PCB layout with diff pairs

30 ohms is still -12dB return loss for a 50 ohm line, which in many cases is still entirely reasonable if not great.
Reply to
Joel Koltner
Loading thread data ...

It is ok, as long as you don't face an intricate phasing or I/Q network. Then the production tolerances can cause a lot of hair loss. For example, in ultrasound we routinely have to be within 0.5dB and in systems production that can become iffy.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

I just love those RF guys. They make a 50% error, and are proud of themselves because the VSWR is only 1.5:1

John

Reply to
John Larkin

Well, you're probably looking at a ~DC to multi-GHz part, right? "Wideband" and "well-matched" tend to trade off pretty directly -- for narrow bandwidths you can of course build matching networks to get very, very close to 1:1, assuming you started with something better than, say, 5:1. (And at low enough frequencies you can play games with op-amps to actively make the input impedance pretty much anything...)

Ultra-low VSWRs are largely overrated anyway -- many automatic antenna tuners use 1.5:1, 1.8:1, or even 2:1 VSWRs as their setpoints for "good enough." :-)

To some degree I'd agree that a lot of commercial RF design does seem a bit sloppy though -- you can find little firms that make relatively minor modifications to commercial radios and gain, e.g., 10dB improvements in sensitivity, largely just by paying close attention to interstage matching, filter responses, and biasing.

---Joel

Reply to
Joel Koltner

The problem with time-domain stuff is that you can really see overshoots and pulse reflections rattling around in the system. Sine wave - reflected, inverted, phase shifted - still look like sine waves.

A good 50 ohm match is one you can TDR and see nothing but a couple of tiny connector wiggles maybe.

There are a couple of MMICS that are good wideband 50 ohm input matches. You can even trim the device current to nail exactly 50 ohms.

John

Reply to
John Larkin

The (traditional) RF designer's assumption of "the only thing that matters is the steady-state response" (after everything is done rattling around) has allowed many grotesque hacks to be quietly swept under the carpet via use of a clever matching network. :-)

The ultra-wideband RF guys pay a lot more attention to time-domain response of the entire system -- out of necessity.

---Joel

Reply to
Joel Koltner

I did that at a board meeting. All financial stuff. "What's your confidence levels in the bottomline you presented there?" ... "Plus minus 3dB" ... Eerie silence, followed by thundering laughter. Later my boss told me to never, ever pull that again. Said his heart almost stopped when I said that in front of all the guys with the expensive designer suits.

enough

Then there's transformers and the more esoteric distributed line schemes but that's usually in the domain of trade secrets.

Some transformers are pretty darn good in bandwidth:

formatting link

RF newbies are always amazed when someone sits there, bends tiny tabs and wire stubs and suddenly there is a straight line on the analyzer.

But probably only over a limited frequency range.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

I got a call from a biggish aerospace company engineer today. Our digital delay generator is spec'd at 25 ns insertion delay, and he's seeing 38. So I started asking him about cables...

DC to about 4 GHz for the one I had in mind.

John

Reply to
John Larkin

I had that a few years ago. Several hours of flight travel, followed by almost two in a rental car, because some test set never worked right. They had dutifully used BNC cables of exactly equal length but what seemed like RG58 was actually deceptively similar looking 75ohm cable. Whoops ...

May this esteemed group of newsgroupers learn which one that is or does it fall under trade secrets?

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

No, there it would be I = 0 = 120 watts PMPO.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

:-)

Nice... although is power proportional to money or money squared?

Agreed, I'm amazed at the bandwidths of some commercial offerings.

---Joel

Reply to
Joel Koltner

In the (usually locked) desk of a sales representative?

formatting link

Prices not being in the open usually means mid 4-digits and up.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

As Jim would say, "Bwahahahaha!"

I'm sure it must have required effort to stiffle some chortles. ;-)

Reply to
Joel Koltner

Yeah. Plus these days there's a large matrix of prices based on the number of routing layers, the licensing model, the optional components desired, discounts for multi-seat purchases, what country you're in (this can be quite significant these days), etc.

As of a year ago just the "high speed design" option for Pulsonix that supports automated routing of differential pairs (and a few other tricks) was $1,600 for a node-locked license. Pulsonix itself ranged from $1,950 (node-locked, no auto-router, 1000 pin limit, none of the fancy option packages) to $8,338 (network-licensed, no pin limit auto-router, still no option packages).

Prices gave gone up in the past year. Some of them significantly so, such as "Pulsonix SPICE," due to marked price increases in SI-Metrix, which is what Pulsonix licenses for their SPICE.

Hmm... Eagle is looking more and more attractive all the time. :-)

Amusingly, when Pulsonix began the schematic capture portion (only) of the program was 100% free (presumably as a "loss leader" to get you into buying the PCB tools or SPICE). Last year it was $563 (node-locked)! Wow...

I doubt Leon would still be using Pulsonix if he were paying for it out of his own pocket... it's definitely moved into the pricing territory where you pretty much have to be a regular company to afford it.

---Joel

Reply to
Joel Koltner

SGA-3586.

John

Reply to
John Larkin

formatting link
? The data sheet (with 35mA bias current) shows ~11dB return loss over most of its operating range. You're saying that by starving it a bit the return loss improves significantly?

Reply to
Joel Koltner

Thanks! Entered into the local wiki file.

Available at a very reasonable price no less, and the fact that Digikey stocks over 10,000 of them means it sure ain't a boutique part:

formatting link

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

I remember the DOS days when they wanted to charge twice as much for a dongled OrCad through a "value-added" reseller in Europe versus the non-dongled version in the US for $495. So ...

It would IMHO be an almost perfect CAD suite if they hadn't bungled the hierarchy (there is none). That is the reason I want to leave Eagle behind but I am in absolutely no hurry.

Per coincidence I found that AutoTrax is still maintained by Kovac in Santa Monica and now runs under Windows. They say they'll even come out with a new version, DEX. I wrote to them on the weekend with a couple questions. No answer. Hmm ...

If node-lock means tied to a MAC address I will never buy it. I expect business people to assume the same honesty in me that I'd like to assume in them. Else it is IMHO not wise to enter into such business.

Yeah, if Cadsoft wouldn't have made that crucial mistake not providing a hierarchy their engineers could be rich now (if they have things like stock options over there).

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

Money = Power

At least that's what a venture capitalist told me. In business I agree with him, in personal life I have a different take on that but this would get into religion now.

There is a lot of voodoo involved. I don't know much about audio transformers but whenever I design a signal transformer that has to cross a high voltage barrier, drawing the fab instructions takes more time than anything else. If they don't follow this by the letter it ain't going to pass final test. This once almost drove a production technician up the tree.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

USB dongled, actually. Sorry for the confusion -- "node locked" does tend to imply something like MAC address checking.

I think Pulsonix is still a good value (and they do support hierarchy :-) ), but they've definitely moved out of the "hobbyist" category that Eagle can still cater to (although $995 for full-featured layout & schematic is definitely starting to get up there...). Too bad about their not supporting hierarchy...

I think the longest outstanding "potentially big change" in Pulsonix that many people want to see is a scripting language.

---Joel

Reply to
Joel Koltner

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.