Difference between revisions of "HERMES - Design Discussions"

Latest revision as of 10:17, 25 January 2010

This page is for the design discussion on the HERMES DUC/DDC Transceiver. Discussion proceeds with most recent at the bottom.

Update (16 December 2009)

picture courtesy Phil Harman VK6APH

Update ( 28 April 09 )

Since the announcement of the Hermes project there has been quiet a bit of traffic ‘in the background’, in summary the hardware details are now reasonably stable. What follows are some snippets of traffic to provide readers with some insight to the development process ......

Phil VK6APH has provided several suggestions, namely:

There needs to be an audio out from the TLV320. JTAG - the JTAG programming can come from the FX2, perhaps we have a standard JTAG header and jumpers to connect the relevant pins on the FX2 so we can use either? Need to show the LPF at the RF input. We could take the forward power from the 15w PA as well since I think there are enough spare inputs to read all we need. ICD header needs to connect to ADC78H90. If there are spare inputs should we bring them out - perhaps for use with an external amplifier etc? 122.88MHz clock directly feeds DAC. Phil is also trying to find a packaged LPF so we can get rid of the toroids on the DAC output. Add a header for I2C signals from the FX2 to control external equipment.

Graham KE9H has made an interesting suggestion regarding the gain distribution of the existing Penelope 1/2 W PA strip :

As far as the gain re-distribution to get the transmit frequency response flat out to 50 MHz, you basically change the transformer T3 from a TC4-1T to a TC1-1T (which will maximize linearity of the AD9744,) and add a second OPA2674 in series with the current OPA2674 OpAmp, and redistribute the gain between them. The current OPA2674 is being run at 24 dB gain, and is hitting gain-bandwidth limits around 30 MHz. With two in series, they would each be set for around 12 dB gain, and the combination should be flat well over 60 MHz.

Graham is also investigating the use of a THS6012 as a substitute PA, this provided a very clean output, at a higher level, a test board has been drafted and a few boards have been ordered to enable to testing. I’m not sure this chip will form part of Hermes as the supply requirements are beyond the current spec, but I guess the final decision will be made once the test results are in :)

While I’m discussing the PA strip, I must reiterate that the PennyWhistle PA is really just a dream for Hermes, I really don’t think there will be enough room on the board, however I will squeeze everything one first, and if there is room, PennyWhistle will be considered.

Graham has also suggested some form of audio PA be included, this has now been added to the hardware block diagram.

Scotty WA2DFI, has made several suggestions and between us we have now settled on some of the connecting hardware. To save board edge space a number of the connectors are now ‘stack’ connectors. Scotty has also made a good start on the schematics, I’m not going to provide a guesstimate for release for comment, mainly due to Scotties Dayton commitments, but it’s going to be weeks not months :)

Phil N8VB has provided some valuable information, namely:

Take a look at how I connect the FX2 to the FPGA using port E for 8 bit parallel configuration. After configuration I use this as the 8 bit communications bus between the FX2 and the FPGA. This allows faster configuration of the FPGA and faster read/writing to the registers in the FPGA (especially for tuning word). You might want to use this scheme in Hermes.

Also, instead of using an Altera configuration device (especially for the larger '40 Cyclone III), consider using a SD card which can have a serial interface to the FX2. The FX2 can read the SD card and configure the FPGA via the parallel mode mentioned above.

Phil’s advice regarding the use of port E will be incorporated in Hermes. The move to an SD card is still in the balance as board real-estate is of a premium and I don’t understand / can’t see any benefits for the change? Can someone out there enlighten me?

Phil N8VB, has also enlightened me on capability of the FX2 (although I’ve had to delve into the technical reference and ‘Google’ to ‘partially’ understand USB end points):

The FX2 has 4 high speed endpoints. This is enough for full duplex operation of the receiver and transmitter via USB.

The current command and control scheme will need to be restructured to achieve this and is well worth pursuing.

Well I guess that pretty well brings everyone up to date with the Hermes story so far, so in summary, the hardware is close to being finalized, Scotty has made a start on the schematics, and I’m climbing the learning curve of OrCAD, in preparation to layout the board :)

Kevin - M0KHZ

23 May Comments

From lstoskopf:[1]

I have one of the QS1R boards which does not have a preamp. Input

voltage for the AD is basically Vdd plus the base breakdown voltage. I popped the AD by turning an external pre-amp on before turning the QS1R on. The preamp alone could not put out more voltage than the AD could stand IF IT WERE TURNED ON. Phil C. left the preamp off so that the QS1R could be used in undersampling mode. I would prefer to have the preamp chip in line to protect the expensive AD, even if it does reduce the dynamic range a bit. Or a jumper

block/solder link to put it in and out. Comments welcome. N0UU

From Lyle Johnson:[2]

Interesting observation. Might need some protection (that doesn't

compromise IMD) at the input to the board with the ADC. One of the

problems with building blocks...

Or, a better preamp. One can be made with discrete parts that could

outperform the LTC6400. Or maybe the dynamic range isn't compromised enough to worry about. I know we faced this issue with Mercury and decided the convenience of the preamp for some applications might outweigh the small reduction in dynmaic range it caused. There was

lively debate on this topic!

From Phil Harman:[3]

Looks like a case of dejavue all over again!! One implementation of Hermes

is to use Alex for its HPF/LPF/antenna switching and 6m preamp. In which

case it will need the LTC6400-20 preamp unless we want to redesign Alex.

A preamp is essential before the LT2208 if you want to have usable

sensitivity on the higher HF bands. As many are finding with Mercury, the inherent selectivity of your antenna system is often sufficient for satisfactory performance on the HF bands without any additional filtering. By not putting a pre-amp on Hermes you are forcing the user to use an

external one.

Having been through this preamp - no preamp loop a few times I recommend we

leave the LTC6400 preamp in circuit at all times and use a switchable

attenuator to take out the excess gain i.e. just like we do with Mercury.

From Lyle Johnson:[4]

I am currently using Mercury with a 3-element SteppIR. I'm sure odd

harmonics are getting in but I suspect it does a passable job of rejecting severely out-of-band energy. I am in a reasonably quiet location. I've found that Mercury doesn't need the preamp on 20 meters although the increase in band noise when I connect the antenna is very

slight.

I've been doing side-by-side comparisons with a K3 (surprise!),

IC-756PRO2 and Mercury. But that's a tale for another day when I have

more anecdotal data to help obfuscate things :-)

24 May 2009 Comments

From Alex in three messages:[5], [6], [7]

would it be possible to add an atlas connector to the hermes board so it

could interface with other boards even if not the right size, it could

use an adapter cable and be mounted externally

is that rf volts or dc volts, if dc then wont the transformer isolate, if rf could we not put protection in.

would it not be better to put linked in/out preamp and let the user choose

25 May 2009 Comments

From Henry Vredegoor:[8]

Form factor / Casing:

I think one of the aims of creating Hermes in the first place was to have a

kind of "stand-alone", compact, Penny/Mercury/Ozy setup?

So not the Atlas-bus interface and -connector requirements/overhead.

Kevin Wheatley: Yes this is correct.

Then why stick to the default form-factor of Atlas/HPSDR (Eurocard) board

size of 200 X 100 mm. ? For a stand alone and "mobile" use, maybe a different form factor could be a

better choice?

What about a mini-ATX form-factor of 170 X 170 mm.?

This could possibly make using (mobile/Car) mATX-PC casings and 12(13.8) Volt PSU's etc. a nice and cheap(-er) option. This slightly bigger size could also provide some extra space required for the small PA idea. A companion mATX-PC (e.g. An Intel Atom based mATX-PC board) in the same

type casing would make a nice set.

Kevin Wheatley: This idea of migrating to a 170x170 definitely has some merit, the extra space will be helpful. I'll do some research while considering this idea.

Mobile/remote use:

It would be nice for mobile/remote use if the power supply input voltage

range could be extended to say from as low as 10-11 Volts and up to 14-15 Volts (typ. 12 Volts) This would require some voltage up-conversion for the +12 Volt supply (if the input voltage is at its lowest spec voltage of 10-11 Volts) since the +5

V, +3,3V are probably regulated down from the higher supply input voltage.

From FRANCIS CARCIA:[9]

1. I wonder if 1394 interface would be worth doing giving more horsepower than USB2.

2.The idea of a transceiver on one board fits right in with my present project interfacing to Penny and Mercury working my way up to 1.5 KW.

3. This would leave atlas for a VNA or other test equipment options.

26 May 2009 Comments

Possibly I am showing my history in not trusting USB interfaces. I would advocate for Ethernet as a first choice and fire-wire as a second. What are the downsides of these two?. Regards, Richard Ames, VK2NRA 07:02, 26 May 2009 (UTC)

Richard you could be right :) however from my perspective I believe the issue is with software developers (or lack of them activly contributing) at present we have an interface to PowerSDR that works, we also have an interface to Phil's KISS console, not being a software man myself and not wanting to impose further work on the software team I think Hermes should remain USB. Hermes can with a some fpga code changes operate full duplex across USB2, this will be an improvment to the current protocols. Perhaps when Ozy2 design has been proven, maybe there will be a need to more Hermes to that interface (whatever that may be) - Hermes2? So from here USB remains the choice for Hermes. 73's Kevin User:M0KHZ 18:01, May 26, 2009 (UTC)

Seems like good reasoning! I suspect the code to do all these interfaces exists in reusable blocks but I am not familiar enough to volunteer! I had thought noise (rfi) would be the problem... Regards, Richard Ames, VK2NRA 21:52, 26 May 2009 (UTC)