Next revision PCB to be submitted Oct. 5

[Steve Haynal]

Hello List,

I made good progress on the next revision PCB this weekend and plan to submit it for production next weekend. All of the most sensitive placement and routing (AD9866, BeMicro connector, power distribution network, bypass caps) is pretty much the same as versions 0.9 and 1.0, which have both been demonstrated to work. There are power supply and RF front-end changes as discussed in earlier posts. I will document these once the board is done. 

Since I plan to submit the design on October 5, this is the last week for me to gather data from those who built either the version 1.0 or 1.1 boards. Please post any data that would be helpful for the next revision. I believe Rob is the only one who has assembled a v1.0 board, but want to make sure to check with others. From his experiences, I have gathered:

1. Using two separate 5V power supplies, one for the BeMicro and one for the Hermes Lite is problematic. To address this, I have tested, and been using, a single power supply for both connected via the BeMicro 80-pin connector for several months now. 

2. Reset may need to be delayed. To address this, the next revision connects reset to the FPGA so that it can be fully controlled.

3. The SMA connectors are too close together. To address this, the next revision separates the TX and RX connectors sufficiently.

73,

Steve

KF7O

[Rob Frohne]

Good work Steve!

I will try and look over your board this week; school has me by the neck at the moment.

Can you explain your one power supply setup in a bit more detail?  I did a brief test, where I shorted the two pins that I expected to allow me to power the whole setup from the AD9866 side, but it didn't work.  I have been going to re-examine the schematic of the BeMicro to see if I misunderstood what was happening with that, but haven't yet.  Are you powering from the BeMicro side?  I was going to try that too, but I keep forgetting to get my USB extension cable from the office, and so I haven't done that yet.  From what I remember of my previous tests, you need to power the AD9866 slightly before the BeMicro on the v-1.0 boards.

I'm still itching to put components on the front of the v-1.1 boards, and I hope to be able to do that this week.  Maybe something I find there will be of use to you as well.

73,

Rob
KL7NA

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 
Rob Frohne, PhD, PE
Professor
EF Cross School of Engineering
Walla Walla University
100 SW 4th Street
College Place, WA 99324
(509) 527-2075

[Steve Haynal]

Tie PWREN2# low. With the prototype I do this by modding the BeMicro SDK. Connect the BeMicro SDK 80-pin 5V pins to the Hermes-Lite 5V. There are pads to jumper this in. The BeMicro regulator can not supply enough current so you must power from the Hermes-Lite. All this has been discussed on this list. Search for PWREN for the details.

73,

Steve

KF7O

[Rob Frohne]

You figured it out for me.  On the v-1.0 boards, I didn't connect pin 80 to 5 volts, as I did in the V-1.1 board.  I'll try that, and see how it works.  I did have the means for grounding PWREN2# on the BeMicro, also called PWPEN# on ours.

Thanks,

Rob

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Rob Frohne]

Hi All,

I experimentally verified that one supply will work for me.  I can power it up from only the Heathkit bench supply feeding the v-1.0 board and then on to the BeMicro.  When running two supplies, I had to turn on the v-1.0 board, then slide the BeMicro USB connector in to the laptop or 5V tablet charger.  When running on one supply it mostly works okay, but there have been times when I powered up and the preamplifier was not controllable as when I powered in the incorrect order with two supplies.

Both boards together draw close to 500 mA on receive.  I didn't test TX with this setup, but I expect it to be around the same because the draw on the v-1.0 board alone was about the same for TX and RX.

I also tested the very cheap tablet USB power supply that I got from China (with two outputs, 2.1A and 1.0A), powering everything.  I was looking for noise.  It is within 1 dB of the linear Heathkit bench supply on 20 meters where I was testing.  I had to use a 50 ohm load as the source for the antenna, as the QSB on the band made measurements within 1 dB impossible.  The cheap power supply was not exactly 5.0 volts though.  It was 5.275 V.  This is a bit high, as the datasheet on page 31 states the the V_cm should not exceed 5 V on the TX output to keep from dissipating too much power in the output stage.  At least it is within the absolute maximum ratings (< 7 V).   Presently I have a 1 ohm dropping resistor to try and insure that doesn't happen, but I don't like the voltage being high by 5%.  I have another of these cheap power supplies at the office, and I intend to measure its voltage too.

I left this v-1.0 running on the cheap Chinese supply, and you can listen on QtRadio or glsdr if you are curious.  The antenna is pointed toward Alaska.

73,

Rob
KL7NA

[Steve Haynal]

Hi Rob,

Thanks for the interesting measurements. I found no measurable noise floor level difference when experimenting with various power supplies a month ago too. I used a 50 Ohm load and was experimenting with regulators versus several PC power supplies. I did find radiated noise from some power supplies when my antenna was connected. This was probably pronounced for me as my antenna is only a foot away from my house.  I see some of this type of noise on your radio too, switching power supply noise at ~21.050MHz, birdies on 10M from monitors or ethernet. I also found that the noise varied by band and by power supply when my antenna was connected. When I had no noise on 20M, I found I had lots of noise on 80M, etc.

Since you are setup to take these types of measurements, would you mind reporting the actually dBm values you see in QtRadio? Do you see variation between your two radios? I suspect that since you flooded copper ground on the top and bottom of your fixes to my prototype, there should be some noise improvements. Although there may be unaccounted for variables, here is what I see. 

Settings:

Hermes-Lite running new firmware and default gain, +20 dB.

hpsdr-server running at 96kHz

QtRadio with defaults (slow AGC, no noise reduction)

USB

filter at 5 kHz

50 Ohm load connected

The readings were taken from the the dBm meter

-130 dBm at 4MHz

-130 dBm at 7MHz

-129 dBm at 10MHz

-128 dBm at 14MHz

-126 dBm at 21MHz

-125 dBm at 28MHz

73,

Steve

KF7O

[Rob Frohne]

Hi Steve,

I'll be happy to do those measurements to compare .  And you are right.  I have quite a bit of noise here due to various things, including now verified noise from Ubiquiti gear connected to Ethernet cables going up my tower.  The 48 V supply makes this noise that drifts around in frequency, and was annoying me. Then there are some really big noises on 15 meters that you also noticed, but they don't appear to be the Ubiquiti gear.  I haven't yet gotten around to switching things off at the breaker box and trying to locate the source of that, or even swinging the antenna.  I'll do a bit of checking on the radiated noise from the cheap Chinese tablet power supplies when I get back to that as well.  I was also using the dBm  meter in QtRadio for my measurements.

This afternoon, I finally got a chance to put the components on the top side of one of the v-1.1 boards.  My new solder stencil was much better than the last one, and I only had a few bridges to clean up.  I didn't even have to use solder wick at all.  I briefly tried firing it up by using the same BeMicro that had the firmware that I was for the v-1.0 boards, and it seems that the A/D is sending back all zeros.  I haven't done any other debugging other than to note that the power supply voltages at the regulators are okay and the current draw is 280 mA instead of 500 mA as it is in the working v-1.0 boards.  I did add the same 0.1 uF capacitor from the RESET* to gnd.

Tomorrow, I will compare the schematics, and try and remember what the differences were between the boards.  One difference is that the RESET* line goes to the FPGA.  If there are tests you want me to try and run before you submit your next board, let me know, and I'll try to do them tomorrow afternoon or Friday afternoon, assuming I can figure out why I'm getting all zeros and fix that first.

73,

Rob
KL7NA

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Rob Frohne]

Hi Steve, et. al.,

Changing the pin assignments solved the problems with v-1.1.

The measurements you describe below are the same within +/-1dB (sometimes + sometimes -) for both the v-1.0 and v1.1, where the 1.0 was powered on the linear supply, and the v-1.1 was on the  the cheap Chinese tablet USB charger.  With the v-1.0, it was roughly 4 dB worse on the USB charger on all bands.  With the v-1.1 on the linear supply, it was exactly the same as yours, except on 10 meters where it was a surprising 6 dB worse.

My other tests were at 48 kHz sampling rate, where all of them have a lower noise floor.  In fact I can't measure it without turning up more gain.  With all the gain in, it is -97 dBm or -98 dBm on all bands.  You can see it jump 3dB when you put the 50 ohm resistor on. 

It appears that as you supposed, I was taking too much trouble trying to improve the S/N ratio.

I did a wide spectrum using KISS Konsole, and it looks like this on V-1.1 running on the linear supply:
Note: I'm not sure about the relative dBm, because there appears to be a bug in KISS Konsole at least on Linux, that when you resize the window, it changes where the spectrum is displayed.  I am not so concerned with that.  I was looking for spurs, and it looks pretty good that way.

I'll leave v-1.1 running tonight so anyone who wants to can have a listen to it.

73,

Rob
KL7NA

On 10/01/2014 07:24 PM, Steve Haynal wrote:

[Steve Haynal]

Hi Rob,

Congratulations on now having v1.0 and v1.1 running! I would like to see the raw data from your measurements. Can you post s list/table of measurements with the settings you used as I did? If most of your measurements are when running at 48 kHz, I can update my settings to match yours, but you will see differences in processing gain between 96 and 48 kHz.

The wide spectrum interfaces to Hermes-Lite in KISS and cuSDR need work to be calibrated. I'm not sure how to interpret the data.

If I interpret everything correctly, we now have 3 versions of Hermes-Lite working, (0.9, 1.0 and 1.1) but have not measured much difference among them except on 10m. At least it is nice that the AD9866 appears to be robust! It may be a lot to ask for an open source hobby electronics project, but I think we should start looking for second order effects in a systematic way. For example, the wideband signals Joe reported a few days ago under the QtRadio receiver topic may be worse for a particular board revision. Also, last night I thought the noise floor was higher on 15M than usual, with some type of pulsing noise around 21 MHz. But I can't tell if this is something with v1.1 or your environment without the ability to quickly change antennas.

73,

Steve

KF7O