The active filter cont.

So I've finally built the device (8th order bandpass based on LTC1562-2) and it works. It turns out that my intuition was, unfortunately, correct: the gain of 36dB built into the filter caused the device to be extremely unstable. It started to oscillate whenever it could, sometimes without any apparent cause. Fortunately, the gain in this application could be set by the value of just one resistor, so when I set it to be 0dB, the thing became super-stable. After the filter I have an opamp buffer stage on MCP6022E (to buffer both Vdd/2 and Vout, as the output is differential). Fortunately, in the very last moment I've changed the PCB layout to contain the place for two more 0805 resistors, so I could choose whether the opamp should work as a follower or a non-inverting amplifier. Then I moved those missing 36dB into the output stage. What I don't understand:

  1. The filter is still stable. What difference does it make if the gain is in the filter or its DC-coupled direct successor stage?

  1. I don't have a decent spectrum analyser, so I can't check for sure, but after the mixer stage (with 76.5kHz LO) the output signal contains a lot of noise, way beyond the 2.6kHz bandwidth of the input signal. I don't know if it comes from the mixer or the input signal.

  2. I wasn't able to obtain similar results with the analog resources of a PSOC5LP (LNA on an opamp, the mixer,
4th order Sallen-Key low-pass). The various internal couplings make the device to oscillate very wildly. But the PSOCs are there *exactly* for mixed-mode applications. My analog frontend design was very conservative, so I doubt that the fault is on my side. Very disappointing.

formatting link

Funny thing: my Farnell order shipment was blocked: they wanted me to make a statement that the device will not be exported to Russia. What the heck, is the LTC1562 part so high-tech? :-)))

Best regards, Piotr

Reply to
Piotr Wyderski
Loading thread data ...

The mixer stage multiplies what's coming out of the local oscillator by what's coming in from the filter.

If the local oscillator has a square wave output its output includes all the odd harmonics of 76.5kHz, with amplitudes inversely proportional to harmonic number.

Even sine wave oscillators tend to have some higher frequency harmonic content, so the mixer output is going to contains a lot of high frequency noise.

If you did have a spectum analyser you'd see spikes in the high frequency spectrum.

--
Bill Sloman, Sydney
Reply to
bill.sloman

Noise, mostly. The post-filter gain follows 8 opamps worth of accumulated noise. Lowest noise results from getting gain early in the signal chain.

70 KHz is getting into the range of passive LC filters. They don't oscillate, and Ls and Cs don't make noise. And aren't bothered by out-of-band signals.

That LTC chip is expensive!

The oscillation could be through the supply rails. Lots of opamps have terrible PSRRs.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Are you sure that changing the gain does not change the filter characteristics?

Don't you watch the news? There's that whole embargo nonsense. They say it's for `election hacking', but maybe it's just econ-war or trump desperately trying to look non-manchurian-candidate.

Reply to
Johann Klammer

The ITAR silliness long predates Trump.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Perhaps just buy some from China, a good deal cheaper too:

Reply to
Clifford Heath

I had bad luck with a Maxim RTC chip (a ready-made module) from China. It had completely insane phase relationships between the

33kHz and 1Hz outputs. I was able to capture *any* phase offset by pressing the manual capture button on a scope. In a chain of asynchronous counters it should be impossible: the phase shift should be *some*, maybe temperature-dependent, but not cycle-by-cycle variable, even negative. A genuine part didn't have this problem. Still a mystery what the Chinese put there.

Best regards, Piotr

Reply to
Piotr Wyderski

Sure, and it can be seen on a scope. But the harmonics are n*76.5kHz, way above the audible range. And I hear the noise (well-behaved white noise) on a speaker.

Best regards, Piotr

Reply to
Piotr Wyderski

Yes. The price is suspiciously low in this case, but a lot of suppliers get partial reels after production stops, and you get genuine parts. The trick is to be ready to test on arrival, and dispute any bad parts - you get your money back quite easily in this case.

Reply to
Clifford Heath

There is a JFET (BF245A) preamp/impedance matcher.

But they are big, have problems with a decent value of Q and I am not sure about their long-term stability. Nonetheless, maybe my expectations are set to high: a resonant transformer wound on a toroid (for bandpass) + a 4th order lowpass after the mixer could be an optimal choice.

Insanely, compared to what it does. But since I needed only one and it also serves as a reference for other experiments, I decided to buy it.

Have no idea how to check it on a ready-made board:

formatting link

Best regards, Piotr

Reply to
Piotr Wyderski

In a the case of a typical opamp it would be a good explanation, but they claim "93dB Typical S/N, Single 5V Supply (Q = 1)". I'd guess it makes a fair share of the price (the other is the LTC logo...). So the other possible sources are the MCP6022 post-amp, the mixer within the psoc5lp processor or the audio amplifier. I have no idea how to check it with just a Rigol scope with a lousy FFT.

Best regards, Piotr

Reply to
Piotr Wyderski

Actually, 3,5,7etc x 76.5KHz - square waves should only include the odd harmonics unless the duty cycle is appreciably off 50%.

Sure, but the point about a mixer is that it mixes any broad-band noise on the signal going in with your 76.5kHz local oscillation, and produces low frequency side-bands from the higher frequency noise content and the higher harmonics.

--
Bill Sloman, Sydney
Reply to
bill.sloman

Cut tracks and bodge in LCR filters on all the power inputs.

If you can squeeze in tantalum capacitors, their equivalent series resistance is usually high enough to damp the kind of resonance you get with ferrite beads.

Ceramic chip capacitors can need a damping resistance.

--
Bill Sloman, Sydney
Reply to
bill.sloman

In a large order filter, stages go from high Q to low Q for the same reason they go from high gain to low gain: noise. This way they have large gain p eaks in the early stages that get rolled off in the later stages which also roll off the noise. If you did it the other way, you would squish the sign al down to the noise and then lift both the signal and noise in the later s tages. Because stages go from high Q to low Q and high gain to low gain, th e early stages have large signals with large gain peaks and require large e nough power supplies to handle these gain peaks. If you don't have large en ough powers supplies you will hit the rails and oscillate. To fix this you need to make trade offs and move some of the Q or some of the gain to later stages. This requires an understanding of the transfer function of the fil ter. Van Valkenberg has a good book on this stuff but it is heavy on the ma th.

Reply to
Wanderer

And adds its own portion of the noise. I am not sure if the mixer can be assumed high-quality, will try to invent a way to check it or compare with, say, an NE612. Or the mixer itself is OK, but the noise is injected through the power lines/couplings from the MCU core (the mixer is on the same chip as the rest of the digital world).

Theoretically there should be no considerable noise on the input signal (8th order Bessel + x40 amplifier), I suspect the problem is in the PSOC.

Best regards, Piotr

Reply to
Piotr Wyderski

I doubt they consciously sell counterfeit ICs or even that someone would ever try to imitate such a niche chip or even a Maxim RTC. The problem is deeper: the part can be entirely genuine, but didn't pass all the quality tests and a very inventive person has just found a way to obtain them and introduce to the market.

In the abovementioned case the RTC worked "correctly" and even would cause no trouble in a typical application circuit. It was my desigh which used it to the extreme: a normal RTC action, but use the 2ppm 33kHz output of the TCXO as an input for a dual-edge counter to create ~1/66e3 second time resolution and the 1Hz output to synchronize the fast counter with the one second resolution of the RTC. 3 days of debugging based on a false assumption that my Verilog is wrong: usually the count was OK, but sometimes

-1 and sometimes +1. In a synchronous circuit! It turned out the input was wrong, hence the GIGO. A part with guaranteed origin made it rock-stable.

You never know how much testing is enough, so its better to pay more. One year ago I would have happily bought the reel, now I have that fresh experience...

Best regards, Piotr

Reply to
Piotr Wyderski

That's not true at all. There's a long history of Chinese vendors selling broken, re-labelled or substandard parts, and sometimes just empty packages. I've only ever heard of it happening with chips that are in demand though NLA, which I wouldn't use anyhow. I've never encountered one on AliExpress, because I buy only products that are readily available from current production, and only from vendors who have happy customers. My first action on getting search results is to order by Orders, not price. I don't think a vendor of fakes would get very far on AE, that's all.

Possible, but I haven't come across that yet. Suppliers who want to hang around in a reputation-based business try not to do that.

Clifford Heath

Reply to
Clifford Heath

I remember having to fill out all sorts of export-control forms for shipping stuff to the former Soviet republics or Serbia or Saudi Arabia, etc. over a decade ago when I worked for an online gear retailer. Had to fill it out for very dangerous dual-use technologies like PC soundcards and DTS-encoding software

Reply to
bitrex

It's the package pin-to-pin capacitance itself. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Is there a nice ground plane on the backside? The oscillations with gain in the filter might be due a poor ground. Where does pin 15 connect to ground?

George H.

Reply to
George Herold

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.