High End Per Vices Cyan Software Defined Radio Released
This blog is mostly concerned with software defined radios that are affordable to most hobbyists, but if you've ever wondered what the cutting edge is, take a look at the recently released Per Vices Cyan. This is a US$73,500 one channel RX/TX SDR with a tuning range from 100 kHz up to 18 GHz, ADC resolution of up to 16 bits, a maximum instantaneous bandwidth of up to 1 GHz (with 16 GHz if all channels are required), and an on board Stratix 10 FPGA. There are also higher end Cyan's, with the Cyan Mid having 8 RX/TX channels for USD$149,500, and the Cyan Pro with 16 RX/TX channels for USD$289,000.
Obviously Cyan is aimed at the research, industrial and possibly military market, but maybe this is the sort of capability we will all have in 10-20 years.
Their press release reads:
Per Vices Releases New High Bandwidth, Compact Software Defined Radio Platform
Ontario, Canada- February 20th, 2019
Per Vices, an industry leader in wireless platforms in North America, has been developing Software defined radio (SDR) solutions since 2006, offering customers high performance solutions. Continuing to grow their influence in the wireless communications, radar, signals intelligence, defence, medical imaging, and test and measurement markets. Their newest product, Cyan, is the best SDR available on the market, offering the highest bandwidth on a compact radio platform. Cyan continues to fulfill the company’s legacy by providing the market with the highest performance radio solutions.Cyan is designed to offer users with a customizable number of independent, phase coherent radio channels, up to 16 total, each offering a standard 1GHz instantaneous RF bandwidth. Featured on a high channel count, ultra wide band, high gain direct conversion quadrature transceiver and signal processing platform. Cyan provides the highest RF and digital bandwidth with an onboard DSP in a compact form factor.
On the digital front, the platform is designed around an Intel Stratix 10 FPGA SoC enabling significant DSP resources for a variety of applications. The platform also features 4 x 40 Gbps digital backhaul enabling ultra-high data throughput while maintaining low latency for applications that require raw radio data to be transferred to another platform.
To learn more about Cyan, Per Vices, or their other product offerings, contact Brandon Malatest at +1 (647) 534-9007, or email [email protected], or visit the website at www.pervices.com
Is there any new information / update on this monstrosity?
Good God, for $73,500 I better be able to use this thing to convert a small Cessna into a DIY AWACS plane!
JTC, it really looks like the kind of equipment that would be kept in a climate controlled laboratory environment rather than installed in a Cessna or in an AWACS :).
You are absolutely correct – our platforms are designed to meet customer requirements in a number of markets where the important aspect is meeting performance metrics and then looking at the price/performance ratio. We have found ourselves doing well in both of these aspects with our customers working with our platforms.
You are also correct that these are not designed as hobbiest or enthusiast use cases…or installation in a Cessna or AWACS 🙂
Just a small question, For the receive only 100 kHz up to 18 GHz radio and all screens and antenna up and running what would be the cost? (full system)
The eval board for the Xilinx RFSoC with 8x 4GHz ADC and 8x 6.5GHz DAC, 4 ARM cores, DDR4, Big FPGA etc.. that compares to the $150K machine is only $9K on a half length PCIe card!!!
To be fair, the fpga used in the pervices system is twice as big (900k logic cells vs 2700k logic elements). In addition the Xilinx board is missing a ton of the rf side. We will have to wait and see once Brandon gives more rf specs especially with praise noise and linearity. Building a system with that bandwidth and that frequency range has a lot of pitfalls. Just go take a look at test and measurement equipment, once you push limits in phase noise, bandwidths and frequency range at the same time cost goes through the roof. The costs associated with this stuff is not linear, it’s exponential.
Also sample level phase coherence across 16 channels, depending on implementation, is impressive at these bandwidths and frequencies.
I have high expectations for rf measurements. It better be amazing at those prices else I agree it’s the common theme of this thread.
Also, 75k for one channel is rediculous no matter how you slice it. A simple price analysis puts the rf boards at around 10k per rx tx combo. Meaning whoever pays for this monstrosity is paying over 50k for the fpga…
By praise noise I mean phase noise
Hi Fred,
Thanks for the comments and reply relating to the FPGA – you are absolutely correct, it is a monster with plenty of resources available for further processing!
We are getting all of the measurements together and working on customer specific requests for measurements. We are seeing very good phase noise and linearity but I will need to wait for all of the data to be formally documented before sharing. Just for reference, we are using Microsemi’s OX-174 with the best phase noise ordering possible as our common clock which allows us to keep to our performance goals.
Thanks also for explaining the cost structures associated with these types of platforms and you are absolutely correct that they are exponential not linear!
I am also happy to let you know that we do have deterministic phase coherency across all radio chains – this was a fun aspect to get locked down but we are very happy to have this met as it is a requirement for many of our customers.
Relating to the price, there are a lot of aspects driving this including the ADCs, DACs, clock source, and, of course, the FPGA. Although I cannot provide specific details on the exact prices of each of these components, I do encourage people to investigate on their own and you will quickly find that the price adds up really quick.
Thanks again for the feedback and conversation on this group – I will do my best to stay on top of answering all of the questions.
All the best,
Brandon
I did notice the avoidance of my one channel comment.
At least tell me the fpga is open to the user like ettus. Such a massive fpga for one channel and all it does is ddc/duc ?? And I’m paying 75k? Plus I’d need to build my own fpga based product to interface with it, I won’t beable to do anything useful in software at full bandwidth.
It was not intentional at all regarding the one channel comment.
The intended audience for this product are those who need a very high performance radio platform and in making such a platform, the costs are higher than let’s say, a lower bandwidth, limited FPGA resource, and ONLY receive platform. We do offer customers the option of different variants up to 16 total radio channels. The FPGA source code is not open source but we work with customers depending on their needs. I encourage you to reach out to us directly if you are interested in the platform and discussing more about your project needs.
Hi, you mentioned the FPGA source code isn’t open source. That is fine, it doesn’t need to be open source but is it able to be licenced, if I buy a unit, can do my own FPGA design?
Hi Mark,
Thanks for joining the discussion and I am happy to confirm that we do offer different licenses for our FPGA source code so you can build your own design. It would be great to learn more about your application to ensure Cyan is a good fit. Please feel free to contact me directly at [email protected]
Thanks and all the best,
Brandon
There is more detail if you go to their website and download the user guide and look at the block diagrams.
My reading shows all pretty standard parts RF mixer front end IQ demod with 390MHz BW, that is sampled by a two channel ADC running at 3GSample/sec. As the I/Q output BW is 390MHz total BW for IQ should be 780MHz, they claim 1GHz which they pump to an external PC over 40GHz optical that has 32Gbit throughput so 1GSample with some loss for filtering. So all the processing happens on a user supplied PC (assume this needs to be top of line beefy machine). But $78K for Gighertz sampling but no processing seems way way overpriced I’d be happy to built similar boards for $5K and make a tidy profit.
Hi Joe,
I apologize for the delay in getting back to you – as I am sure you can imagine, we have been pretty busy lately!
Thanks for looking into the User Guide and providing your feedback. The major block that relates to the BW is our ADC which as you correctly pointed out is a 1 GSPS dual-channel ADC – this allows us to reach a 1GHz sampling bandwidth and sampling 16bit I and 16bit Q.
The processing portion really depends on the application and the users needs. The FPGA offers plenty of resources, especially if there is not a need for all 16 radio channels, and we are able to put a lot of the DSP there. For some users it is more useful for them to process on external equipment or even store for future analysis.
Regarding the price side of things, we do our best to make products that are as competitive as possible considering the performance metrics. The FPGA is one of the latest and greatest from Intel (FPGA SoC) along with near top of the line converters (ADC and DAC), and a high stability clock to drive it all. Happy to answer any questions if you have them.
The reliance on that iq downconvefter is killing the viability of the rx board. Perhaps you want to do advanced filtering on the fpga, either way to say the user is getting 1ghz bandwidth is very misleading.
Hi Fred,
Sorry for the delay in getting back to you. You are correct, there is a difference between sampling bandwidth and application bandwidth. The metrics we are referring to are sampling bandwidth which is typically the metric of interest and presented for SDRs (as we have found).
Thanks for getting back to me and my criticisms.
Its typical to post RF bandwidth and Sample rate. In your spec sheet you have Sample rate and “RX Sampling Bandwidth” which is set to be the same as the sample rate, not sure what the purpose is there. It doesnt feel nice that we have to go digging to know what RF bandwidth we can get out of this expensive device. Also, does the IQ down converter change when upgrading the ADC to the 3Gsamp variant?
The truth is there is nothing comparable in the market. I am trying to figure out the business case. Technically, especially when all the channels are populated, it is an impressive accomplishment (although that opinion is pending more in depth RF metrics). Though for 1 channel it seems like I would be paying for a lot of hardware I have no need for (Huge FPGA, 4 40G interfaces and all that is needed to support that, what must be an impressive clock distribution for all the channels I wont be using, Power delivery ect.) For one channel it wouldnt be hard to build this with off the shelf subsystems and get likely better performance for what I need it for. In this price range you arent going to get people who are just looking to “play around”, they are going to have a specific need. Unless that need is to use the entire frequency range and bandwidth you provide then it wont be hard to create a system that caters to their needs for a much cheaper price and without the hassle of this 40G backhaul.
When I ask myself the question “Why isnt there anything else out there like this” I am less inclined to think that its because no else is capable of doing this, and I am more inclined to think that there is no need for it. The cost of trying to be good at everything (as an SDR needs to) with that channel count, that frequency range, and that bandwidth is far to great as you have demonstrated, it is far more cost effective to build something specific to your needs, and likely do it better (ex. out of band filtering). And again, this point becomes much more salient for the 1 channel case where phase coherency and synchronization is not a factor.
Earlier I asked if the FPGA was open as perhaps someone can make use of that massive FPGA to do some crazy low latency DSP… but if its not open then… whats the use case, who would buy this and for what purpose. Above you noted that one of the benifits of this product is that it doesnt have limited FPGA resources, my question would be what are you doing with the FPGA resources? If I am buying Cyan it means Id want access to very large bandwidths meaning there isnt much needed on the FPGA in terms of DDC or DUC.. Filters can be costly, but not that costly…
Anyway, as I said, likely (pending RF measurements) an impressive technological accomplishment. I just dont understand the “why”.
Hi Fred,
I completely understand your point and I am sorry if things appeared to be misleading. If you look at the architecture diagram you will notice that there are multiple paths depending on the frequency that you are operating at. If you are at baseband then the sample rate is not impacted by the IQ Downconverter, which makes it challenging to post all different scenarios in press releases, quick specs, etc. We always like to understand the end customers application prior to selling the equipment to ensure it is the right fit and that any details are shared directly with the customer purchasing. The 3GSPS does have some architectural changes but we have not pushed that version out yet so I do not have the final details to share.
You are also absolutely correct that our customers are not typically just purchasing it to play around with it and they come to us with a specific problem that needs to be addressed; we can then optimise a solution to fit their specific needs – sometimes this involves wideband operation and others it requires additional filtering or other changes to meet their specific requirements. We have experience working with all types of customers in varying markets and there is no one solution fits all BUT we are able to meet many requirements with the standard product and are able to meet all requirements with smaller modifications which reduces the time to market for many of our customers.
It is likely that there is not another product out there that does what we do because it is a challenge, to say the least. Although when we share the architecture diagrams it may not look challenging, there are a lot of things that come into play when designing a system of this complexity. We have already observed a need and have customers working with Cyan in a variety of markets. We do our best to just share which markets and for obvious reasons do not get into specific applications. Some of these customers have us perform changes to the FPGA firmware, others purchase a license to the source code, and some do not need any changes at all as most of the processing is done on another system but they need the high channel count and high bandwidth over the operating frequency we offer.
It would be great to speak with you about your application in case there is a fit but if you are looking for a single channel and limited bandwidth, it is likely that our Crimson TNG or another vendor’s hardware would be more applicable. Feel free to e-mail me directly at [email protected]
I don’t have an application in mind. I am very curious on the details of what these applications are, especially the ones I’m which apply to the one channel option. I feel I could make a pretty compelling system with off the shelve lets to fit most needs in that frequency range. And if I couldn’t, just selecting the parts you chose for the rf front end and making my own solution with a more reasonable fpga and without the extra bloat for extra unused channels, I could build a FAR cheaper system.
But hey, since I can’t think of an application that would actually make use of your one channel system maybe I can’t see what I am missing if I am missing something.
On the mention of your low band stage I took note that it isn’t iq so it’s bandwidth is halved.
18 or 20GHz tuners aren’t very standard and very expensive.
https://www.dsinstruments.com/store/products/20ghz-mixer-integrated-lo-hybrid-ku-band-converter/
1,200 dollars.
That is a mixer, not a tuner, many more components go into making a good tuner.
What do you mean ?
It’s a down converter with a built in lo. Put an amplifier in front , output it into another sdr that doesn’t cost 75k. Maybe add a filter or two.
That 1200 box will essentially (although not seemlessly) extend then frequency range of a much cheaper sdr.
The missing piece is the high bandwidth part as most if not all other SDRs have far less sampling bandwidth. But after this down converter it won’t be too hard to build the board yourself using the same adc pervices uses. Hell you can even use eval boards.
Yes that’s a lot of work, but as per my earlier comments, because of the 40g backhaul, you were likely to be developing a system anyway. It’s not the typical “function out of the box” that you get with other sdrs. You aren’t going to hook this up to your pc at 40g and run 256qam at full rate in gnuradio.
Hi Everyone,
I am from Per Vices and just wanted to let you know that we are releasing more details in the coming weeks. The product is currently in production and our engineers are getting further measurements to provide more technical info. If there is anything you would like to see, please e-mail us at [email protected].
Thanks!
“Per Vices”? What a terrible name for a company.
Too bad, that there is only little technical info available.
Wow, I wonder how many RTL dongles they managed to squeeze in that blue box to get THAT performance…
Thats actually a great idea. Bunch of dongles chained together via software with a fast UI like gr-fosphor… 🙂
Glad someone knew I was joking, thought I’d get more takers / responses claiming it was RTL-based.
Oh well, must try harder, I guess.