[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xylo-SDR] Earlier Wolfson vs TI Discussions



Checked back and it was almost exactly a month ago when Phil_Harman – VK6APH mentioned his early experiments with the Xylo and his student. MAN has this little group moved fast!


In spite of the fact that I will be sending out the Wolfson chips to as many as I can. Phil Covington – N8VB posted an excelling rebuttal of the “Wolfson Decision” and had done quite a bit more research into A/D and Codecs after Phil_H’s and our search last Spring. The TI chips are readily available, inexpensive and do some other things that Wolfson does not. I am republishing that early conversation, since the TI chips ARE the way to go if this project indeed is to eliminate reliance on commercial sound cards for SDR in any implementation. (Ahti’s comments re: do it all in the FPGA, not withstanding).


I am republishing that somewhat limited distribution conversation here for reading. Phil_C makes the point and I sure have been convinced. I also intend to order up the TI stuff. I am not trying to quash any experimentation. TI is the way to go in the future. All things even being equal (which they are not) the availability of TI stuff turns the tide.




Hi Eric,


On 12/5/05, ecellison <ecellison@comcast.net> wrote:

> Phil


> Great going. I think PhilH, (and I agree) will stay with the Wolfson.

> Also will use the Wolfson 8971 for general purpose mic in, with ALC,

> and speaker out, simple I2C interface. Parts in volume about 6 bux but

> who knows. I called local (Alabama) distributor and they are hunting

> 10 samples of 8785's. Had not heard anything as of today, but will

> call tomorrow and also inquire about the 8971. I think that if we

> develop a 'product' based on these chips, we will be able to get

> hundered quantities or more. I think they have distributors in GB and

> Germany where they can be purchased retail and I know folks in both countries pretty well enough to get quantities.


(Phil Covington’s rebuttal)


I feel very strongly about this so I will offer the following argument:


<argue mode>

I could understand staying with the Wolfson parts if the design was finished, the parts easy to get, or they had some compelling performance advantage.  If you were only talking about building one or two projects for yourself it wouldn't make much difference either.

Unfortunately the parts are difficult to get and do not offer any performance advantage.  The PCM4202 is equivalent or superior to the WM8785, readily available (in stock at Digikey), and there is a twin

PCM4204 part which gives you 4 channels instead of two - if more

channels are ever needed.   The WM8971 does not offer any compelling

advantages either - the TI 24 bit CODEC TLV320AIC23B gives you *all* that the WM8971 has and more.  It also has two DACs with PGAs and output drivers right on the chip.  The WM8971 is only an ADC - you will need to add a DAC whether it is an external chip or it is done with PWM in the FPGA. It would be much easier to just use the TLV320AIC23B's DACs.  The TLV320AIC23B is in stock at Digikey now.

Samples can be ordered from TI's website.

</argue mode>




I have researched a lot of ADCs, DACs, and CODECS, comparing their specs, availability, etc...  I have samples of all of them except the AKM parts.


> As long as the "Xylo-Phreaks" have enough to test and develop on I

> think we will manage to gettem in any end product. I offered my Xylo

> to PhilH however apparently he is getting a replacement. BTW: they are

> out of Xylos till Feb or so, which puts a quash on our group if it

> grows more. Dale has created an SVN for this project on Hamsdr.com.

> However, neither Phil nor I (or others?) know how to get a project

> like this going and how it will work. Perhaps you can help since I

> know you have done project management like this. PowerSDR is also going on there.


When there is some code to check in, send it my way and I'll get it into SVN.  I just need to get Dale to set up a user name and password (we will all need them for development).  On PowerSDR: We were considering Berlios until Dale offered to host the projects on Hamsdr which is a great thing as far as speed and availability.  I am very happy that PowerSDR development will be done through SVN -  I think a lot of people have been holding back because of the difficulty of coordinating PowerSDR development.


> PhilH asked if I would do a circuit board based on the Wolfson 8587.

> Then we discovered the 8951 or 8971 and also thought we should put I2C

> 16 bit Microchip 23016s on the board. Feature creep went rapidly and

> pointed to the problem of not having an interchangeable hardware system capable of updates.

> My thoughts are to create a 'motherboard' and buss structure board

> with perhaps 40 pin .100 pins as on your board. Include analog and

> digital regulated power 12, 5, 3.3 and 1.2v and run the I2C, user

> assignable Xylo pins, and I2C busses clks on the mother board. Watcher think?


Since we are using 24 bit ADCs in an attempt to get high dynamic range, it may be better to put only a 12 and 5 V supply on your motherboard and do the lower voltage regulators on the ADC board(s).


Where will the FPGA reside and will you stay with the Cyclone (and which part)? - assuming that you will not use the Xylo board in the final design.  If you are going to make this into a project (as opposed to a couple of units for personal use) I would highly recommend that the Xylo board is not a part of the design.  (Lack of) Quality problems alone with the Xylo will cause it to be a nightmare for users.  The Xylo board is very easy to duplicate but is fine for development use.


Altera is the way to go.  The only reason I used the XILINX Spartan 3 in my project s that it has 18 bit multipliers which makes certain things much easier and faster when you want to do the DSP in the FPGA.

 I would have much preferred to use an Altera FPGA otherwise since they are much easier to get.


> I'm willing to

> invest the time to learn to layout the stuff in ExpressPCB or other,

> since I am not a programmer and will probably be an 'assistant' where I can help.


This is where I would STRONGLY recommend that the schematic and PCB stuff be done in an open source tool.  If the design files are going to be shared then it does no good to have a "open" hardware project

using closed or proprietary $$$ tools.   Sometimes there is no choice,

but in the EDA end there is.  Using ExpressPCB locks people into using their service if they want to duplicate the boards.  Luckily the Altera and Xilinx design tools are free for download.


I would recommend either gEDA or Kicad.  IMO, Kicad is much easier to use and more polished of the two.  I have used PADS, ORCAD, Eagle, etc... and Kicad feels more like the professional tools.  I have the full blown $$$ version of PADS logic and PADS layout here for my paying projects, but I plan on doing all my open hardware PCB work in Kicad.  The free but limited version tools such as Eagle are an aggravation in the end - you always run into the size limits eventually.


Now if you think that no one would be interested in duplicating the design themselves then ExpressPCB would be ok.


> That way we can add 'daughter' cards interfaced to the Xylo. The board

> can be placed down on the CONIO1 pins. In the end the entire prototype

> system could be a product external to the SDR-1000. The Xylo also

> gives us graphics display capability etc.


With the motherboard and other daughterboards you are not too far away from just replacing the SDR-1000 especially for RX only use.  All you really need to do is add the QSD on the ADC board and design a DDS board :-)


> In any case, I have lost track of all the xylo owners but know a few

> and will put together a mailing list.


> Eric


Maybe we can ask Dale to create a Xylo section on his Hamsdr forums?


73 de Phil N8VB