dkozel
commented
7 years ago Hi @JohnDSutter. Thanks for the bug report. I should be able to check this this weekend.
Open JohnDSutter opened 7 years ago
dkozel
commented
7 years ago Hi @JohnDSutter. Thanks for the bug report. I should be able to check this this weekend.
etchorner
commented
7 years ago This may be IDE version dependent. On 1.8.1, this bug does not exist. A user on the HR360 forum points out that this bug is likely the result of variable overflow, in the const arithmetic of SetDDSFreq, where the 4294967295 const is already at the limit of a 32 bit variable.
An alternative to that of @JohnDSutter would be to replace the const division term (4294967295/125000000) with the const 34.35973836.
JohnDSutter
commented
7 years ago There is a difference between
(Freq_Hz * 4294967295)/125000000
and
Freq_Hz (4294967295/125000000) => Freq_Hz 34
when using integers
Band
Center Frequency
(Freq_Hz * 4294967295)/125000000
Freq_Hz * (4294967295/125000000)
Actual Frequency
Error (kHz)
% from Center
160
1900000
65283503
64600000
1880107.448
19.89
20%
80
3750000
128849019
127500000
3710738.384
39.26
16%
40
7150000
245672129
243100000
7075141.186
74.86
50%
30
10125000
347892351
344250000
10018993.64
106.01
141%
20
14175000
487049291
481950000
14026591.09
148.41
85%
10
28850000
991278452
980900000
28547947.3
302.05
86%
6
52000000
1786706395
1768000000
51455572.26
544.43
27%
From: Christopher Horner [mailto:notifications@github.com] Sent: Friday, February 03, 2017 1:05 PM To: HamRadio360/Antenna-Analyzer Cc: John Sutter; Mention Subject: Re: [HamRadio360/Antenna-Analyzer] SetDDSFreq calculation ambiguous (#7)
This may be IDE version dependent. On 1.8.1, this bug does not exist. A user on the HR360 forum points out that this bug is likely the result of variable overflow, in the const arithmetic of SetDDSFreq, where the 4294967295 const is already at the limit of a 32 bit variable.
An alternative to that of @JohnDSutter https://github.com/JohnDSutter would be to replace the const division term (4294967295/125000000) with the simple integer 34. Not all version of the IDE may support uint64_t or long long casts for const arithmetic.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/HamRadio360/Antenna-Analyzer/issues/7#issuecomment-277361981 , or mute the thread https://github.com/notifications/unsubscribe-auth/ABDiiAwVc6nVOGBar_2BPveamVH2Fh81ks5rY5ZpgaJpZM4LpB0X . https://github.com/notifications/beacon/ABDiiJAorFL2uKqjgo9uUDaa5JgzUQ-Gks5rY5ZpgaJpZM4LpB0X.gif
etchorner
commented
7 years ago @JohnDSutter
Agreed. I think your comment and my edit to my earlier comment crossed in ether. I was having a related discussion on the HR360 forum with a fellow who pointed out the error in my integer multiplier suggestion. My (amended) suggestion is to use 34.35973836 as the tuning word multiplier. Conceptually, the 64 bit cast method you propose may likely work as well, but I'm running 1.8.1 not 1.6.5 so I cannot test it here.
r/ Chris
JohnDSutter
commented
7 years ago If there's room for a floating point library that's just as good.
The difference is negligible.
Thanks,
From: Christopher Horner [mailto:notifications@github.com] Sent: Friday, February 03, 2017 2:08 PM To: HamRadio360/Antenna-Analyzer Cc: John Sutter; Mention Subject: Re: [HamRadio360/Antenna-Analyzer] SetDDSFreq calculation ambiguous (#7)
@JohnDSutter https://github.com/JohnDSutter
Agreed. I think your comment and my edit to my earlier comment crossed in ether. I was having a related discussion on the HR360 forum with a fellow who pointed out the error in my integer multiplier suggestion. My (amended) suggestion is to use 34.35973836 as the tuning word multiplier. Conceptually, the 64 bit cast method you propose may likely work as well, but I'm running 1.8.1 not 1.6.5 so I cannot test it here.
r/ Chris
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/HamRadio360/Antenna-Analyzer/issues/7#issuecomment-277375663 , or mute the thread https://github.com/notifications/unsubscribe-auth/ABDiiNpMC-PcytECVkgZ697R82QF3Gweks5rY6VJgaJpZM4LpB0X . https://github.com/notifications/beacon/ABDiiMpPZdR8TGG6wwbaovUsBzGvudclks5rY6VJgaJpZM4LpB0X.gif
On Arduino 1.6.5 it consistently gets the wrong answer giving me a nice ~1Hz signal. The following is a bit more explicit:
void SetDDSFreq(long Freq_Hz) { // Calculate the DDS word - from AD9850 Datasheet uint32_t f = (uint32_t) (((uint64_t)Freq_Hz * 4294967295ULL)/125000000ULL);