|
| From: | David Halls |
| Subject: | Re: [Discuss-gnuradio] Source Block - Flow Control |
| Date: | Thu, 16 Oct 2014 14:23:45 +0000 |
John thanks for all your help, this works, although it is necessary to add ‘tx_sob’ and ‘tx_eob’ tags (or certainly it is for those not using the latest gr-uhd).
If you are using MIMO it is more complicated, as you have to add a ‘tx_time’ tag as well. This is not obvious how to generate in a tx flow graph.
My solution was to use a UHD source (the same address as the tx USRP), which generates rx_time, I then created a block which reads in this rx_time tag and sends it as a message to a block just before my UHD sink which converts this to a tx_time based on the number of bursts sent, and the burst interval that I desire.
As I am using MIMO, i.e. multiple USRPs in one UHD sink, I also had to use multiple USRPs in the UHD source to produce the rx_time (although the multiple rx_times will be equal).
From: John Malsbury [mailto:address@hidden
Sent: 15 October 2014 20:03
To: David Halls
Cc: Matt Ettus; GNURadio
Subject: Re: [Discuss-gnuradio] Source Block - Flow Control
Hmmmm.. I've never done burst transmissions with a 2x MIMO case. That may or may not put a minor wrinkle in things. Let's go for the gusto and try this anyway...
(or were you not looking for a burst transmission?)
This assumes you're using a relatively recent (past couple months?) build of UHD and GR.
In summary:
Message Strobe -> Random PDU -> PDU to Stream -> Modulator -> [see how the constellation works]
Give it a shot. If it doesnt work, give a brief try with the SISO case.
You'll want to add some leading/trailing samples to flush FIR-filters and such for the full frame to get out of the USRP unscathed. We can address this next...
-John
On Wed, Oct 15, 2014 at 11:47 AM, David Halls <address@hidden> wrote:
Hi John,
I have been trying at this all day! Not familiar with the async message stuff, I have tried putting ‘sleep()’ in the source block, I have tried a throttle outside it and even discarding the first X items just before the UHD sink to flush the buffer away.
This all causes weird underflows and things at the USRP and results in a horrible looking constellation at the RX (I am transmitting 2x1 MISO which requires perfect TX sync). This is exactly the behaviour, I think, that Matt warned of!!
Could you give me a bit more of an idea of how to even do the simpler idea, that I had thought of… I need to basically trigger the source at certain intervals. How does the message queue interact with the block, does the program flow jump to ‘parse_header_data_msg()’ automatically when a message arrives?
DH
From: John Malsbury [mailto:address@hidden]
Sent: 14 October 2014 23:42
To: David Halls
Cc: Matt Ettus; GNURadio
Subject: Re: [Discuss-gnuradio] Source Block - Flow Control
In the implementation I have in mind, the upstream logic didn't make decisions on when the data generating block should generate data per se... Instead, the upstream block provided feedback on the number of packets received by the USRP (via the old async message block). With this feedback and knowledge of the interpolation steps between itself and the USRP, the data generating block could throttle its own output to achieve a specified latency [on the order of 10's of ms].
I think using a simpler scheme of triggering with an async message would be a more convenient place to start though... What do you think, David?
-John
On Tue, Oct 14, 2014 at 3:23 PM, David Halls <address@hidden> wrote:
Hi John,
Nice to hear from you.
Perhaps in a similar fashion to Martin's HPD block; I can pass a message back from later in the flow graph to indicate when to release a packet from the source?
David
-------- Original message --------
From: John Malsbury
Date:2014/10/14 23:08 (GMT+00:00)
To: David Halls
Cc: Matt Ettus ,GNURadio
Subject: Re: [Discuss-gnuradio] Source Block - Flow Control
David,
Perhaps you can use an async message to trigger the blocks output?In some applications that require filler between valid data frames, I've seen people throttle based off of the number and size of received messages at the USRP.
-John
On Tue, Oct 14, 2014 at 3:02 PM, David Halls <address@hidden> wrote:
That sounds promising. The only thing is that the data *is* valid from time zero, but I just want to send the values from my source block, say once per second.
What can I use to block in my block, just not produce any items for some period of time or a number of calls? and is there anyway to know when I can stop blocking? What will fill the buffers further down the chain?
thanks,
David
-------- Original message --------
From: Matt Ettus
Date:2014/10/14 22:56 (GMT+00:00)
To: David Halls
Cc: Jeff Long ,GNURadio
Subject: Re: [Discuss-gnuradio] Source Block - Flow Control
No, if you don't have anything useful to output in a source block, you can (and should) block while you wait. If you have something useful, but not as much as requested, you can output only what you have. There is no need to generate filler in GNU Radio.
Matt
On Tue, Oct 14, 2014 at 2:43 PM, David Halls <address@hidden> wrote:
Matt,
In my source block I can limit the calls to the DB ok, but I will still need to output something from the block, won't I? This will then propagate and fill the buffers?
Thanks,
David
-------- Original message --------
From: Matt Ettus
Date:2014/10/14 19:09 (GMT+00:00)
To: Jeff Long
Cc: GNURadio
Subject: Re: [Discuss-gnuradio] Source Block - Flow Control
Jeff,
If there is a hardware device like a USRP in the chain, then you should not use a throttle block. What you are seeing is the initial startup burst. When everything starts up, all the buffers are empty, and GNU Radio will generate data until something backs up. Once they fill up, you are seeing the rate settle down. This is all normal.
If you really need to rate limit what gets requested of the database during the initial transient (which I don't recommend), you should do that within your source block.
Matt
On Tue, Oct 14, 2014 at 10:56 AM, Jeff Long <address@hidden> wrote:
Should have mentioned ... set the throttle rate just slightly higher than what the chain would consume at steady state when all the buffers are filled and the USRP is transmitting. How well this works depends on what the rest of the chain looks like.
- Jeff
On 10/14/2014 05:52 PM, Jeff Long wrote:Use a throttle block immediately after your source block, setting the
vector size to match your source.
- Jeff
On 10/14/2014 04:34 PM, David Halls wrote:Dear All,
I am wondering how to add some flow control to a source block, so that
it doesn’t fire out items so quickly.
Later stages of my flow graph are slowed by various bits of processing
and combining, before transmission via USRP, with bursts being
transmitted every few seconds.
What happens is that the block fires out 1,000s of vectors, and then
seems to slow down and settle into sync with the rest of the flow graph
after a few seconds. As my source block is reading in from a Database, I
want to slow this initial rate significantly.
The block outputs vectors of bytes (v_len=144). Any ideas?
Regards,
David
---------------------------------------------------------------
David Halls Ph.D.
Research Engineer
Toshiba Research Europe Limited
32 Queen Square, Bristol, BS1 4ND, UK
Tel: +44 (0) 117 906 0790
------------------------------------------------------------------------
NOTE: The information in this email and any attachments may be
confidential and/or legally privileged. This message may be read, copied
and used only by the intended recipient. If you are not the intended
recipient, please destroy this message, delete any copies held on your
system and notify the sender immediately.
Toshiba Research Europe Limited, registered in England and Wales
(2519556). Registered Office 208 Cambridge Science Park, Milton Road,
Cambridge CB4 0GZ, England. Web: www.toshiba.eu/research/trl
------------------------------------------------------------------------
This email has been scanned for email related threats and delivered
safely by Mimecast.
For more information please visit http://www.mimecast.com
------------------------------------------------------------------------
_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
NOTE: The information in this email and any attachments may be confidential and/or legally privileged. This message may be read, copied and used only by the intended recipient. If you are not the intended recipient, please destroy this message, delete any copies held on your system and notify the sender immediately.
Toshiba Research Europe Limited, registered in England and Wales (2519556). Registered Office 208 Cambridge Science Park, Milton Road, Cambridge CB4 0GZ, England. Web: www.toshiba.eu/research/trl
This email has been scanned for email related threats and delivered safely by Mimecast.
For more information please visit http://www.mimecast.com
NOTE: The information in this email and any attachments may be confidential and/or legally privileged. This message may be read, copied and used only by the intended recipient. If you are not the intended recipient, please destroy this message, delete any copies held on your system and notify the sender immediately.
Toshiba Research Europe Limited, registered in England and Wales (2519556). Registered Office 208 Cambridge Science Park, Milton Road, Cambridge CB4 0GZ, England. Web: www.toshiba.eu/research/trl
This email has been scanned for email related threats and delivered safely by Mimecast.
For more information please visit http://www.mimecast.com
_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
NOTE: The information in this email and any attachments may be confidential and/or legally privileged. This message may be read, copied and used only by the intended recipient. If you are not the intended recipient, please destroy this message, delete any copies held on your system and notify the sender immediately.
Toshiba Research Europe Limited, registered in England and Wales (2519556). Registered Office 208 Cambridge Science Park, Milton Road, Cambridge CB4 0GZ, England. Web: www.toshiba.eu/research/trl
This email has been scanned for email related threats and delivered safely by Mimecast.
For more information please visit http://www.mimecast.com
NOTE: The information in this email and any attachments may be confidential and/or legally privileged. This message may be read, copied and used only by the intended recipient. If you are not the intended recipient, please destroy this message, delete any copies held on your system and notify the sender immediately.
Toshiba Research Europe Limited, registered in England and Wales (2519556). Registered Office 208 Cambridge Science Park, Milton Road, Cambridge CB4 0GZ, England. Web: www.toshiba.eu/research/trl
This email has been scanned for email related threats and delivered safely by Mimecast.
For more information please visit http://www.mimecast.com
| [Prev in Thread] | Current Thread | [Next in Thread] |