Analog multipliers and watt measurement

Sounds like misapplication to me... but it figures... hasn't TTL been out of favor for 20-30 years ?:-)

So he just acts that way ?:-) ...Jim Thompson

an

te:

rley"

ribution of the

uency and span a

metering. My stuff

as well as it

to reduce that

ANSI C12-class

ff about that

u to was on

in a journal

learly be

so

some

ysis.

our

I may be useless, but John Larkin calling anybody else a pompous fathead is remarkably ironic.

I disagree with you so I must be claiming to be an expert? I certainly don't claim to be an expert on "customer bases" and I'd distrust anybody who talks about the people who buy their stuff as if they could be usefully divided into broad-brush catagories

Stuff I've never done? Where I've worked the engineers have developed stuff that the customer wanted, more or less despite marketing department who wanted us to develop the stuff that marketing thought they wanted (first and foremost with a more exciting colour scheme).

You, at least, have the advantage of being able to talk to the customers directly in the course of your work. We mostly had to exploit opportunities to talk to them at conferences and exhibitions.

John Larkin does enjoy his comforting delusions.

Backplane buses may have vanished, but there are still busses on the motherboard, connecting your gigabytes of RAM to the processor/s and its/their caches.

A bus stops being a bus if it doesn't get off the motherboard?

PCI Express isn't actually a bus, it's a serial

Granting how crummy the PCI bus was, one can see why they'd go for fast serial links.

VME was a well-designed and extensible backplane bus. We took advantage of the extensibility back in the 1980s, and I'm not surprised that people are still exploiting it.

-- Bill Sloman, Nijmegen

They didn't let you have telephones?

I design stuff that works and sells. There's no room for delusion in that process.

It's academic if you can't access it. You could once design boards that connected to ISA or PCI, and you could write code that ran on the PC and could access registers on your boards. PCs barely have slots any more, and they are (for now) PCi Express, with absurd complexity and latency.

The only way to interface to computers nowadays is through USB or PCIe, which were designed for system use, high throughput, horrible complexity, and horrible latency.

Even many "embedded" chips, like ARM and PPC, emphasize DRAM and cache as their memory model, and make it difficult, often impossible, to do a memory-mapped realtime interface.

It takes about a microsecond to do anything on PCIe, like read one byte for example. At least PCI would do a 32-bit parallel transfer in real time.

Oh, I thought you were dismissing it as antique.

Multibus is gone, ISA is gone, PCI is almost gone, Thunderbolt will trash PCIe. VME is doing fine.

I just got an order for 20 VME boards that I haven't designed yet. LVDT/Synchro simulators. The board will be mostly covered with transformers, hardly any room left for electronics. The Brat wants to put the transformers on top and all the electronics on the bottom. Yuk.

John

They got a bit picky about international calls, and since we rarely got to meet the customers, getting their phone numbers could be tricky.

As you keep on telling us.

Not about what you design and sell, perhaps. You manage to be remarkably deluded in many other areas.

Presumably the instructions telling your hardware what to do have got to come from a computer, and the data your hardware generates has to be eventually fed into a computer for storage and examination. How much latency does it take to mess up either of these operations?

But presumably you can still use them to generate and read serial data strings.

My interactions with the software engineers started with them refusing to promise response times of less than a second, and got better when some of them got into real-time operating systems, and could promise millisecond responses.

It's certainly old. When we wanted to move lots of data around reliably, we planned on using balanced ECL signals - roughly the same voltage swings as LVDS on a 72-bit wide backplane ( 8 error detection/ correction code bits tacked onto a 64-bit word). The VME bus wasn't good enough.

We never got to build it. At precisely the point when we were ready to get boards laid out, the project planning software got its act together and told management how much money and engineering effort was going to be required to complete the project as a whole. The money wasn't a problem - it cost us just as much to buy off our contractual obligations to finish the project as it would have done to finish the project - but the engineering effort would have tied up every engineer at Cambridge Instruments for more than a year, and the company couldn't afford to have all its engineers tied up in one project for that long. Extending the development time wasn't an option.

Transformers traditionally stick up far enough that you might start thinking of plug-in daughter-boards between the transfomers; they might also be a useful way of making the board stiffer.

Printed windings could be fun.You can't get as much copper into printed windings as you can into copper wire or copper tape, but it seems that you can get quite close.

-- Bill Sloman, Nijmegen

That's crazy. When I was 20 years old, my employer would take me on trips to meet customers and sell systems to them. How else would I know what they need? And what customer would believe marketing droids who don't actually understand the technology?

Well, I keep designing stuff. This is s.e.d. The topic is designing stuff.

What difference would that make? Heck, most people (and most professions) are deluded about stuff they don't get real feedback on.

This week, two microseconds is borderline intolerable.

John

Of course it was crazy, but the English don't really believe that engineers have any place in polite society. There was a civil service aphorism - scientists should be on tap, not on top - that describes the attitude pretty exactly.

The Technical Director of Cambridge Instruments when I joined was a very intelligent man, with a degree in pure mathematics. He could understand anything put to him as a self-contained proposition - usually better than the people who had constructed the proposition - but he didn't understand detail work. Most of the managers I worked with were rather less intelligent, and even more ignorant of what the engineers were actually doing or needed to know.

But you don't need to keep on telling us that you design stuff.

And you get upset, aggressive and unpleasant when you get real feedback.

To whom? If it's just you, you can always stick a processor into your system. If it's your customer, they've got to offer a fast enough interface.

-- Bill Sloman, Nijmegen

Your opinions..

But you pontificate on off-topic subjects about which you have no useful feedback.

The fastest interface available out of their PowerPC controller is PCIe over cable. That's the problem. Latency is close to a microsecond to do register i/o, impossibly slow. Even DMA has to be managed very carefully to have a chance of working. That's how "computers" are nowadays.

John

ns.

If you bothered to check my "opinions" you'd find that they are based on solid evidence - but you prefer your own opinions to any kind of evidence.

I read widely, and that does tend to discipline my opinions. Your reading seems to be confined to feel-good nostrums for right-wingers, which manages your opinions in a way that may suit the more lunatic fringe of the Republican Party, but does leave you rather out of touch with reality in areas not directly involving electronic hardware.

he

nd

the

That how cheap computers are nowadays. There are situations where slightly more powerful computing hardware might be appropriate.

-- Bill Sloman, Nijmegen

Such as? The more powerful a computer, the less likely it is to have a parallel interface bus or slot available. PCIe, USB, and Ethernet have big latencies. You have to go down to little embedded-type chips to interface directly to hardware or to an FPGA.

John

tions.

ds

n.

n the

s and

ds

on the

ots

xity

ible

got

to

Do your own research. Or pay me if you want me to do it for you.

-- Bill Sloman, Nijmegen

exhibitions.

the

Pay you? What a hilarious concept.

You are really, really unclear on how the world works.

John

he

to

ibitions.

rely

on

roids

ice

the

ing

k on.

s on the

or/s and

d?

oards

an on the

slots

plexity

B or

orrible

ve got

as to

How

our

ond

People do work for you for nothing, just because they really like you?

It isn't me who is out of touch with reality.

-- Bill Sloman, Nijmegen

exhibitions.

the

and

on the

slots

complexity

horrible

got

No, I pay them well because they are useful, and because I really like them.

John

No, when they were doing their power metering, that is what they intended to do. Then along came the internet and the epiphany was that the internet was going to be huge, and there was a place for wireless in that.

That is, it was a vision change, not some juke and an excuse. The company's name reflected it's intent: Metricom is a contraction of Meter communications. Bob Dilworth thought the internet would change everything. He re-steered the company.

Later, Schlumberger licensed all the technology for metering (UtiliNet) -- they took it over totally before the train wreck. The metering business actually made money, just not a lot of it.

It scaled just fine. Metricom didn't fail because of the technology.

No, they didn't. They were dead before WiFi and cell phone networks ever came into their own for data transmission.

I've never been a ham. I did do this one:

Well, I met with Dilworth and Baron a couple of times, while the meters were still serious stuff, and that's what they said. They did want meters in every house, but the electrical metering was just an excuse to get in and do other stuff... I'm not sure what, maybe home automation, maybe internet. I think that's why they did the metering badly; it didn't seem to really interest them.

The meters were too expensive.

It was, what, 38 kbaud if everything was right?

What is that? Looks expensive.

The things I'm saying here are what Baron and Dilworth and some other guys told me.

John

That's Gen 1 Ricochet speed.

Gen 2 Ricochet was ISDN (128bps) speed. Practically, I would regularly get 200+ in San Jose. And it was always on. (Remember that it was the time when a lot of people were still dialing in for 56k.)

Real work was being done for Gen 3 Ricochet @ 1Mbps. These speeds may sound slow now, but it certainly wasn't then for wide area wireless data networks.

Now it could be that it may have failed technologically eventually. But we'll never know.

The 900 MHz transceiver in the poletop/ethernet microcell.

It was, I suppose.

We were making good progress stripping costs out. What you see is the first cut and "just make it work, and ignore costs" was the first edict. We were to take costs out later. But they built more than