhz and feeding the output into the 36..4.43MHz converter in the hope that I can watch this on the scope.

It's great fun building up my test equipment like this. So far what I've built is probably pretty rubbish but it can hopefully be used to design and build something better and so on. Anyway I think I've now got it to a point where I can actually use it for something useful!

Cheers

Dom

[G8HQP Dave]

I think you are seeing a beat effect between the filter frequency and the applied frequency. When you apply a signal to a narrow filter, the output is at the same frequency as the input (obvious, really!). When the input signal stops, or otherwise changes, the output contains two components:
1. A growing signal at the new input frequency,
2. An exponentially decaying signal at the filter resonance frequency.
If you change the input quicker than the output can change, then these two output components beat with each other as they are often at different frequencies. This effect can be easily seen from the mathematics of a forced damped oscillation but does not seem to be sufficiently well-known among practical people - I only discovered it for myself when I was investigating the modulation behaviour of a Class E power amp for my PhD. I don't think my supervisor had ever thought about it either, although he had been teaching the maths of it for years.

[dominicbeesley]

Ah, you might be onto something there...I was quite worried by the last picture in the first post...there the log-detector sees a bigger signal when the signal is off than when it is on.

I wonder if one of the later stages isn't tuned correctly and a beat frequency is being amplified more than the intended frequency...more investigation tonight. Will give me something to do other than my so far fruitless efforts at building a 36MHz IF strip!

Dom

[dominicbeesley]

Well, I've got a bit further with this. I've realised what I really need is to do a very slow scan...trouble is that would need a decent storage scope - which I don't have. Instead for about £10 I've built a little PIC uController setup which will store 256 samples from any sweep rate between 100seconds to about 50Hz. It also does some averaging and filtering to improve matters.

The two attached pictures show my valved AM transmitter modulating an NBTV colour signal (with 15kHz subcarrier - note sidebands) and a 5kHz sine wave with a lot of 2nd order harmonics - not sure if that is down to the tranmitter or audacity.

Next up I'm going to try and build all this together into a real machine with its own screen...

Cheers

Dom

[Andrew2]

Ooh, that's nice! Watchout HP, Dom's snapping at your heels!