Phase detector with a MC1496

Buy a better phase detector. The MC1496 was designed by Jim Thompson a very long time ago.

Analog Devices have a much faster device, better suited to higher frequencies - the data sheet says up to 200MHz

Try reducing signal levels, your 500mv pp maybe too hot, data sheet suggests 60mv rms (ca 170mv pp) ?

I had no idea Jim invented the MC1496 - up until now. Given the deteriorating relationship between you two towards the end, it's very magnanimous of you, Bill, to recognize Jim's achievement.

Thank you for the interesting thread, Jean-Pierre. Even old, obsolete devices have something to teach me. If nothing else, a little hands-on is better than textbook theory or simulation. Plus, you never know. Sometimes old becomes new again. For instance, I remember talking about AI with a small group of enthusiasts back in the early 1980s. And now look at it. Speaking of AI, does anyone know how to coax AI into creating art? Or is AI art an add-on you're expected to pay for?

I suspect he didn't. It's one of Jim's versions of the Barry Gilbert circuit, and the Analog Devices variations on the idea were a whole lot easier to use than Jim's - and I did use them from time to time at Cambridge Instruments.

The relationship wasn't helped by my low opinion of the MC1495 and MC1496. His claim to have reported me to the FBI for being dangerously anti-American didn't help either - and my pointing that I'd been security cleared to a level that had once got me into US Army ECOM at Fort Monmouth, NJ won't have made him any happier.

The latest IEEE Spectrum has some revealing discussion. Currently AI is designed to produce output that the users like, rather than output that is accurate - rather like Donald Trump. There are optimisation schemes that skew the balance more towards accuracy and thus may improve the long term bottom line.

Getting AI to produce output that the users find attractive might be seen as creating art. Making it more useful at the expense of making it less attractive is the art-versus-engineering dilemma.

It is an ancient part, barely adequate at 10 MHz.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

So tell us about a more modern part with better performance.

Front that with a dual LVDS receiver if you have a low level input. John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Am 16.10.25 um 19:18 schrieb john larkin:

I have packed the innards of the 9901 into a little corner of a Xilinx Coolrunner. The XOR has has nonlinearities of Kp exactly where it hurts the PLL performance most: in the middle of the phase range. The 9901 moves that the to start/end of the phase range, which are never visited in locked state. The 4 FFs & their add-on gates are about half a page of VHDL.

Cheers, Gerhard

My FPGA guys have done that too, or something close. But they were trying to build a PLL with a VCXO, which is maybe not the issue here.

We don't really know what Jean-Pierre wants to do.

My favorite picosecond-jitter PLL phase detector is an ecl flip-flop working in infinite gain bang-bang mode. Works great. Academically frowned on.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

I used the MCH12140 in a low-noise PLL and found that it suffered from the occasional meta-stability hiccup. This mightily upset Jim Thompson. I solved the issue by pulling the phase slightly to one side. A small constant phase error was more acceptable to me than the occasional bump.

Jeroen Belleman

I tried but the useful signal decreases to much!

Thanks.

The MC1496 is essentially a four quadrant multiplier - the MC1495 - with some of the linearisation components left out to make it faster.

It's still got much better linearity than an exclusive-or gate.

Again, not a particularly linear part.

You can build your phase detector out of discrete transistors

E.A.Faulkner and D.W.Harding published a scheme in J.Sci. Instruments volume 43 from page 97 in 1966.

E.A.Faulkner and J.B. Grimsby published a slioghtly more elaborate scheme in Electronic Engineering in volume 39 from page 565 in 1967.

Monolithic dual transistors would help, and there are now some wide-band discrete parts around which would work at higher frequencies. And even some arrays including a matched pair.

This sounds a bit like the problem that the 74HCT9046A was designed to solve. The digital phase detector on the 4046 had a dead zone inside which the phase could drift around. The 9046 modified its digital phase detector so it didn't.

For an xor fed with square waves, the average voltage out vs phase difference in is a beautiful triangle.

A 1 ns CMOS XOR costs us 25 cents. A slightly slower one is 12 cents.

The only thing that keeps one from using XORs or D-flops or multiplexers as phase detectors is mental rigidity.

They make great, cheap comparators.

Not stocked. $12. How many would you need? Hilarious.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Am 17.10.25 um 19:38 schrieb john larkin:

You can always use a doubly balanced Schottky ring mixer in saturated mode if everything else fails. No surprises.

I have used these in my dual slope 1:1000 pulse width magnification gear box with ps resolution. One of the last few RF PNPs that still are available.

I even could get them space qualified in these ceramic flat packs. They are now happily circling around the earth on the ISS.

Cheers, Gerhard

My fave RF PNPs are gone. Pity that phemts and GaN and such are always n-channel.

I miss BFT92!

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Silicon has the lightest holes of any semiconductor, iirc. Dunno about germanium, but all III-IV systems have super heavy holes. So no complementary devices to the BFP640 or SAV-541 are in the cards. :(

I have three or so reels of BFT92s in stock, but that’s no use for client work.

Cheers

Phil Hobbs