that PLL again

Earth is cool. You can find obscure roads and trails and hills, old railways and such, see the elevations, all that. We're considering moving the factory, and Earth is fabulous for looking at new buildings and neighborhoods.

It's also fun to explore other cities and countries with Earth. And I look at the sites of potential vendors and competitors (along with Street View) to see how serious they are.

I finally got fed up with AT&T DSL (800 Kbps in dry weather, approaching 0 when it rains) and expensive phone service, so I signed up with Comcast. So we have internet, phone, and teevee all in one hose. It's been pretty reliable so far. They claim I'm getting 50 Mbps Internet, and I'm actually seeing about 15.

I don't watch teevee, but my wife does, sometimes. I figure that, in 5 or 10 years, there won't be classic scheduled-time TV; everything will be Internet streaming, and Comcast will only really provide DSL. No wonder they hate Net Neutrality.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

True. But for finding trails of any kind regular satellite view is fine for me. Elevations and steep hills come as a surprise, almost everything that hurts builds strength. Maybe I'll try Google Earth when the new computer gets here. The GT-720 card should definitely work.

I "upgraded" AT&T from 1.5M to 6M and get 3M on a good day. But after starting a long PDF download it always gradually falls back to the old speed. Smells like they oversold their server capacity. After raising a big stink it gets better for a while and then falls back. Not cool. I'll have to call them this week for a discount.

Cable TV Internet is expensive out here unless you also take their TV bundle. No way. Then we have Calnet RF but we are right at the fringe and afraid it might fail at times. I can't tell my clients to push the online conferencing session to Wednesday because it rains right now. If AT&T won't sober up soon I'll do the switch though and maybe build myself some massive directional antenna.

--
Regards, Joerg 

http://www.analogconsultants.com/

[snip]

I design stuff for "smart" phones, but wouldn't own one myself. And I will never go near anything labeled "social media".

We have cable, but only watch ~6 channels. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142     Skype: skypeanalog  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

I stopped by the Cambridge Microcenter store, a huge store the size of a supermarket, and discovered their standard price is the same as Dell's special Cyber-Monday price.

Guy at the computer store claimed lots of engineering software does its math calcs in the GPU, rather than the CPU, so a more powerful graphics card greatly speeds up stuff, 3D rendering, etc. I asked about Dell's $1600 XPS 8700, which uses an AMD Radeon R9 270 ($170 at the store). He said that was a start, but claimed he had engineers coming in to spend $600 to $3k on GPU add-on cards. That had me floored, dunno what to think.

--
 Thanks, 
    - Win

That may all be true, but LTspice is not on the list of software that works with the GPU... at least last time I checked.

--

Rick

What JL said, if you're a Windows user. FoxIt is light weight and "just works" while CutePDF is a print driver for Ghostscript (it receives the Postscript output of standard Windows PS drivers) which produces better PDF files than you get from Adobe's. The Adobe software is terribly bloated; more than an order of magnitude bigger than required.

Personally, I now use a Mac, and PDF is one of the things built into the core display rendering engine and hence the Preview app. Combine that with system-wide full-text search and retrieval in Spotlight, and it's a no-brainer. There's a lot not to like about Apple's behaviour, but they did get PDF right.

Clifford Heath.

On a sunny day (Tue, 02 Dec 2014 16:28:10 -0800) it happened Joerg wrote in :

Its true, you only need a tent and a warm climate and a fire in it to keep the mosquitos away:

You can read Dutch right? Download the 2 free .pdf files, makes good reading. It is translated from French BTW.

On a sunny day (2 Dec 2014 17:39:18 -0800) it happened Winfield Hill wrote in :

Buy a Raspberri Pi, that probably does all you need. :-)

On a sunny day (Wed, 03 Dec 2014 13:43:54 +1100) it happened Clifford Heath wrote in :

I use xpdf in Linux, works on most files. Flies on a good graphics card, slow on AMD build in graphics chips.

The huge graphics card will also consume almost as much power as the rest of the machine put together which is a waste if you don't use it!

Some applications can use the GPU to offload certain types of calculation but unless your application supports it you have a high performance and expensive piece of hardware idling to no good end.

That is par for the course. Gamer benchmarks are all 3D.

It isn't such a bad tactic to select a reputable gamer machine and tell them to delete the graphics card (they will take some persuading). You get a frugal machine that runs cool and has ample 2D graphics - i7/i5 on chip 2D support is generally faster than the 3D cards can manage.

Theoretically you are chewing up some memory bandwidth with the shared graphics memory but the speed hit incurred is only around 5%.

Spend the money saved on extra ram or an SSD or gofaster SSD cache.

You might have to live with a moniker on the side along the lines of Barbarian Megadeath Exterminator or some such but the price performance available this route is extremely good if you choose wisely.

--
Regards, 
Martin Brown

On second thought, I don't think this will work at the jitter level required. The original sorta10 signal will be periodic, so the graph of K_phi vs phi will actually be a very fine staircase rather than a ramp, so that there will be thousands of very small deadbands. Unless these are very small (which maybe the delta-sigma thing will accomplish), the lock will wander around inside one of them in an uncontrolled way.

Since you have an FPGA, maybe gate the XOR output to select only the region near the edges. That will give the same relative drift improvement, assuming that all the amps and stuff downstream are more stable than the logic's power rails and duty cycle. (Which they probably are.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

No choice, those machines always come with a high-powered graphics card. The on ly option would be to pull it out but then the mobo probably won't have VGA. At least not for two monitors and I need that.

But then you need to buy another graphics card because now the machine won't have any VGA and DVI connectors.

What's a good fast SSD in your opinion? Something mainly for LTSpice RAW files, does not need to be large. The main concerns would be speed and wear.

Or Dearth Vader looking at you. My road bike saddle is a "Vader" edition. A bit silly but it is very good and cost me less than $15 (shipped).

--
Regards, Joerg 

http://www.analogconsultants.com/

Yes, but is it a Dell? I found over the years that this name brand does stand for quality. Wish I could say that for others. I just spent 2x the money replacing worn fans with Papst. BIG mistake, those things are noisy.

I paid slightly under $700, including sales tax and shipping. Has 8GB of RAM which should be fine but for under $200 I could bring that to the limit which I believe is 32GB.

Could he name such a software? I mean one that is somewhat mainstream and not super expensive.

I could imagine that to make sense if they write their own software or use boutique programs. But I've never had or seen a regular simulator or beamfield calculator making any use of graphics processors.

--
Regards, Joerg 

http://www.analogconsultants.com/

According to Rob, the 125:1944 adder algorithm produces pseudo-random edges in sorta10, with the statistics of a 1st order delta-sigma. If the average pulse width output of the phase detector is, say, 2 ns, and varies from 0.5 to 3.5 as the sorts10 edges jitter around, and you lowpass filter the PD output, you get some noise spectrum in the VCO input; the next issue is to quantify that. He says that the noise spectrum is already shaped towards high frequencies. We can, with a good VCXO, filter pretty hard.

Rob says that he may be able to do 2nd order d-s on the sorta10 edges, pushing the noise spectrum further up. Textbooks analyze that extensively without saying exactly how to generate it. That may not be necessary.

The frequency where the PLL has to discipline a good oscillator is in the range below about 100 Hz. It might work.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

[Snip!]

I think you may want to look up the architecture of sigma/delta *D->A* (not A->D) converters. Those do that sort of thing. It's basically a digital implementation of a sigma/delta modulator.

Jeroen Belleman

How old was your last motherboard? Most i5 & i7 MOBOs have a DVI out. The main CPU has its own very capable 2D GPU built in. 3D isn't so hot, but I was surprised how well it worked considering its limitations.

It depends whether or not your data will be mostly incompressible in which case I favour the Samsung Pro 840 or whatever their new model is now and a size ~256GB which is the sweet spot for populating all channels on the controller. Crucial or Intel would be my second choice(s) though it is a couple of years now since I last looked.

Anandtech has more benchmark details than you will ever want

Basically you need to decide what sort of size read and writes your profile is likely to use and pick wisely. But even the worst SSD on a SATA3 connector will vastly outperform a spinning rusty disk.

So long as you don't regularly torment them with benchmarks SSDs seem to last fine in normal to heavy regular use.

Obviously it is a compromise. I get teased a bit about mine sometimes. (actually both the recent ones are now anonymous black monoliths)

--
Regards, 
Martin Brown

Probably around 5 years, would have to dig it up in the books. Dell has a tendency to economize. For example, they leave off any connectors and other parts that aren't absolutely necessary. So when the PC comes standard with a graphics card I imagine that if the mobo had connectors they aren't stuffed. But can't tell yet because the XPS 8700 won't be here until the end of December.

Thanks. They even got a cheaper 128GB version which would be plenty big for me:

All I'd do is run scores of LTSpice RAW files into it and then pick results from there. Since most of my sims are big which is the reason to buy this new PC there won't be large writes more often than once every

10-20 minutes. I assume those drives have wear-leveling and when data never gets above a few GB it should last.

I might also put some of the more bloated software in there. Like OpenOffice which is the slowest starter on my PC.

Nothing that a good can of spray paint can't fix :-)

--
Regards, Joerg 

http://www.analogconsultants.com/

As I understand it:

Making 10 MHz from 155.52 means a ratio of 15.552. If you make that from logic clocked at 155.52, you get periods that jump between 15 and

16 clocks, such as to average 15.552. The pattern repeats - magically! - at 80 KHz. If you go for 2nd order noise shaping, some of those periods have to be 14s and 17s, too. So the 2nd order noise shaping increases the RMS noise amplitude, but pushes more energy into higher frequencies, which makes the PD output easier to filter.

This is awfully complex. The simple bang-bang detector keeps looking better. Remember, I need a *time* locked loop, not a frequency locked loop. Anything that averages a lot of jitterey edges is going to make time errors.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I dimly remember a paper posted way back in this group--maybe by Rick Karlquist--that had some theory on how to do that.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net