Here's some quick comparisons, grabbed with a passive whip GPS antenna into a 20dB preamp:
Noise floor,
datecode 1652 (FLiP pre-production with EPSON),
datecode 1712 (FLiP rev A with SITIME)
Will be interesting how Tracy Allen's sample SiTime works out...
Those spectrum, plots are a little different from the earlier ones.
One easy other change to try, would be to vary the P1 CLK Register OSCMx (Bits 4:3)
That would change the CL, and lower Slew (at the expense of more current peaks).
...
Some adjustment to the osc local caps may be in order.
To reduce the fast edge effects, a simple series R, (smallest package they can assemble), very close to the OSC would help.
The OSC itself should be directly decoupled, and maybe with not too large a C - 3v3 is decoupled elsewhere, so a smaller C that gives lower Z at GPS area may help ?
To reduce the fast edge effects, a simple series R, (smallest package they can assemble), very close to the OSC would help.
But might the offending edges also be internal, somewhere up the divider chain?
Well, yes, but those nodes are rather harder to tack any cap onto....
So, you work with what you can reach and change...
Generally length/area/slew determines what radiates how much at a given frequency - a series R greatly reduces how much total trace + prop lead frame has the ns edge speed.
The SiT8918 chip with slower edges on the output seems to have done the trick. I can now set the FLIP down touching the edge of the GPS without it losing lock. I take it as a preliminary result. I have only one FLIP, now modified, so I have lost my control group. Maybe the GPS signal is just stronger today, or the background lower or it's just a fluke.
@jmg, interesting idea about changing the clock setting so as to use the higher capacitance at xtal1 to xtal3. But no. For one thing, I realized earlier programs tested using the stock SiT8918, I was in fact using the xtal1 setting, not xinput as it should have been. Xtal1 puts 36pF at the X1 Prop input. According to the SiTTIME data sheet, that should have slowed down the rise time from ~0.5ns to ~1.5ns. I did try both xinput+pll16 and xtal1+pll16 on my current modified FLIP. The GPS locked in either case, just a 70µA increase in FLIP power supply current on the xtalN setting.
The rise time applies to the voltage slew rate. Apparently the output from the SiT8918 is current limited. Working backwards, 3nanoseconds rise of 3 volts into 60pF, that comes out to 60mA current on the circuit board trace during the transition, compared with 10mA per transition for the version I have on the FLIP now. From the VonSzarvis' spectrum plots, the peaks are definitely spaced at 5MHz, not some higher multiple.
The FLIP pcb layout is impeccable, I think. In the attachment you can see the pcb with the SiT8918 removed. There are three bypass capacitors in close proximity. As seen in the pic, the output is at upper left next to the Prop X1 input. Clockwise from that are ground, OE, and Vdd.
I take it as a preliminary result. I have only one FLIP, now modified, so I have lost my control group. Maybe the GPS signal is just stronger today, or the background lower or it's just a fluke.
I have two. PM an address and I will Priority Mail one tomorrow. I cannot believe Parallax hasn't offered you a lifetime supply for finding their bug.
Sure seems that checking bulk lead times/MOQ's and qualifying that SiTime variant on a larger sample should be considered, alongside the Epson part, for better GPS compatibility. I'll feedback your results to the team pronto!
Mostly, I'm happy you got your FLiP back!
Whilst writing this, I feel compelled to say thank you for sharing all your educating insights here. I'm sure it's been said many times over the years, but still you continue to inspire me with all that you post. (And help make the task of Moderating enjoyable too!)
Sure seems that checking bulk lead times/MOQ's and qualifying that SiTime variant on a larger sample should be considered, alongside the Epson part, for better GPS compatibility. I'll feedback your results to the team pronto!
If talking to SiTime, there are also SiT1602B (58c/1k) and SiT8008 (59c/1k), which seem cheaper ?
SiT1602B does not formally mention 5MHz, but does spec 3.57 ~ 77.76, including 6MHz, but maybe SiTime would like the "visible use" FLiP has, and consider adding 5MHz.
You can only ask
I also notice the (2.00mm x 1.60mm) package, is cheaper again at 52c/1k, which may result it that becoming the preferred package, so it may fit on the same pads ?
8008 spec says this : "Operating temperature from -40°C to 85°C. For 125°C and/or -55°C options, refer to SiT1618, SiT8918, SiT8920"
Seems for a FLiP like product, -40°C to 85°C is fine ?
cheapest, lowest slew part code I think is SiT1602BCL73-33E-5.000000E ?
I like to help out Parallax because of their people and culture, and also because they have given me support and indeed many years of price breaks on oem Stamps and Propellers. This particular troubleshooting issue came up in a thread that I started because I like the FLIP design; then the troubleshooting itself has become an obsession moved along by contributions from other forumistas. The forum is a learning experience for us all. Thank you for your kind remarks, and Tom, for the offer of the extra FLIP for the control group. VonSzarvas, I have four more samples of the SiT8918 and could get them to the development team if they want them for a further trial alongside the Epson and side by side visualization on the spectrum analyzer. I'm curious what difference it makes quantitatively.
jmg, I wonder about the difference in spectra too, but I'll place a higher wager on the Agilent spectrum analyzer. I won't bet much money, though, unless the two are operating side by side under the same conditions and with the same antenna. That might be interesting. A poor man's spectrum analyzer is very appealing!
The northern CA rep for SiTime is NorComp, and I'm think they could put Parallax in touch with an FAE for suggestions. These parts are programmable, that is, it is relatively easy for them to factory program in bulk with the settings that you want. They also offer a field programmer.
I would definitely trust the Agilent as well - it's propose-built. The Airspy is an excellent receiver as far as SDR goes but it is just a general purpose radio receiver. Most SDR gui software just happens to typically have a spectrum/waterfall plot that is useful for sniffing out things like this, but it's usually not instrumentation grade.
I see that the spectrum analyzer was Siglent, not Aligent. I'm not sure either how the SDR performs as a spectrum analyzer, but seeing all the buttons and knobs and menus on a "real" spectrum analyzer is enough to convince anyone that there is much to it. Or a browse through the extensive HP/Agilent/Keysight app notes.
The Siglent looks like a nice modern instrument. So let me ask @VonSzarvas, how do you like it? I've thought about buying a used HP8594 off of ebay etc., solid, 9kHz--2.9GHz, cheap for raw bandwidth, nervertheless, 1980s technology with a CRT, no USB or modern interface, and a curous mess of options that had to have been factory installed.
So let me ask @VonSzarvas, how do you like it? I've thought about buying a used HP8594 off of ebay etc., solid, 9kHz--2.9GHz, cheap for raw bandwidth, nervertheless, 1980s technology with a CRT, no USB or modern interface, and a curous mess of options that had to have been factory installed.
I like it. It's a tool I found makes certain tasks much more productive, and has provided great learning. Things like searching for faulty components on a PCB (parts which might appear to have power, but may not be active inside), understanding and optimizing passives, or looking at emissions of course
One thing above all is the speed and simplicity of working. Of course it can't replace a scope- but I think a good example of simplicity compared is when you'd want to probe a node with a scope, and have to carefully place the scope probe electrically (takes time/care), or bridge a capacitor with a prong-probe (I mean the scope tip and very short gnd)- then again that can be fiddly. Typically with the spectrum analyzer you work with insulated hoops or probes, so you can run them directly across components and traces- move about quickly and with no fear of damage or influence to what your testing.
As for new vs. old- If you can find a good working model sub $500 then maybe, but otherwise personally I'd be weary of spending the prices I could find on eBy (~$1000+) on such an old device. It's a huge investment for something without warranty. The new mid-range SA's can be had for not much more (logarithmically), and are likely not any less accurate then the vintage (and un-calibrated?) models. Either way, this is a big investment for sure. Not just the SA, but the probes, pre-amps, connector/adapters and antenna's that follow ! You can easily double your costs! (At least you have some nice hoops though, so you probably only need a silly-expensive N to BNC adapter!).
Afterthought- I suppose a used model with all the extra's included would be somewhat more appealing. You'd probably manage without an expensive external pre-amp with a modern SA, but I think that would be essential kit for a 1980's model. That alone could be $500 worth of consideration in your calculations.
@VonSzarvas, Thanks for the detailed response. You mentioned functions there that I wouldn't have thought of for a spectrum analyzer. Certainly great for learning. It looks like a big learning curve to delve into it and to make use of it effectively and to avoid errors of interpretation. I have do have a fancy 'scope with a spectrum function, but it is implemented as math post-processing, not with mixers and analog filters and all of the other goodies that make up a true spectrum analyzer, and not the frequency range of course. I often find it hard to interpret the spectrum as displayed on the scope, but the real deal such as the photos you posted makes it so clear. I do see what you mean about the new generation of SAs compared to the earlier generation. Indeed, comparing specs, so much of the functionality that cost extra $$$ back then is now included or much improved.
I've bought quite a few things from ebay and other surplus dealers. Despite a few lemons, the performance has in general been good. Basic equipment, the 'scope for one thing, and nice power supplies. I'll wait on the SA though. Most of my work is close to DC, but as the GPS issue with FLIP shows, anything with fast edges can cause a EM compatibility concern.
I have a couple extra samples of this chip that I'm sending up to Stephanie at Parallax. Maybe she can send one of those for you to compare with your equipment.
Somewhat late for the start of the party but, I received my Flip today.
Such a cool little board, I wish I had got the breadboard offered for it. I can certainly see the impatience to power it up, my first cable choice was a USB power only cable, gave me a few minutes of wondering what the problem was. But after that it's off and running, and busy in a multi-cog Collatz conjecture.
Just received my "FLIP", and the package says "STOP-INSTALL SOFTWARE FIRST!"...My question is, can I still use the software that loads .spin and also the software that loads .pbas, that I am using now on the Propeller Activity Board? I am not going to be use "Blocky Prop"..thanks..DennO
I use the same software for Activity Board, PPDB, prop plug, and flip, no problem. I use propellor tools, not basic. Windows 7. Just a different USB cable.
If I recall correctly, that sticker is meant more for FTDI drivers. During the initial release of the FLiP, some people had driver issues with older FTDI drivers.
Once again, thanks to all....for almost 2 decades, I have been using the Basic Stamp for all my projects, mostly the BS2sx with the 8 8 program slots and there is/has been alot of them. But it is time to get into the Prop, and learn how to use all 8 processors to get things done at the same time...DennO
At wits end...LOL...using .pbas how do I preform a math operation on the following:
ascii = STR inValue, 4
ascii = 2724 and I want to divide it by 819...the answer should be 3.32, but I keep getting INVALID NUMBER OF PARAMETERS when I try... volt = ascii/819.
And yes, I have declared the various VARIABLES as LONG(s)
You cannot directly do math on a string value (ascii) - it's only ASCII text - it's NOT 2724 rather it's the characters "2","7","2" and "4" NOT values. You have the long integer value right there - invalue. To get the precision you need you'll need to do scaled integer math - for example
tmpval1 = invalue * 100 ' if invalue = 2724 tmpval1 will be 272400.
tmpval2 = tmpval1/819 ' tmpval2 is set to 332
Consult the manual paying extra attention to the string commands (you already know how to use STR) - you'll have to build a new string reflecting the decimal point you want to display by taking a couple of substrings, namely the first "3" and the last two charactors "32" and concatenate those along with adding a decimal point. There are several different ways to do this and ways to optimize it but this will hopefully give you enough to propel yourself forward.
See https://en.wikipedia.org/wiki/Scale_factor_(computer_science) In particular see the Fractional Values Scaled to Integers paragraph.
In reality using powers of 2 help speed things significantly but disregard this for the moment - after you get your head wrapped around what's going on here then you can optimize and expand. See for more. I'm absolutely certain there are forum members who are much, much better than I at explaining this - my forte' is in biology, physiology and the health sciences.
I do understand about NOT being able to do math on a string value...
I have done ALOT of integer math, over the years, with the Basic Stamps, as you can see from the following example with great success, but I am guessing there is no such thing, as integer math, in .pbas...maybe I am wrong. JUst having a little trouble with making SUBSTRINGS and placing the decimal. Used to just divide the whole number by 1000, or 100, or 10 to break out the decimal...etc. See following code snippit...
'SUB ROUTINE
calculate_24voltage:
'remember...12 bit ADC = 4096 resolution...@ 30 volts...means that the resolution is 136.5 to 1 volt.
total_volt1 = (ADC_24volt_average*10)/1365 'integer math for units..first and second digit
total_volt2 = ((ADC_24volt_average*10//1365)*10)/1365 'integer math for third digit
total_volt3 = (((ADC_24volt_average*100//1365)*100)/1365)*100/1365 'integer math for forth digit
SEROUT LCD,baud,[posCMD,73,DEC2 total_volt1,".",DEC1 total_volt2,DEC1 total_volt3]
RETURN
however in the following code snippit is the MCPValue a string or is it a binary number? With the Basic Stamp, (in the old days), I always thought it was a binary number that one could convert to a decimal number...
Comments
Those spectrum, plots are a little different from the earlier ones.
One easy other change to try, would be to vary the P1 CLK Register OSCMx (Bits 4:3)
That would change the CL, and lower Slew (at the expense of more current peaks).
To reduce the fast edge effects, a simple series R, (smallest package they can assemble), very close to the OSC would help.
The OSC itself should be directly decoupled, and maybe with not too large a C - 3v3 is decoupled elsewhere, so a smaller C that gives lower Z at GPS area may help ?
-Phil
So, you work with what you can reach and change...
Generally length/area/slew determines what radiates how much at a given frequency - a series R greatly reduces how much total trace + prop lead frame has the ns edge speed.
@jmg, interesting idea about changing the clock setting so as to use the higher capacitance at xtal1 to xtal3. But no. For one thing, I realized earlier programs tested using the stock SiT8918, I was in fact using the xtal1 setting, not xinput as it should have been. Xtal1 puts 36pF at the X1 Prop input. According to the SiTTIME data sheet, that should have slowed down the rise time from ~0.5ns to ~1.5ns. I did try both xinput+pll16 and xtal1+pll16 on my current modified FLIP. The GPS locked in either case, just a 70µA increase in FLIP power supply current on the xtalN setting.
The rise time applies to the voltage slew rate. Apparently the output from the SiT8918 is current limited. Working backwards, 3nanoseconds rise of 3 volts into 60pF, that comes out to 60mA current on the circuit board trace during the transition, compared with 10mA per transition for the version I have on the FLIP now. From the VonSzarvis' spectrum plots, the peaks are definitely spaced at 5MHz, not some higher multiple.
The FLIP pcb layout is impeccable, I think. In the attachment you can see the pcb with the SiT8918 removed. There are three bypass capacitors in close proximity. As seen in the pic, the output is at upper left next to the Prop X1 input. Clockwise from that are ground, OE, and Vdd.
I have two. PM an address and I will Priority Mail one tomorrow. I cannot believe Parallax hasn't offered you a lifetime supply for finding their bug.
tc
Sure seems that checking bulk lead times/MOQ's and qualifying that SiTime variant on a larger sample should be considered, alongside the Epson part, for better GPS compatibility. I'll feedback your results to the team pronto!
Mostly, I'm happy you got your FLiP back!
Whilst writing this, I feel compelled to say thank you for sharing all your educating insights here. I'm sure it's been said many times over the years, but still you continue to inspire me with all that you post. (And help make the task of Moderating enjoyable too!)
Michael.
Yes, but the other spectrum images differed, so I wonder if they test different things ?
If talking to SiTime, there are also SiT1602B (58c/1k) and SiT8008 (59c/1k), which seem cheaper ?
SiT1602B does not formally mention 5MHz, but does spec 3.57 ~ 77.76, including 6MHz, but maybe SiTime would like the "visible use" FLiP has, and consider adding 5MHz.
You can only ask
I also notice the (2.00mm x 1.60mm) package, is cheaper again at 52c/1k, which may result it that becoming the preferred package, so it may fit on the same pads ?
8008 spec says this : "Operating temperature from -40°C to 85°C. For 125°C and/or -55°C options, refer to SiT1618, SiT8918, SiT8920"
Seems for a FLiP like product, -40°C to 85°C is fine ?
cheapest, lowest slew part code I think is SiT1602BCL73-33E-5.000000E ?
The northern CA rep for SiTime is NorComp, and I'm think they could put Parallax in touch with an FAE for suggestions. These parts are programmable, that is, it is relatively easy for them to factory program in bulk with the settings that you want. They also offer a field programmer.
Cheers,
Jesse
The Siglent looks like a nice modern instrument. So let me ask @VonSzarvas, how do you like it? I've thought about buying a used HP8594 off of ebay etc., solid, 9kHz--2.9GHz, cheap for raw bandwidth, nervertheless, 1980s technology with a CRT, no USB or modern interface, and a curous mess of options that had to have been factory installed.
I like it. It's a tool I found makes certain tasks much more productive, and has provided great learning. Things like searching for faulty components on a PCB (parts which might appear to have power, but may not be active inside), understanding and optimizing passives, or looking at emissions of course
One thing above all is the speed and simplicity of working. Of course it can't replace a scope- but I think a good example of simplicity compared is when you'd want to probe a node with a scope, and have to carefully place the scope probe electrically (takes time/care), or bridge a capacitor with a prong-probe (I mean the scope tip and very short gnd)- then again that can be fiddly. Typically with the spectrum analyzer you work with insulated hoops or probes, so you can run them directly across components and traces- move about quickly and with no fear of damage or influence to what your testing.
As for new vs. old- If you can find a good working model sub $500 then maybe, but otherwise personally I'd be weary of spending the prices I could find on eBy (~$1000+) on such an old device. It's a huge investment for something without warranty. The new mid-range SA's can be had for not much more (logarithmically), and are likely not any less accurate then the vintage (and un-calibrated?) models. Either way, this is a big investment for sure. Not just the SA, but the probes, pre-amps, connector/adapters and antenna's that follow ! You can easily double your costs! (At least you have some nice hoops though, so you probably only need a silly-expensive N to BNC adapter!).
Afterthought- I suppose a used model with all the extra's included would be somewhat more appealing. You'd probably manage without an expensive external pre-amp with a modern SA, but I think that would be essential kit for a 1980's model. That alone could be $500 worth of consideration in your calculations.
I've bought quite a few things from ebay and other surplus dealers. Despite a few lemons, the performance has in general been good. Basic equipment, the 'scope for one thing, and nice power supplies. I'll wait on the SA though. Most of my work is close to DC, but as the GPS issue with FLIP shows, anything with fast edges can cause a EM compatibility concern.
I have a couple extra samples of this chip that I'm sending up to Stephanie at Parallax. Maybe she can send one of those for you to compare with your equipment.
Such a cool little board, I wish I had got the breadboard offered for it. I can certainly see the impatience to power it up, my first cable choice was a USB power only cable, gave me a few minutes of wondering what the problem was. But after that it's off and running, and busy in a multi-cog Collatz conjecture.
Great little addition to the family!
-Phil
"You will flip over the FLip"
ascii = STR inValue, 4
ascii = 2724 and I want to divide it by 819...the answer should be 3.32, but I keep getting INVALID NUMBER OF PARAMETERS when I try... volt = ascii/819.
And yes, I have declared the various VARIABLES as LONG(s)
DennO
See https://en.wikipedia.org/wiki/Scale_factor_(computer_science) In particular see the Fractional Values Scaled to Integers paragraph.
In reality using powers of 2 help speed things significantly but disregard this for the moment - after you get your head wrapped around what's going on here then you can optimize and expand. See for more. I'm absolutely certain there are forum members who are much, much better than I at explaining this - my forte' is in biology, physiology and the health sciences.
Mike R...
I have done ALOT of integer math, over the years, with the Basic Stamps, as you can see from the following example with great success, but I am guessing there is no such thing, as integer math, in .pbas...maybe I am wrong. JUst having a little trouble with making SUBSTRINGS and placing the decimal. Used to just divide the whole number by 1000, or 100, or 10 to break out the decimal...etc. See following code snippit... however in the following code snippit is the MCPValue a string or is it a binary number? With the Basic Stamp, (in the old days), I always thought it was a binary number that one could convert to a decimal number...