Open Research Institute Weekly Reports, 22 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: ORI FPGA Meetup 22 November 2022

Description

COBS Decoder, downlink on Pluto, and more.

A

Thank you so much for being here, it's a November, 22nd and today is our stand-up meeting for fpga and remote Labs work at open, Research Institute, and what we do here is we we care about what we've done over the past week and what we have planned for the next week.

A

If we need any resources to accomplish this work and if we have any roadblocks and there's a plenty going on so I will, what I'll do is I'll turn it over to James first, to give him an opportunity to talk about remote, lab South um and before he before he takes it over um we're in the I've talked with with Keith uh I'm, not sure I've talked with you yet, but we're we're gonna initiate an attempt to move equipment uh pretty soon so I'm I don't know exactly when, but uh since things are good at remote, lab South we'll be able to ship some equipment to you all right.

A

So anyway, the floor is yours. James.

B

Oh, thank you Michelle. Actually, that's pretty much. The primary thing to report this week is that we're getting ready for the new shipment to come over and.

A

B

A

Should have let you set it.

B

That's that's fantastic because it just means that everybody's, on the same page, that we're getting ready for the big shipment of new equipment, we're getting ready to get that all set up and uh remote lab South will be expanding its capabilities. Very soon is the Hope.

A

Okay, yeah I'll do all that I can to do it so yeah, sorry for sorry, for that you should have. You should have been the one to to present.

B

It no, that is entirely fun because it better shows that we're both on the same.

A

Page yeah I guess you're right uh yeah, so so those that uh that are listening, um so vermontlab South is a a wider footprint in terms of Science and Tech than uh than Royal Labs West, which is uh focused on digital Communications. uh So there'll be lots more posted and printed and and photographed and videoed about it over the next couple of months. If all goes well, so thank you so much James. We really appreciate you and everybody. There.

A

Because all right I know Everest has some some reports and progress, so I will I think the floor should be his.

C

Okay, hello, everyone um yeah uh some progress about uh BB frame receiving uh uh so with the mini tuner in order to receive after that dvdc.

C

So right now the the fragmentation of the BB frame is okay and so I can send some UDP packet with DB frame inside and then we can analyze it on Wireshark. So it's a good start. Then um then I need to uh implement the dvbjc, and there is several solution to do that.

C

uh The first is to uh adapt the Run code, the gear dvb dvb GSE.

C

The second option is to use lib GSE from uh from GitHub, and the order is used rust version of Daniel E4 gbz, but it's only include the decoder.

C

uh So here is where I am uh on the fpga side. uh So I get back from uh well. uh uh Prior I changed the biswits from 60 bits, 64 bits to a 32., but um the the the mod code is Well written by The Encore. But the data are not okay, I, don't know exactly why so I get back to uh 64 bits and then of the uh the semia issue, with some mod callers before so I went for you uh cops details if it could uh help on that back to you.

A

Thank you so much. It's super interesting all right, I'll turn it over to Paul. He might be able to answer your question and has a report on the on the Cobbs decoder work.

D

Hello um yeah, it's all sorts of assessment going on in the Cobbs decoder. um Let me show you a screen if I may um see, if that's gonna work.

D

I did set this up in advance.

D

I'm not seeing the window in my choices. Well, I wonder why not.

D

I can use this I think there. It is.

D

That failed, didn't it. What are you seeing now.

A

Random random uh windows, not good, it's not quite there. Yet, okay.

D

A

To you're able to share so it's uh it's all! You.

D

Yeah, that's great.

D

I think this program doesn't really know how to be a Macintosh program is the problem. um Let me worry about that later.

D

What we've been doing is trying to create the the decoder that receives Cobb's encoded data in eight bits, wide axi interface, actually stream interface, and we have made quite a bit of progress thanks to help from Michelle and I've been generating timing diagrams that show that exact logic needed to decode Cobbs and that turns out to be more complicated than I thought. It was, but not too terribly complicated and I think I'm pretty close to having a set that'll work.

D

I definitely have a set that satisfies the first six timing, diagrams worth of where the test cases, which is what I wanted to show you, and maybe a little bit later in the meeting I'll, be able to do that.

D

um I'm, not sure what the answer is to your question ever is because uh I didn't really understand the question, but the idea is that Cobbs will help with dividing up the incoming byte stream into uh BB frames of whatever size they need to be. So if you're going to mix different BB's frame sizes, you're definitely going to want something like that, or else you'll have to rely on the individual dma transfers. Being the boundaries and I think that's nearly too fragile to use. This is why I propose introducing Cobbs last week.

D

um I don't want to go over all that ground again. So, if there's something uh something more specific I can answer, please ask: that's. uh That's currently the status.

C

Yeah, okay, so it's it should solve the problem of alignment mainly for the dma, uh which means that uh right now uh the the unignments and the teles is used, um and if the transfer transfer size of the BB frame is not aligned to that, then, uh where we we do some uh bytes and with your uh cops solution um there should not well, uh we we don't have any issue with that, because we we can uh well. You know exactly when the BB frame uh starts right.

D

That's correct, yes,.

C

uh And the on the uh I wonder well on the encoding side, uh I think the process is not very complicated.

C

uh Well on the on the to run on the on the PS, uh uh which means that uh there is not a lot of computation, I think uh uh where it's more computation than uh just memory copy, but I hope that it doesn't uh well for a little uh arm CPU. It doesn't uh be too much of uh of user CPU.

D

Usage, I, don't I, don't think it should be. It's a very lightweight protocol and it's designed to be easy to implement for software that has access to the whole buffer all at once. The only reason it's complicated for me is that we're trying to pipeline stream it- and it's not designed for that. It's designed to be buffer at a time, and that only adds you know a couple cycles of delay and some State machines to mess with. So it's not um it's not much of a load. I, don't think uh you.

D

My intention was that it would be used for loading mainly from files where you have all the time in the world to build them, but if you have to build them on the Fly, that should be very feasible too um I. Don't intend that this will be necessarily used for on-chip.

D

You know, stuff that comes from the receiver is probably already going to be nicely partitioned into into blocks and won't need cobs in order to stay partitioned correctly, but anything. That's a byte stream, independent of uh axi style processing could benefit from cops for the framing.

C

Okay, so this block would be in front of the encoder of the room of the current implementation encoder, or do you have to uh modify the in color, particularly the metadata input.

D

Yep, it would go in front sorry about the phone ringing in the background. It would go in front of the inline configuration adapter and there will probably need to be another. Some some width adaptation, because this block we're designing is single byte and single byte out, so uh an additional width adapter might be necessary or a modification of the inline configured after it. I don't know exactly how it's going to fit in, but for now we're working in by one bite in one bite out.

A

Yes, I think that would mean it was. It would be after yeah. It would be. It would require something like that, um but yeah it'll be separate. We're not going to put it as part of the encoder we'd like the encoder to stand alone and be able to handle pretty much anything and be reusable.

A

So there may be a little bit of of overhead in terms of programmable logic or fabric usage, but hopefully not too much. We've been trying very hard to make it efficient.

A

So Paul, do you want to try to work on getting some some of your um audio visual examples and then, if, if ever East has anything else that he'd like to talk about, then we can let him have the floor and then uh then Sasha can talk about some of the RF activities.

D

Okay, actually I'm gonna get a try right now. Let me let me go ahead and see if that's going to work.

A

Oh okay, yeah go ahead.

D

um Are you seeing timing diagrams now yeah.

A

Yeah, we sure are thank you. Okay,.

D

This is I, don't want to go through this. This is way too complicated uh to go through in this meeting, but there is a document that walks through it in excruciating. Detail and I would encourage anybody who's. Any interest in this to go, read it and find my mistakes, because I'm sure there are still mistakes in this document, and then these timing, diagrams and very possibly in the logic equations that we're going to use to implement them um I'm using this new tool new to me anyway, a new tool called wave drum.

D

You can see some of the language up at the top, there's a whole whole mess of this little cryptic code. That's how you create the timing diagram and then the diagram draws itself in this little online. It can be an online tool or it can be a standalone tool like I'm running here or their various plugins. You can use that integrate it with other tools, but you can see.

D

We've got quite a few um signals here involved in getting the logic right and two delays, although it's more complicated than that with all the handshaking going on between the input data and the output data. Unfortunately, there's no way to implement this as a standard axis stream that has no delay, which you sometimes see in symbol, axi stream blocks.

D

But this has the minimal amount of delay that will allow a nice synchronous, implementation I think we could do it in one cycle less, but there's no, no reason to push it that hard and that would lead to Messier low level timing. So this is what I've been beating my head against for the last week and I'm now approaching some sort of conclusion on. uh Please read the the document. If you can and- um and let me know what I've done wrong.

A

Yeah, thank you yeah. It's a real big step forward to be able to have timing diagrams described with a bit of code because all of us that have used um Vizio and neck.

D

Paint back pain, I'm in diagrams and Mac paint back on the day, so yeah.

A

It's just so nice to have a description, it's very similar to me uh to open S CAD for 3D modeling, where you describe the shape and it renders it and you can change it and then share it.

A

It's really nice and it's the lovely, colors and and good fonts, and all that, so we're really super happy to find it and to incorporate this into our documentation. It's called waved, ROM and I. Think there's a well I'll make sure that there's a link to it in this uh in this recording sorry.

A

Yeah we're looking forward to it I've been mainly in charge of the test bench and I'm learning how to make not stupid test benches. um So a lot big learning curve for me to get back into the swing of this and I've found several really neat open source methods. So so far, what I've done is I've used the standard, IEEE library to open files in a test bench and then read in the data for the Cobbs decoder.

A

uh You know so this is essentially the input data and some of the control signals and then produce a simulation in vavado and and sawato has a hole. He uses a a much more sophisticated test.

A

Bench Suite with his fpga cores and I, didn't think that we needed all of that for this particular project, so so we're doing sort of the minimal just open a file and then visually verify, but the next step up will be attempting to use either swato's methods or use coco TV, which is a neat framework and python that does helps you do test benches, so we're on the test side of it, because we we developed the test test benches along with the the code that we do.

A

You know very important in order to make it reusable. You know so we're we're prioritizing that, and so this will. This is moving forward and we'll be published at the same time so that we don't have a situation where we're going to dump a bhdl code block on on someone without a way to to independently verify it. At least you know at a at a basic level, you get all of the stuff that you need in order to verify it. So that's that's what we're out to do all right! All right, Sasha, Sasha, hey.

E

Does my microphone work.

A

Yeah you sound good cool.

E

So I've been looking at the up converter and the whole everything like from IQ samples to an x-band and I, had I I decided to give like direct conversion so like modulating directly at x-band a look, but there aren't that many parts and also there's the question of how good we can make the how clean we can meet that amazing mask and all of that and I also talked to some friends and colleagues who have worked on similar projects and they all went with Superhead and I.

E

Think I'm going to go with super hat as well just to minimize technical risk and the additional cost is actually not that significant and another advantage of super hat is that it lets us be very illnesses. Have lots of choice in how we generate the initial analog signal because we can use the a the the Ada.

E

Let me look at the 9371 style Parts if we can find them in stock, but The Analog Devices had that has the adrf 9000 Series, which is a bit more recent and those have the disadvantage of slightly less bandwidth and also not supporting je the jesd they're. All like I think CMOS for lvds, but they have the advantage of being an act. They have the advantage of accurately being in stock, which is a nice thing, with the chip shortage being what it is, but also super hot thoughts just use.

E

If, if that won't end up so we can use, we can have, we can be pretty free and how we generate the signal we can use and we can use like discrete a pair of discrete tags and Analog. Devices has a bunch of these and an analog IQ modulator to generate like I had, which is probably I. I looked at the numbers of it and I. Think around like one to two gigahertz is a good place to put the if to avoid having lots of Spurs and yeah I I.

E

Also I I also started looking at parts for the actual, like x-band up, converter and also parts for the. What's it called the X, the x-band, the the the uh the Lo for the up converter, which is not necessarily x-band, but it is pretty high, so it will need a it will need a pretty capable synthesizer and that synthesizer itself will need a pretty good, very low phase noise oscillator, because Faith noise there is going to because of how the because of the the PLL is going to amplify the phase noise.

E

We want the phase noise to be as low as possible on the oscillator, so we've been looking at parts for that and also I think I converged on on an initial set of parts and I'm going to try and get my company to expand step four through these and now what I'm looking at is figuring out, how the the easiest way to to just test it and the things that I would I'm I'm thinking of for testing is it's the the both I want the to both look at the Spectrum quality.

E

So that's going to need to be a spectrum analyzer of some of some flavor and also I, want to I want to get a look to see how how the modulation itself gets distorted by the up conversion and I'm I I had a question. I know that there's the DAC tag there is a ptie dvds2 receiver cards and those are kind of expensive I'm wondering do you all know of of some like a bit less expensive, dvbs to receiver? It's fine.

E

If they're, like software defined or pcie cards, the form factor doesn't matter much but I'm wondering are there ones that are like slightly less expensive? That will give that will give like evm output and, like signal condolence ratio.

A

A

A

E

Have and we have cards in remote Labs, so if it's going to be an open source project.

A

And you can use them. Oh.

E

What what cards are those.

A

E

B

Model number is on.

A

E

A

E

A

Me go get that.

E

Okay, yeah I'll finish giving the end of my status update while you search for that and yeah I I, don't yeah like I'm I'm, totally fine, making my work open source.

E

It's just that like getting getting like, I'm, probably gonna, be playing around with that boards and stuff, and then it's just the physical Logistics might be a bit complicated of getting like the signal to the card from dashboards that I have physically here in Seattle, so that that would that would be my only life difficulties not concerned.

E

So yeah I've been looking at yeah. Just also the other thing that um that I had a question for y'all was. Do you all know if, like good sound converters from like X-Men, whether it's like 10, gigahertz or egg, it hurts to like a lower thing that uh that a like SDR receiver card can ingest yeah.

A

Yeah everybody uses a standard lnb like from DirecTV DirecTV, okay and then that hooks up directly to an RTL SDR, it's incredibly cheap. So you can go from 10 gigahertz all the way down to whatever, whatever.

E

A

E

A

The performances.

A

E

Can you also like me, like the information about the direct, the the down converter itself, the lnb? Yes, thank you so yeah. If you can, if you can get me, the information on the receiver card and the lnb I'd be super grateful, because that gets me closer to like having something I can test those about converter Dev boards with so yeah.

A

Yeah Paul double check me I, put in chat the things that we.

E

Have we have a.

A

Tbs cards a set of TBS cards.

D

That looks right, I'm looking at the same page, probably the remote.

E

Labs test equipment directory has.

D

Documentation for all the items.

E

Oh interesting and do those do those cooler cars like give? You have like signal open noise ratio and evm and all of that I.

A

Think they're doing software, they say again: I think that the tuner cards, depending on the software that.

E

You use to drive them. Okay, I'll, take a look at that and I think I'm just going to order myself uh one or two of those thanks for giving me the part numbers and all of that just so I have a known quantity to work with yeah.

A

Those yeah, those those are older cards and they're. Maybe.

E

A updated version of it.

A

But that's the sort of sort of like a ballpark and.

D

The the manufacturer and the.

A

Particular model line yeah.

D

I'm not sure how cheap.

E

They are, they are those are kind of expensive cards, I believe I, I Googled it and I see a shop selling them for like 160 Euros, which is expensive, but not as expensive, as one of my friends pointed me to a DAC Tech card which was like 2K dollars, and that, like that is offensive. Yeah.

A

Cards were much more expensive.

E

But those are the ones.

C

We were only able to afford, with the.

E

A

Grant and the cards.

E

The tuner cards, those.

A

We were able to just buy on our own, so I think there's an order of magnitude between the two sides: yeah oh yeah, the mini tuner, which may not be available.

D

E

Think you might have trouble finding one.

D

E

A nice little box and they're much.

A

B

Expensive and they do have yeah.

E

Very nicely I'm gonna take a look at that as well thanks for pointing me forward, but.

A

E

Yeah anything that we have in remote Labs is available for.

C

D

To use- and it will help you get information on.

A

Whatever you need, whatever you need, yeah.

E

That, like even if I, don't end up using the physical ones over there because of just I'm going to be playing with Hardware myself, where I physically am knowing knowing the part numbers knowing the the the software, knowing that, like people have actually used this and that works. That in itself is extremely useful.

E

A

Yeah very happy to help all right, any other comments or questions or anybody need anything um have any particular roadblocks.

A

All right we're looking um and we're working very hard on getting a a working revision. A working version of the the little Cobbs decoder into the reference, design and and I should also um note that the pro processor side code from from Everest actually does work with the encoder. And um when last we left that I think this was last week.

A

I was totally thrilled, so every processor side code that that minimal dvb test works and the stuff that we tried to to use uh the the code that we built up uh from the from some of the example code from from Analog Devices. We're not really sure why it times out times out every time. So this has been a remarkably interesting thing. That has happened to us uh and we haven't gone back to that.

A

Yet because we've been working on cobs, but we will very quickly in the near future, come back to processor, side, stuff um and I'm sure that the the secret is some minor difference between the way that every east walks through setting up and configuring all of the iio devices and channels, and things like that and the way that we do it.

A

So there's uh there's very little differences, but it's got to be in there somewhere and it's very exciting, to see a real honest to goodness transmission over the air for the encoder and as soon as we get that, then we can start working hard on the um the the sort of the multiplexing and and bringing in the the receiver side. So it comes in on the receiver multiplexed in and figured out and then transmitted down correctly on the dvbs2 and S2X frames. So all this will come together.

A

uh We we're looking forward to having some some folks come back uh that are that are on leave um so and also the holidays are coming up. So if you are traveling or away uh we we look forward to to you coming back and we'll all be working hard here to make things move forward. So thank you so much everyone for your time and effort and energy, it's totally worth it.

A

I, do have a brief update. I'll tell you about contacted by a school, so this is from Mastodon. So the there's a big free and open source software community on Mastodon and an amateur radio and radio Digital radio, a bunch of different communities on Mastodon and we got contacted by a school- sounds like a University, Research satellite. It's going to be operated on Leo and Mio, so medium earth orbit, and they are doing a dbb, S2, adaptive coding and modulation downlink system.

A

So this is not going to be available for amateurs to use as a communication resource. However, it's going to be used for University Research and if we can help them out in any way with the dvbs2 IP cores, then we will so I've. Given them all the information to get in touch with us. I look forward to hearing from from them and it's very exciting to to see that happen.

A

So if there's anything that they have that we can use that's open source, then we will incorporate it and to make our our projects better and- and vice versa. So it's uh it's nice to nice to see that so, if I over the next week, if you see anybody new in fpga, I'll make sure to to introduce them and uh and identify what they're, what they're about all right, thanks, everybody I think we'll close out. Oh sorry, go ahead! Everest.

D

Left left an interesting link to another, uh the ATV receiver. In the chat there looks like maybe a clone of the mini tuner. uh Do you have any comments on that for us Everest.

C

uh Just just because uh there is some stock on me, tuna- and this is uh an adaptation from elad I- think it's an Italian uh manufacturer, but it is a miniature inside.

C

um It is just because it is available and it's in it's in a box and uh that's it, but there is other option to uh to order a mini tuner. The the main issue right now is that the neem, so the the the the.

A

Doorbell rang and somebody knocked on it and I could not get away I'm running a meeting just like a minute or so ago. I couldn't get away, I'm running a meeting Michelle.

D

You're not muted.

A

D

Dear sorry, the.

A

D

A

Thank you. I tried to mute, I, really apologize, I've erased.

C

No problem, apparently.

A

There's there's a delivery that came so I'll fix it and post all right. I'll be quiet. Now.

C

um Yeah uh the the main issue with the mini the with the mini tuna, what it's the neem- and this is um what the the receiver and the the modulator the St the modulator, which is uh not available. Whether I think that there is around 200 pieces at the batc.

C

uh So the British TV um and the interesting thing is that uh well we can. We can still find some meat tuna pescibi, but uh we can maybe drive uh uh uh the the this name uh using uh raspberry Pico and it could be very easy to uh to implement and have a direct internet.

C

Output, well, that's another story, but it's in the project. I just speak with Brian uh an English guide which is working on the on this project. That's it on the mini tuna. Sorry, uh just just another question about the because it's very good! That's uh you! uh You have work uh well that my God is working on the zc 706.

C

um Have you uh tried to uh reduce the the the the simple well, the sample rate uh of uh the ad board? That's the first thing and the other is. If you really want to um receive it on a dvbs2 receiver, you need to add the um error, C filter, because else it's uh it doesn't work at all. So you.

A

C

uh To add the filter uh I just use a fear instead of the body phase filter, which is very consuming which consume a lot of cells. uh So it's a fair filter, but you have the example already I. Think I, don't know if you have implemented on the on the design.

A

Yes, sort of so we we've been able to um experiment, I guess uh with the with modifying what they call the profile, and this has all of the clocks, and it has a lot of information in it. So the profile for the adrv 9371 is a um it's human readable, but it's a file that has all these configurations, and it has, very importantly, it has the symbol right in it, and we've been so far operating with just the default.

A

That just comes with it out of pedal Linux and everything is wide open, like at the highest clock rates and so then, and with a default filter too, and so we've we're now realizing.

A

um We knew it all along, but now we we're getting a little bit more confidence with uh updating and incorporating the the correct profile. So I don't think we're at the point where it will work receiving it off of the air. Because of the things that you just said that you have to have the right rights and you have to have the right filters um and so very soon we will have to tighten that up and make it uh make it correct.

A

But we have been able to make things change over the air to go from. You know more than 100 megahertz wide signal to down to 50 megahertz, it's it's not as simple as um so in your code when you send off the command to the Pluto and you're able to change these things directly with the device rights, um at least on the on the 9371.

A

Some of it's tunable and some of it's not so the and and the transmit sample rate is one of the things that you have to actually set in the profile, and you can't. Apparently you can't write to the the um write that value and and change it dynamically. At least we haven't been able to um and then I'm not I, understand that we need to change the filters and and I and we do have the the filter uh utility.

A

So it's a it's a program from Analog Devices that we've been able to get running and we've been able to to change the filters um and and output the coefficients, and we haven't close the loop and put them back onto the device yet into the into the dev board. But we've got a good part of the recipe working.

A

So that's that's going to be very important, very quickly, I think as we clear up these other challenges and get these things working, that we'll have to tighten up the filters and the sample rates and the clock rates and and we're good, as Paul has pointed out before to me like, um since we have a relatively narrow band, uh the the 9371, the you know for 10 megahertz signals. This is actually the very lowest this chip can do.

A

It's the narrowest band, really it's capable of very wide band Communications, which is great and and people should use it for that. But for for the amateur satellite sub-band that we're talking about, we don't have a whole lot of bandwidth, that's pretty narrow, um and so in our case we definitely are going to have to have a new profile and new filters, and and all of that, so anybody that knows how to do that better than us, uh we'll we'll just keep plugging away and learning.

A

um But everything that every Easter said is is correct and, and we need to worry about.

D

Yeah, we will probably need an additional filter uh to do the rrc before the band limiting filter, that's built into the reference design. Is there an existing design for that? That's already been used on the project.

A

The Pluto has one so there's a good example. I think that's what Everest meant when he was talking about an example. I have a.

C

Yeah exactly yeah.

A

It's it's in the Pluto design. It's good yeah.

C

You can use the the filter uh like this and then it just apps sample by four and make a rrc filter. So so it reach more well higher bandwidth. So if you target 10 megahertz, then the sample rate on the profile should be 40. Megahertz.

A

Right yeah, with our with our roll-off that we picked.

D

Okay, that's I didn't realize that that wasn't already part of the encoder, but um good.

A

Enough, no, no, it's not it's not it's not part of the encoder. No, the.

C

No, no just it just after the encoder you just plug the rsf shooter, maybe um yeah.

D

Gotcha: okay, no problem.

E

A

Birds now see what I did there. It is not Dull here.

A

Packages and birds- oh my gosh, okay, I'm, just gonna. Let the rest of the household fix that um okay, any other comments or questions before we close out the business part of the stand-up.

A

All right, thank you. Everybody thank you for uh for hanging in there and doing such amazing work. It's uh it's a huge honor to be able to be even a small part of this, so we'll keep working and we'll be on slack and we'll do this again next week and uh and very much appreciate everybody all right and I'll be here for office hours and uh open discussion after this, uh but uh we'll we'll shut down and see you next week.