Undefined variables in ldpc_dvb/dec/ldpc_dvb_dec_types.svh

[hamzabinabid]

following are the errors:
** Error: C:/Users/Hamza.Abid/Documents/DVBS-2/ldpc_dvb_dec_types.svh(32): (vlog-2730) Undefined variable: 'cZC_MAX'.
** Error: (vlog-13069) C:/Users/Hamza.Abid/Documents/DVBS-2/ldpc_dvb_dec_types.svh(46): near "col_t": syntax error, unexpected IDENTIFIER.
** Error: C:/Users/Hamza.Abid/Documents/DVBS-2/ldpc_dvb_dec_types.svh(47): (vlog-13205) Syntax error found in the scope following 'cycle_idx_t'. Is there a missing '::'?
if you have these somewhere else kindly let me know but in the current file theses variables are not define

[dshekhalev]

I'm sorry, it was my mistake: not full "manual" merge with my non public repo. It's remnants from full codec source. Nowtime all works fine. I actualize some codecs state, check all tests & scripts. All works fine now.

[hamzabinabid]

I was able to resolve that issue earlier. There is another issue iam facing that is iam getting a constant bit error rate and also the buffer gets saturated if i increase the frame rate also if i add some delay in the code to prevent the saturation it gets too slow.

Following are the details:
For code rate = 2/3, ber = 0, at SNR ~ 3.25
For code rate = 1/2, ber = 0, at SNR ~ 1.25
The falling trend of curves is not smooth.
Which i have observed on Matlab!

Regards!
M.Hamza Bin Abid

[dshekhalev]

Testbench is easy demonstration that codec is work. Moreover it's LDPC only codec, no any BCH inside and no parity bits transponse)

If you set optimal IP LDPC settings: pNODE_W = 6, pLLR_W = 4, pNORM_FACTOR = 7, pUSE_SC_MODE =1, add parity bits transpose, add BCH code you get something like this https://github.com/dshekhalev/FEC/blob/main/netlist/dvb_s2/expected_results.png

[dshekhalev]

Something about buffer overflow: codec IP has ordy ports on interfaces. This signal works like "AXI stream ready" & input interface use AXI - like handshake. If you send your data without any action to ordy signal you just drop you data, because it don't write to input buffer (see source unit).

To make maximum decoder performance you should do the following: set pLLR_NUM = 8 metrics by tick, pDAT_W = 8 bits by tick, make assign ireq = 1 (prevent decoder/encoder output handshake), set iclk ~= 300MHz (code clk), iclk_in/iclk_out ~= 250MHz. Then you can get near 500Mbps at ~25-30 fixed iterations.

[hamzabinabid]

Thank you for your response.I will test this and let you know of the results.

Regards!
M.Hamza Bin Abid

[hamzabinabid]

1.Moreover, are you having the architecture diagram for this? With DVB-S2 standard activated.
2.what was the number of frames for the curves you have provided in your earlier response
3.If i increase the number of frames in the testbench will it be having any impact on the curves?

[dshekhalev]

1. Sorry, but the purpose of this repo creating is to show that there is nothing complicated in FEC, it's simple and sometimes primitive logic and every developer can make own codec with decent quality. Yes, it may have gap to best codes realization of 0.5-0.7db but it will work. This is not a commercial repository, so there is not any documentation, diagrams or algorithms descriptions. Anybody who need them can use reverse engineering, the code is self-documented and fairly easy to read.
2. I use BER metric, whereas standard DVB use FER. These curves measured for RTL bertest AWGN channel at 1e9 data bits. I have 1e10-1e12 measurements but it's not inside this repo)
3. I suppose bertest DVB LDPC simulation in an RTL simulator is practically impossible. It takes a huge amount of time. Therefore, I use RTL simulation on a demo board (Xilinx US) and even in this case it can take a couple week for DVB-S2 codec only.

[hamzabinabid]

USED questa sim for Simulation

1. I have plotted curve for bertest.sv as default settings in the file and BER became =0 at SNR=1 but when i chnaged number of frames from default (3)to 6 BER became constant= 2.5e^-1 for the SNR range from 0.5 to 3
   FOLLOWING is the output

bits 194400 EbNo(SNR) = 0.50(0.50) : ber = 3.17e-01. fer = 1.12e-01

bits 194400 EbNo(SNR) = 1.00(1.00) : ber = 2.50e-01. fer = 1.22e-01

bits 194400 EbNo(SNR) = 1.50(1.50) : ber = 2.49e-01. fer = 1.14e-01

bits 194400 EbNo(SNR) = 2.00(2.00) : ber = 2.48e-01. fer = 1.08e-01

bits 194400 EbNo(SNR) = 2.50(2.50) : ber = 2.49e-01. fer = 1.00e-01

test done 17235040

3. Moreover it would be so kind of you to refer me some theoretical resources related to DVB-S2 LDPC implementation so that i can relate to the approach used in this repo.

4.kindly please suggest some software and hardware based tools necessary for the implementation of this .