Keyboard Shortcuts
Likes
- BITX20
- Messages
Search
Re: Si5351A: facts and myths
Hello Jerry Your findings re the Si5351A slow register programming are fascinating to me. They kind of imply the Si5351A DOES double-buffer the registers! Amazing. I want to do lots more experiments myself, now. I love the way the understanding of this wonderful little $1 chip keeps being pushed forward - despite chronic lack of hard facts in the documentation.? > Your older comment about floating point taking less flash than? > 64bit integer math was interesting.? I'd guess that if you have a > lot of computations, 64bit integer loses due to all of those 64 bit? > args getting bussed around, where standard floating point is? > just 32 bit per arg.? For just a couple computations the integer? > math might win, as the floating point multiply and divide? > subroutines are relatively large. ?? My experiences with this were based on a single line of code involving one floating point multiplication and one floating point division.?It actually was only a couple of computations, as you said - but the floating point STILL won.?? This was my original Si5351A code in the Ultimate3S QRSS/WSPR/etc transmitter kit an also later used in the other kits.?As the code evolved and code space ran out, I kept having to find places to compact the code and eventually came to the tricky one, getting rid of the floating point arithmetic. The easy thing to do was to replace it with 64-bit integers and that was when to my horror, I saw that the flash code space was even BIGGER! Just to be clear again: the entire rest of my code uses unsigned 8, 16 or 32-bit variables. This single line of code was the only place I used a float (or uint64_t).? In the end I spent many hours headscratching with pen and paper, and Excel simulations, and finally the C code... and eventually managed to figure out some neat tricks to accomplish the arithmetic I needed, using only 32-bit, without losing any precision! When I'd done that and floating point libraries were gone from my code I reclaimed 5 or 6% of the Flash space if I recall (even after adding all my integer tricks).? 73 Hans G0UPL |
||
Re: Bitx TFT display code
Do you have a hardware list for your project.. I did not see a listing on your site I want of make sure I have the correct type of tft or I need to order one.. Joe VE1BWV On Fri, Jun 30, 2017 at 10:44 PM, SP Bhatnagar <vu2spf@...> wrote:
|
||
Re: Bitx TFT display code
Wow!, The features look great.... cant wait to try it out. Your website on the Colour TFT display and si5351 looks very interesting.I like the idea of using the si5351 Thanks for the info Joe VE1BWV On Fri, Jun 30, 2017 at 10:44 PM, SP Bhatnagar <vu2spf@...> wrote:
|
||
Re: Si5351A: facts and myths
Yes it is using the same SDA and SCL lines as the Si5351 just a different address. The display is a 5v one. I didn't know there was a choice. I separated the Nano and display so I could mount the Nano with the USB port at the rear of the case and use a 20x4 display. I wanted to add an S meter and maybe a output/swr meter so I wanted the additional lines. It made more sense to run only 4 wires to the front panel and it was pretty easy to change the sketch to accommodate the new display, only three or four lines to change.
Your other post about birdies I have not done a side by side comparison but I use the Bitx a lot for digital so I have tuned around on USB and it sure seems to have less birdies and what is there seem to be at a much lower level. ? |
||
Fw: [BITX20] Bitx TFT display code
Jack Purdum
I forgot to mention that I bought the wrong Mega from AliExpress. Jack, W8TEE ----- Forwarded Message ----- From: Jack Purdum <econjack@...>
To: "[email protected]" <[email protected]> Sent: Friday, June 30, 2017 10:26 PM Subject: Re: [BITX20] Bitx TFT display code All: I bought a Mega 2560 Pro Mini from RobotDyn and it does NOT fit my PCB. ? This is too bad, as it sell for about half the price of the one that does fit. Jack, W8TEE From: Joe <joeman2116@...> To: [email protected] Sent: Friday, June 30, 2017 5:51 PM Subject: Re: [BITX20] Bitx TFT display code In searching for tft displays, i noticed there were serial and parallel. The ones with connectors on the top and bottom are generally pin compatible with uno and atmega2560.? I have several of these and the 3.5" just plugged and Jacks display worked great.. These were mci friends type marked on the board. My other tft's were 2.8 inch.
One i had to change some text size for the display to look correct. The other 2.8 from elegoo was also mcufriend, but Jacks display code worked with no changes. His code is able to identify several tft drivers and select the correct one.. Very cool. The serial type tft usually have single set of pins - 8-10 pins. The usually say spi or serial in the description.
I see that some Japanese and Russian hams use these type. Looking forward to Jack"s next code run... Joe
VE1BWV On Fri, Jun 30, 2017, 4:57 PM Jack Purdum via Groups.Io <econjack=[email protected]> wrote:
|
||
Re: Bitx TFT display code
Jack Purdum
All: I bought a Mega 2560 Pro Mini from RobotDyn and it does NOT fit my PCB. ? This is too bad, as it sell for about half the price of the one that does fit. Jack, W8TEE From: Joe <joeman2116@...> To: [email protected] Sent: Friday, June 30, 2017 5:51 PM Subject: Re: [BITX20] Bitx TFT display code In searching for tft displays, i noticed there were serial and parallel. The ones with connectors on the top and bottom are generally pin compatible with uno and atmega2560.? I have several of these and the 3.5" just plugged and Jacks display worked great.. These were mci friends type marked on the board. My other tft's were 2.8 inch.
One i had to change some text size for the display to look correct. The other 2.8 from elegoo was also mcufriend, but Jacks display code worked with no changes. His code is able to identify several tft drivers and select the correct one.. Very cool. The serial type tft usually have single set of pins - 8-10 pins. The usually say spi or serial in the description.
I see that some Japanese and Russian hams use these type. Looking forward to Jack"s next code run... Joe
VE1BWV On Fri, Jun 30, 2017, 4:57 PM Jack Purdum via Groups.Io <econjack=[email protected]> wrote:
|
||
Re: Si5351A: facts and myths
Would be interesting to know if any birds are still around when operating the VFO at 19mhz.
toggle quoted message
Show quoted text
This can be done with Allard's sketch by going to USB. Jerry On Fri, Jun 30, 2017 at 05:07 pm, Michael Babineau wrote: All that I have left are a few birdies, which I suspect can only be eliminated by better filtering on the DDS input to the main board. |
||
Re: Si5351A: facts and myths
Hans,
toggle quoted message
Show quoted text
Actually, there is a significant change when "a" increments. Going up from 7.0 mhz, MS0_P2 transitions from 0x00000 to 0xd51f8. If they didn't somehow buffer those register writes, I should have heard it given the 10ms delays.. And from what I see on the scope, they make a very clean transition to the new frequency without so much as a runt cycle. ?Not bad for $1. Jerry On Fri, Jun 30, 2017 at 08:29 am, Jerry Gaffke wrote:
When dumping the registers, noticed that the values don't jump much on integer transitions of "a". |
||
Re: Si5351A: facts and myths
Allard et al :
What I did that really got rid of noise from the Raduino was to use either multi conductor shielded cable or RG-174 for every connection? coming out of the Raduino and making sure to ground the shield at both ends. ?Also I avoided running any wires under the Raduino or the main board. All that I have left are a few birdies, which I suspect can only be eliminated by better filtering on the DDS input to the main board. Reducing the Raduino output to 2mA from the default 4mA, significantly reduced birdies without a big impact to the performance of the rig.? Cheers Michael VE3WMB |
||
Re: Si5351A: facts and myths
¿ªÔÆÌåÓýI have a SI5351 Signal Generator running an I2C 4 line LCD, I just had to jumper a non stock address on the I2C adapter. You can run the I2C address scanner to find out where things are? Mike, WA6ISP On 6/30/2017 3:43 PM, Jerry Gaffke via
Groups.Io wrote:
So it's an i2c display, using the same SDA and SCL lines as the Si5351? -- Mike Hagen, WA6ISP 10917 Bryant Street Yucaipa, Ca. 92399 (909) 918-0058 PayPal ID "MotDog@..." Mike@... |
||
Re: Si5351A: facts and myths
So it's an i2c display, using the same SDA and SCL lines as the Si5351?
toggle quoted message
Show quoted text
That seems ideal if there aren't any issues, as we're about out of pins on the Nano. That's relevant. Displays with the little i2c adapter board added on are quite cheap. If not specified they are likely operating at 5v as that's what most of the Arduino world wants. I'd choose a 3.3v display and not 5v though. ?If it had pullups to 5v on board there would be some current getting dumped into the Si5351 die's SDA and SCL protection diodes. If the Si5351 were on the main board, it would be easy to wire up the an i2c display and tuning pot to the processor board of our choosing. ?No need for a Raduino board. (I know, broken record here.) Jerry, KE7ER On Fri, Jun 30, 2017 at 03:08 pm, Jeff wrote: I am not sure it is relevant but I changed to a serial display and all the tuning clicks disappeared with the new display. Nice discussion by the way I am getting my education! |
||
Re: Bitx TFT display code
In searching for tft displays, i noticed there were serial and parallel. The ones with connectors on the top and bottom are generally pin compatible with uno and atmega2560.? I have several of these and the 3.5" just plugged and Jacks display worked great.. These were mci friends type marked on the board. My other tft's were 2.8 inch. The serial type tft usually have single set of pins - 8-10 pins. The usually say spi or serial in the description. Joe On Fri, Jun 30, 2017, 4:57 PM Jack Purdum via Groups.Io <econjack=[email protected]> wrote:
|
||
Re: Si5351A: facts and myths
On Fri, June 30, 2017 22:16, Hans Summers wrote:
In my opinion (reinforced by this result) the tuning "pops" which peopleYes you're absolutely right! There is a lot of confusion about the origin of the tuning clicks. The SI5351 library v2.02, 2.03 and 2.04 had the PLL reset issue which indeed gave clicks because the PLL was unneccessarily reset each time. Those clicks were extra strong when you tuned near to a strong signal. But that has all been fixed recently in v2.05. Any clicks that still remain are coming from the LCD display which can be fixed by better shielding, extra caps, etc. 73 Allard PE1NWL |
||
Re: Si5351A: facts and myths
We had some pops in there due to extra PLL resets in the Etherkit library, but should not be an issue going forward ?Some have found that just a filter cap across the incoming 12v at the Raduino solved their tuning tick issues. ?So filter the 12v in well, and would be best to have a good 3.3v regulator for the Si5351. ?We may have RF from the i2c lines into the Si5351 also. ?If I were scratch buillding for maximum performance, I'd put the Si5351 on the main board under a 1/2" copper plumbing endcap for shieldng, and make sure those i2c lines were in coax and well filtered for RF. ?And use multiple Si5351s and multiple endcaps if needing multiple clocks.
toggle quoted message
Show quoted text
Haven't tried it yet, but I think you could cut off a 1/4" piece of 1/2" copper pipe and solder it to the board, testpoints coming up through the board inside the pipe to center it and give something to solder to. Then the endcap would be a friction fit over the top. ?And keep some 3/4" and 1" endcaps on hand too for larger bits of circuit that need shielding. ?Signals from one endcap to the next are routed as short traces through the bottomside groundplane, perhaps covered with kapton tape and then copper tape soldered to the groundplane. ?Should take out most of them birdies. Your older comment about floating point taking less flash than 64bit integer math was interesting. ?I'd guess that if you have a lot of computations, 64bit integer loses due to all of those 64 bit args getting bussed around, where standard floating point is just 32 bit per arg. ?For just a couple computations the integer math might win, as the floating point multiply and divide subroutines are relatively large. ?? As I said previously, those computations on a,b,c to generate the register values do muddy things up a bit. I'll someday find those frequencies at which an entire register flips from 0xFF to 0x00 and scan through that looking for ticks. ?But I doubt they are there, and if those ticks are there they are infrequently encountered.? Jerry ? On Fri, Jun 30, 2017 at 01:16 pm, Hans Summers wrote:
|
||
Re: Do not solder ground. Instructions mess
Well, you're about to find out anyway: ?No consequences.
toggle quoted message
Show quoted text
Check the back of your board, the SPK1 connector ties into the same big happy ground plane as TUNING1 Some of the ground symbols got labeled "GNDREF" and some did not, but that's all one big ground net. On Fri, Jun 30, 2017 at 11:29 am, Andrew Krause wrote:
|