>>4. A 74HCT40103 costs more than twice as much as a CD4020, both
>>from Farnell.
>>
>>There are a number of ways in which the same effect could be had
>>differently. But that doesn't make it "nuts" to use an older
>>device.
>
>---
>Indeed.
>
>Since you're fairly new to the group and haven't interacted much, if
>at all, with Sloman, you might not be aware that his agenda isn't
>supplying technical expertise, and help, it's self-glorification.
Likewise Larkin. Both best killfiled. Likewise those that feed them ;-)
>Sadly, for him, his efforts fall short because he's easily found out
>and his cause derailed.
>
>---
>JF
...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Remember: Once you go over the hill, you pick up speed
Maybe not with _one_, no (well, it'll count from 0 to 1, but I think that doesn't count, excuse the pun.)
_But_, with careful selection of operating frequency and timing components, they _can_ be used to frequency-divide, over a fairly small (clock freq.) range. And what's a counter but a string of dividers? :-)
The 4020 is a ripple counter, and thus pretty much a string of dividers. The 40102 and 40103 are synchronous counters, which means that the individual bistables all change state on the same clock edge (if they are going to change state - which depends on the states of the outputs of the lower order bistables).
Neither can be equated with a non-retriggerable monostable, which is what Rich would be thinking about, if one wanted to dignify his ramblings by calling them "thoughts".
Jim Thompson kill-files everybody who doesn't share his silly ideas. He does know stuff about designing analog integrated circuits, though the times that I used parts that he designed aren't amongst my happiest memories. He seems to be like Linus Pauling in being able to get 90% of the way to a good idea, so the 4046 is lot easier to use than his 4024/4044 combination, and Barry Gilbert's four quadrant multipliers are much easier to use than Jim's MC1495, while the Analog Devices modulators are rather more popular than the MC1496.
On every other subject he comes across as an ignorant hill-billy.
Kill-filing him, rather than John Larkin, will lose you a lot less useful information.
Who cares about jitter when using an RC oscillator??? A ripple counter is a drop in the bucket.
State decoding can be a bit more challenging when using a ripple counter, on account of glitches, but there's a place in the world for the 4020, especially when it replaces a pair of synchronous counters that cost three times as much--a cost saving of 83%, not counting the board space. There's also several times more stock available of the
4020, I notice, which is very comforting.
I still use HC4017s for simple state machines, too, especially in situations where I don't want a micro around for signal integrity reasons.
I recently did a really old-school analogue board to lock a $1k DFB laser to a temperature stabilized etalon. It uses a tri-wave sweep for lock acquisition, an analogue switch to open and close the loop, and a bunch of amplifiers. It even uses class B push-pull drivers for the thermoelectric cooler. It acquires lock using temperature alone, and then switches to temperature + current tuning to get 500 kHz lock bandwidth. I'm building another one today, actually. Doing it in analogue improves life by a lot. (It does have an ATMega88 in one corner to supervise and tweak a setpoint or two, to make sure both the temperature and current loops stay in range.) The result is a laser good enough to hit 1-10 ppb drift over 8 hours, plus a greatly reduced linewidth. (I don't know how much narrower yet, but it should be about 4 orders of magnitude.)
So I'm a big fan of the old school for some jobs.
Cheers
Phil HObbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
There's not a lot of skill required to follow the operations of your mind.
Why would you think that CO2 is going to kill us all? It does seem to be the engine behind anthropogenic global warming, and if we lett that warming get progessively worse for another century or so we might be able to provoke some kind of event like the Younger Dryas or the Paleocene-Eocene thermal Maximum, which might well cause the human population to crash, but that isn't going to kill us all - since we will all be safely dead by then - though it might drasticallly thin out the descendants of those of us who have descendants.
It could be used in a circuit which might act as a frequency divider over a narrow range of frequencies.
Why would I need to? I already knew what you were talking about.
I believe so. The Dutch medical insurance system recently removed Freudian psycholanalysis from the list of treatments that they'd pay for, but only because it doesn't work. Are you contemplating immigrating?
The synchronous counters are likely to work at much higher frequncies than the 4020, which runs out of puff at 4MHz even when the supply voltage is 15V - 5V only gets you 1MHz.=A0
The 4020 is a legacy part, like the 555. People who know enough to use faster parts usually know enough to have moved on from there to programmable logic a decade or so ago.
.
A CoolRunner programmable part will use as little current, and could confine all the switching logic within one package, which might do an even better job of keeping radiated noise out of the rest of the circuit. It will be faster, so you might need some RC filtering on the outputs to slow down the edge speeds to those you'd get from a 4017.
That's belt and braces. My thermoelectric temperature controller filtered the hell out of the class-D switching drive at the driver, and swapping it for the Class-AB (LM12-based) bridge we used in the proof-of-principle system didn't amke any obvious difference to our noise levels.
Physicists do have this enthusiasm for ancient circuits that they know to have worked - the investment required to keep up with the state of the art is hard to recover on devices being built in limited numbers. I've published a comment in Review of Scientific Instruments berating an author for getting excited about the speed offered by 10K ECL at a time when ECLinPS was widely available. They emphasised - IIRR - that
10K was roughly four times faster than TTL when ECLinPS was roughly four times faster again.
krw can't get anything right. I'm not afraid of the 555 - I just don't think that is a good solution for most of the problems that it is used to solve, and I think that this has been true since around 1980.
And I'm certainly not afraid of carbon - it's pretty much my favourite element. I don't happen to think that the atmosphere needs any more CO2 than it has at the moment, but it isn't the CO2 as such that worries me, just it's effect on the surface temperature of the planet.
So what? This is engineering, not fashion design. If I need speed, I'll pay for it if necessary. But if not, why pay six times the price and twice the board space for nothing?
C'mon, Bill, is name-calling the best argument you've got? People who build in needless cost get their lunch eaten by their competitors. And that doesn't just mean per-piece cost.
Besides, I associate 'legacy' with happy events such as long-lost uncles leaving me a lot of money. ;)
Even if that were true, so what? Slow is beautiful when you don't need speed. CoolRunners don't cost pennies, and 4017s don't need yet another development system. Besides, they'll be around longer than either of us.
You say "belt and braces" as though that were a bad thing, whereas it's one of the keys to making very hard measurements solid and reliable in the face of uncertainties. If I had a belt-and-suspenders solution for all the problems this customer runs into, I'd be a very happy guy. Cheap insurance is a wonderful thing--nail the problem to the floor and move on. There are always ten more coming.
You won't catch this trout with that mouldy bit of bait, sorry.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
Getting an old ripple counter to do what you want is always a little messy. If you want the service engineers and the final test guys to understand what is going on, more transparent design can be handy. If you are designing one-offs that only you are likely to have to fix if one of the parts turns up marginal when you put it in the circuit or after few years in service, this is a marginal benefit - you can usually remember what you had in mind.
Designing for even low volume production makes transparent design more attractive.
Precisely. And the cost of fixing stuff and getting it working is one of those costs.
You wouldn't if you'd had to wade through a few legacy designs.
ons.
r
s.
Coolrunners don't cost all that much and the basic developments systems aren't expensive either. I got the same argument when I started work at Haffmans BV in 2000, and didn't bother argueing with the boss - why should I care if he was prepared to pay me and a technician while we spent a week or two stitching together a pile of CMOS to do a job I could have managed in an afternoon with a CoolRunner. You don't need to spend much time at all finding errors in hand-wired boards before you've paid for the development system several times over.
for
"Soldi and reliable" is what you've done already, so you know that it works. Carry it to its logical conclusion, and the whooe system is made of brass and ivory and riven by steam.=A0
Perfectly true. But class-B push-pull drivers for Peltier junctions waste a fair bit of heat, so you've got another set of heat-sinks to squeeze into the space you've got - class-D drivers are at least efficient, so they do kill off one set of problems, even if they do make bigger demands on your RFI skills.
4
It's not a "mouldy bit of bait". It reflects a real conflict in instrument development. We are both agreed that doing a job the same way you did it last time is easier and quicker than taking advantage of the latest technical development, and often gives you something that works just as well as a solution that took advantage of something new - albeit usually in a bigger package, or with a worse thermal management problem or whatever.
The problem with the authors that I berated was almost certainly that they didn't want to built circuits with surface mount devices, and ECLinPs is only available in surface mount, probably because the packages interface micely with microstrip transmission lines on the printed circuit board - you've got to spend money on the tools you need to put the boards together, and nobody wants to do that, but at some point you've got to raise your game or get left behind.
What I've kept in mind when I've been making my - mostly conservative
- choices has been that at some point, dumping what you've done before can move you up into a rather different ballpark.
I had several reasons for using an A/D converter in my Peltier- junction based milli-degree controller, but the best reason for doing it was that putting the control loop into the digital domain meant that it was practical - in fact easy - to keep the feedback loop critically damped as the watts per amp coefficient of the Peltier junction went up and down by a factor of seven as the temperature difference across the Peltier junction went from one of the specified range to the other. The joke is that I wasn't conscious of this as an advantage when I made the choice ...
I can't remember ever having a problem with a ripple counter. What problems are you referring to?
You're arguing my side now. A 4017 on a schematic is a lot more transparent than a programmable logic block. If they're doing the same thing, what's the advantage?
Have you ever built a board where a ripple counter didn't work? Have you ever even heard of one? So where's the cost? You don't have to debug MSI parts.
I rarely need to repair things, except for the occasional spectrum analyzer. I also almost never do maintenance on old designs. But if I did, it would be much more pleasant if I didn't have to exhume some wart-encrusted development system that only runs on Win 95 and needs a parallel port dongle that doesn't work any more. I recently bought an EPROM programmer--something I thought I'd never need again--specifically in order to copy and archive the contents of the EPROMs in some of my Tek and HP gear.
Digikey's cheapest Coolrunner that runs on at least 3.3V costs $4 in quantity 100, whereas a 4017 costs 15 cents. (And the 4017 is synchronous!)
Painful memories at work. I sympathize with that part--I've had PHBs too. But that has nothing to do with the facts of the case here. I like programmable logic for some things too--I still have a box I built
20 years ago, with a socket for a 22V10 and breakout pins for controlling various other things. I haven't used it in awhile, but sometimes programmable logic is just the thing.
Less so nowadays, though--processors are so cheap and so powerful that unless you really need hardware timing accuracy, it's usually cheaper to do the Jan Panteltje thing and use a PIC or the equivalent. I have a small pile of development systems for various modern micros kicking around, from 8 to 32-bits.
Another moldy lure. No thanks. Performance is what I care about.
This particular board runs the laser right around room temperature, so there's no big power dissipation--maybe 700 mW total for the entire board. That's one of the big advantages of Peltiers in instruments--you usually don't have to have a big delta-T in order to get good control. The board runs off a single -9V supply because the laser is common-anode.
I didn't agree with any such thing, because it's nonsense.
"Raising your game" doesn't mean learning a new programming language every six months, for instance. That's just churn and fashion. Learning how to do really new things is another matter. Knowing 10 ways to do something is usually a win--provided you actually get something working in the time required. Otherwise it doesn't matter at all.
Sure. Of the last three temperature control loops I've built, two have been microprocessor-controlled. Why not? For most things, it's better than good enough, sometimes it's better period, and since you get the ADC and PWM for free with a $2 micro, it's almost always cheaper. You use a PWM-controlled heater, and avoid hunting by dithering the PWM ratio or extending the number of bits via a timer interrupt handler, like a dual modulus prescaler.
If all you need is millikelvin stability, that works fine. The limits are mostly due to thermal forcing anyway, on account of the limited bandwidth you can get with most temperature sensors. Besides, with a micro you don't have to try to make 1000-second RC time constants.
In the present case, it would be a disaster, because I'm not trying to maintain a set temperature but a set wavelength. Which is a very different kettle of fish, especially with the extreme SNR and (especially) 1/f and drift requirements. (The sensor's operating bandwidth is 10**-4 Hz to 100 Hz.) To get enough loop gain to suppress the natural line width of the laser and the thermal forcing, I need about 500 kHz closed-loop bandwidth, which is a bit faster than your average temperature controller.
A nice op amp like an OPA378, which has 40 nV noise asymptotically constant as f->0, is a dramatic improvement over any ADC you can name for this job. Its set point does need a bit of tweaking from the micro very occasionally, because the current tuning sensitivity of the laser is so very low that I had to put a gain-of-5000 amplifier between the two control points in order to get the bandwidth I needed. If the current tuning loop pegs, the temperature loop continues unperturbed, but the loop bandwidth drops by six orders of magnitude. So it had better not peg, hence the micro. The tweak is done by a quad dpot in a sort of Kelvin-Varley configuration, so the micro can run very slowly and spend most of its time asleep.
The point is that performance and cost are what count, not fashion or prejudice.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net
The kind of problem that I had in mind were out-of-spec parts that weren't far enough out of spec to give problems in final test, but did give problems when the machine had been running long enough to get properly warmed up - one was a Texas Innstruments shift register in a TTL-based shift-or-add multiplier and the other a Hewlett-Packard logic-output opto-isolator. I've not had a problem with a ripple counter yet, but I'm pretty careful about doing worst-case timing analysis. I've cleaned up after other engineers to be aware that not everybody is this careful, and on one occasion I managed to miss a crucial timing chain in a draft circuit that I had to turn over to a less senior engineer.
They aren't if you know how to program programmable logic. Programming languages for PLD's aren't exactly high level, but they are higher level than tracks on a circuit diagram.
Less board space, fewer connections, fewer tracks to get shorted or open-circuited, faster, more easily modified if you need to upgrade or correct a drop-off.
e
o
And
I certainly built one board where the ripple counter wasn't working the way I expected - it was very much an experimental prototype, thrown together to see if the idea could work, which it did, but my worst case analysis of what I'd built turned it into a slower, if very reliable product. You do have to debug MSI parts, and their connections. Not every part you get from the manufacturer meets it's specification, and production are remarkably ingenious in finding new ways of screwing up connectivity.
es
Lucky you.
We used to have to redesign them to get rid of parts that had become obsolete. =A0
One of my happier memories is of the day when the UK rep for ICT pointed out to me that the PA7024 PLD was a drop-in replacement for the 22V10. I'd inherited a design done by a guy who had departed to become a lecturer in Manchester, whose academic expertise hadn't extended to doing worst case timing analysis. I was able to rescue his board by - amongst other things - turning a couple of the extra buried bistables available in the 7024 into a short delay line. These days you'd need an SMD-to-DIP conversion socket to pull the same kind of trick.
asons.
e
eed
ther
f us.
ous!)
But you have to decode the 4017's outputs to make a state machine - apple and pears.
Getting genuine parallel processing out of a processor is trickier, and special purpoe hardware can be a lot faster than cheap micros. In 1992 we wanted a microstepping controller, and it turned out that we needed a - cheap - Tranputer to control. At that time regular processors just weren't fast enough.
p for
a
he
s
d
u
in
he
er
ed
ut 4
w
You've been lucky.
Agreed, but that a straw man exaggeration of my position. The counter- example is John Fields' persistent affection for the 555, which was effectively obsolescent in 1980. =A0
s
The advantages of doing something slightly different aren't always immediately obvious.
er
I was still able to get away with putting that into the milli-degree temperature controller paper, which was published in 1996 probably because the whole idea of using PWM to drive a Peltier junction sitting right next to a milli-degree thermistor sensor was new to the instrument literature.
Compare and contrast
Bradley C C, Chen J and Huet R G 1990 Rev. Sci. Instrum. 61
2097=96101
Barone F, Calloni E, Grado A, de Rosa R ,di Fiore L, Milano L and Russo G 1995 Rev. Sci. Instrum. 66 4051=964
Don't pay much attention to Barone et al's performance claims - they'd obviously never heard of thermocouple voltages.
Perhaps. But it probably isn't anywhere near as good as an AC-excited thermistor, where the crucial noise frequency is at the AC excitation frequency rather than down in the 1/f region. Crystal Semiconductor used to sell a sigma-delta ADC that was designed to work with "reversing DC" (as mentioned in my 1996 paper, but we couldn't use it because it needed bipolar rails), and at least one of Linear Technology's "24-bit" A/D chips is set up so it can work when digitising an alternated signal.
Unfortunately, all our design decisions are driven by fashion and prejudice - from what we have fashioned earlier and the prejudices we developed on the basis if that experience. Checking out a "crazy" idea from time to time can be theraputic.
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.