V2 Test Frontend with pre-V2 Amp

[John Williams]

pre-V2 amp team,

Wondering if any of the kits are finished and functional. I sent 9 of
them out.

Anxious to know how well the TX reconstruction filter works as a system
married to the amp.

Anyone have any spectrum data captured?

Any operational anomalies with the band switching.

How does power output look per band - relatively balanced and hitting
the 5W goal?

John

[Sid Boyce]

On Saturday I couldn't find 2 RD16HHF1 which I thought I had here a long
time.
I ordered 2 from a UK supplier via EBay expecting them to arrive Monday
or Tuesday.
Ready or collection tomorrow.

I'll build the Pre-v2 frontend using the build notes - bypassing the
PE5429's, etc. to test it with.
73 ... Sid.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[Glenn P]

Hi Steve
Been sidetracked here with house duties. Currently tiling the floor in a large room.

I am a little concerend though that we seem to have two or three versions of re construction filter of which i have assembled one onto my v1.2 rf board. I published a sweep of it and also the Rx LPF here recently but yet to hook up to my pre V2 PA/lpf board. The latter only has one lpf fitted and no relays.

I did receive the 12 way IDC sockets though yesterday from ebay.

I am also unsure which transformer version one should be using on the RF board for the Rx input?   I wound mine according to the wiki info. But the 1.2 RF Sch shows a different ratio?

I am also have difficulties with PowerSDR on my new PC.  I installed it on a Win8.1 PC but cannot get any audio output from it, when connected to H-L.   I followed the instructions to install P/sdr on the wiki but alas, no worky. Alans software works just fine though. Its the only PC i have that is 64 bit windows and grunty. 

glenn
vk3pe

[Steve Haynal]

Hi Glenn,

A lot of stuff is up in the air right now. Sorry about that, but this project is 80% "science project" (which I enjoy) and 20% "make a product" to me. It should eventually get sorted out. I am experimenting with a 8:1 impedance transformer on the receiver and hence the difference on the schematic. Either 9:1 or 8:1 should work okay. Virtual audio with PowerSDR is always a struggle. You can ask on the apache labs or openhpsdr list for help.

73,

Steve

KF7O

[John Williams]

Glenn,

Per the wiki, did you setup the VAC1 screen and enable it "Next, to hear audio, enable VAC1 as identified by a red arrow near the bottom as shown below. Unlike the Hermes, the Hermes-Lite has no onboard audio DAC so virtual audio must always be used."

Select the appropriate input and output devices on the screen. Then click the enable option on the upper left corner.

Also, on mine, I have to select the stereo option on the VAC1 screen to get audio.

I can send a screenshot if you need one.

John

--
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.

[Glenn P]

yes i did all that (Per the wiki, did you setup the VAC1 screen and enable it "Next, to hear audio, enable VAC1 as identified by a red arrow near the bottom as shown below. )
 but just in case a screen shot may help....... i am stumped. I had no such problems on my Win7 Laptop. Its just worked. I am not a PC nerd. (nor any other nerd)

The new PC is an Pent I7 running win8.1. Works just fine with Alans software which i have tried for WSPR decoding (works great)  But really want PowerSDR on the new PC.

glenn

[Glenn P]

OK Steve
I do think its best to all test using the same hardware as much as possible.  Get more meaningful results for the20% or at least find if on the wrong track.

glenn

[Stew KF5KOG]

Hi John,

I should have mine done today...finally some hobby time :)

73, Stew

KF5KOG

[Stew KF5KOG]

I see the T1/2 ratios are different on schematic, how should I wind them?  Or is it OK to use the standard ratios as in the original front end board?  

73, Stew

KF5KOG

On Wednesday, November 18, 2015 at 4:30:54 PM UTC-5, John Williams wrote:

[John Williams]

See my wiki article. Tx is same and rx is 8:1 with no center tap. Also, I skipped the rx filter. Not performing well.

--

[kf5kog]

That"s the info I needed, Thanks!

Stew

[Stew KF5KOG]

Here are my scans of the TX and RX filters for the V2 Test front end.  I used the SMD inductors per BOM.  RX looks good.  TX not so good. 

PE switches are not installed. Connection made directly to PE switch output and SMA connector.

 I will try toroids for the TX filter and retest.

73, Stew

KF5KOG

On Wednesday, November 18, 2015 at 4:30:54 PM UTC-5, John Williams wrote:

[John Williams]

Stew,

The filter that we are trying now for TX is in the wiki. Unfortunately this got chosen after I built the kits. You may be able to get close if you have some 100pf caps and using the 270nh and 390nh in the kit. Just double the 100pf caps for 200pf. Since we have a 10M roofing filter on the amp, I skipped the RX filter.

John

--

[Stew KF5KOG]

Ok John I will build it that way.  Anyways, here are my results using T37-10 Toroids.  L1 7 turns, L2 10 Turns.  Looks much better.

Stew

[Stew KF5KOG]

Here is my results for TX per the wiki.  I do not have smd inductors so I used T37-10.  L1/3 11 Turns, L2 12 turns.  All equally spaced 270deg around cores.  I did not try to adjust spacing.

Looks pretty good.

Stew

KF5KOG

On Wednesday, November 18, 2015 at 4:30:54 PM UTC-5, John Williams wrote:

[John Williams]

Stew,

The main concern with the TX reconstruction filter is suppression of  spurs in the VHF and up regions. Can you run a spectrum analysis into that frequency range? We need to have all spurs be a minimum 60db down above 30MHz.

Thanks for your help,
John - W9JSW

--

[Glenn P]

I did this in another thread:-  https://groups.google.com/forum/#!topic/hermes-lite/MaooZxtlHws

My post also shows the values used.
Only plotted though to 200Mhz from memory and the suppression was getting worse as freq went up...... some due to radiation in test setup between cables but not all.

glenn

[Stew KF5KOG]

I have attached my transmit sweeps using the V2 Test Frontend and Pre V2 5 Watt Amplifier.  I am still using separate 30 and 20 meter filters.

Output from HL is ~10dbm

The first PDF shows each band out to 300mhz.  Second PDF shows a few of the bands out to 3ghz.

From what I can see it looks like 12 meters is the worst.  Maybe use its own filter?  Use the empty spot when combining 30/20 meters?

Other than 12 meters, the spectrum looks pretty clean on my test equipment.

73, Stew

KF5KOG

On Wednesday, November 18, 2015 at 4:30:54 PM UTC-5, John Williams wrote:

[John Williams]

Was thinking that a measurement with hand wound toroids would be another data point.

[Glenn P]

Stew

is your reconstruction filter built per the Sch on the wiki?  We tested that before, Graeme and I and at the time concluded that the spread of parts meant that it was probably not repeatable.

The test I did was using the the 7 pole version 100p, 330uh,200p,390uH,200p,330uH,100p

Glenn

[John Williams]

Glenn,

Correct me if I am wrong, but I think the issue with repeatability was with the 17M/15M filter.

I thought that the TX reconstruction filter was pretty solid. You used SMD components, correct? Stew used had wound components.

Am I recalling this correctly?

John

--

[kf5kog]

Glenn,  I used the Wiki for the final assembly.

Stew

--

[Glenn P]

Yes i think you are correct John.    I should keep a log book of these tests....

But for some reason I thought we decided on the 100p, 330nh,200p,390nH,200p,330nH,100p filter ?

You can see from my and Stew's results that they are vastly different in attenuation of around 60MHz for example.

I used all SMD parts in my build on the pre v2 RF 1.2 pcb.  The 330nH were 0603, (Q=25)  the 390nH as supplied, 0805? ( I have a small kit of smd inductors, 0603)
Caps all 0603


glenn

[Glenn P]

Stew, just to confirm, your pdf' plots are for the Tx from the prev2 RF1.2, to the PreV2 5w stage output including all filters.

ie you have a fully built Tx chain.

I don't have this.  I have only tested filters in isolation. 

My prev2 RF1.2 is fully loaded with values as stated. 100p, 330nh,200p,390nH,200p,330nH,100p

My preV2 5W board has one filter only fitted....17/15m ? No relays yet either.

So I am not in a position yet to do a full system test.   [I also need to somehow get my PC to work with PowerSDR. Not tried it again as yet]

glenn

[kf5kog]

Glenn,

Yes I am using that setup with fully built TX chain.

I am using these values for reconstruction filter, 100p, 330nh,200p,390nH,200p,330nH,100p

This inductors are hand wound on T37-10 cores.  I used the turns suggested by mini ring core calculator.  No adjustments made.

Stew

[John Williams]

Perhaps the Q of the hand wound toroid is significant?

[Glenn P]

I managed to get PowerSDR running with virtual audio.   I changed from a wireless dongle on the PC to a wired in LAN connection.

Not really related to this thread but I can hear -130dBm from the HP8640B sig gen with Alan's ap. on 40M. That's directly into the V1.2 front end which is fitted with all parts.

But Only can hear about -100dBm signal on PowerSDR...... I would assume that i have a setting wrong in Psdr.

glenn

On Tuesday, November 24, 2015 at 11:25:36 AM UTC+11, John Williams wrote:

Perhaps the Q of the hand wound toroid is significant?

On Mon, Nov 23, 2015, 6:04 PM kf5kog <stewk...@gmail.com> wrote:

On Mon, Nov 23, 2015 at 6:53 PM, Glenn P <glen...@gmail.com> wrote:

snipped:---

[John Williams]

Try varying the Step-Attenuator setting. See the wiki for more info. Alan's program probably has a HLoptimized default. PowerSDR does not.

[John Williams]

Excellent work, Stew. Plots look pretty good. The 30/20 and 17/15 will make it a bit better. I was concerned that there may be spurs above 30 MHz that the Cauer may be less effective at filtering. Does not look like there will be a problem. 12M is messy but still meets FCC requirements.

John

--

[Stew KF5KOG]

Thanks John,

I'm thinking about trying the 30/20 filter.  Then put in a case with some shielding between stages.  May not have time until after Thanksgiving.  Dinner at our house this year.

73, Stew

[Steve Haynal]

Hi Stew,

Thanks for the data. Congratulations! You sweeps looking surprisingly clean. I am starting to test the reconstruction filter (without PA) and am not getting such clean results, perhaps partly due to my scope/spectrum analyzer. Just to confirm your design:

• T1 for TX was wound on a BN-43-2402 as per the original frontend, 4T bifilar on AD9866 side (8T with center tap) and 6T on filter side for (8.0/6.0)^2 = 1.78 impedance transformation?
• A 100 Ohm resistor was added across the AD9866 side of T1 to reduce power as per John's latest wiki instructions?
• Instead of AIML SMT  inductors, you wound your own on T37-10 cores?
• Did you reduce power at all by audio level or power level adjustment in PSDR?

I'm using the AIML SMT inductors, original T1 winding, full power and no 100 Ohm resistor. I added copper braid from the ground of T1 to all filter capacitor grounds. This improved things by 5-10 dB. I'm looking at specific spurs and see the following with and without filter when transmitting on 29 MHz:

2.456 MHz (5th harmonic aliased back in the DAC)

No Filter: -59.7 dBc

With Filter: -34.4 dBc

This is surprising. Is the filter putting additional load on the DAC? Do you see this spur?

58 MHz (2nd harmonic, no aliasing)

No Filter: -46.2 dBc

With Filter: -37.7 dBc

Again surprising. The 2nd harmonic is worse for some reason?

87 MHz (3rd harmonic, no aliasing)

No Filter: -34 dBc

With Filter: -44 dBc

Finally some filter improvement.

44.728 MHz (Troublesome alias at 73.728 - Ftx)

No Filter: -49.7 dBc

With Filter: less than -65 dBc

118.456 MHz (Expected alias at 2*73.728 - Ftx)

No Filter: -13.7 dBc

With Filter: - 54.9 dBc

Filter providing around 40 dB here

176.456 MHz (Expected alias at 2*73.728 + Ftx)

No Filter: -31.7 dBc

With Filter: - 48.8 dBc

147.456 MHz (Frequency that the ADC interface is running at)

No Filter: -51.7 dBc

With Filter: -39.8 dBc

Is the filter picking up this frequency?

I'd appreciate any data you can provide for a similar setup on your end. There may be problems with my filter. I need to try with the 100 Ohm resistor, and inductors on T37-10 cores. With the first version of John's PA, I measured output with a proper spectrum analyzer at OIT. I too saw that the PA+filters had an overall very good effect.

73,

Steve

KF7O

[John Williams]

Stew's results include the entire TX chain, so there is the reconstruction filter on the frontend, being further filtered by the roofing filter on the PA that is in line for all bands for both RX and TX. It is a cauer filter.

--

[John Williams]

Steve,

I think using a 4 layer board with the filters on the top and the copper
layer on level 2 makes the filters perform superior to a 2 layer board.
Using this on the PA has given us much more repeatable results from
initial testing on a breadboard to similar performance in circuit. I
have no way to validate it but am guessing that the stray capacitance is
much lower on the 4 layer board.

Just my 2 cents...

John

On 12/7/2015 12:40 AM, Steve Haynal wrote:

[Glenn P]

Do you know the layup of the 4 layer board?  typical is 0.4, 0.8, 0.4mm   ie 0.8mm internal glass.   Some manufacturers may vary this.

I imagine if you didn't clear the internal copper under the filters then the capacitance from either side to next copper capacitance will be higher.

glenn

[John Williams]

Glenn,

The total thickness is 1.6mm so your typical is correct. The layout is filters on top, copper on next, power on next and control traces on the bottom, for the most part.

What would be your hypothesis on this? Is it perhaps because of managing impedance to and from the relays?

John

[kf5kog]

Hi Steve,

T1 is per original design specs,  T2 wound per wiki. 8T(no center tap) and 14 Turns

I did not add the 100 ohm resistor.  I limited my power to 41% with drive slider.  I will try the added resistor with full drive tomorrow.

Yes I am using T37-10 cores.

My test were done with the Amplifier connected.  I am no longer hardwired to the Frontend so I can easily test the output of the HL/new frontend.

I should be able to do some testing with your setup tomorrow.

Stew

KF5KOG

--

[Glenn P]

Hi John,

for that pcb layup, the calculated track width for a 50R track is approx 20mils (0.5mm)  .   I doubt that at HF, though there is much to gain by changing it from whatever it is now.

On Tuesday, December 8, 2015 at 8:46:14 AM UTC+11, John Williams wrote:

Glenn,

The total thickness is 1.6mm so your typical is correct. The layout is filters on top, copper on next, power on next and control traces on the bottom, for the most part.

In  a past life at higher freqs we used to clear the layer down to reduce capacitance effects but I think the relatively small 'C' induced will be swamped by the filter values anyway in most cases.

glenn
 

[in3otd]

Hello Glenn,

for that pcb layup, the calculated track width for a 50R track is approx 20mils (0.5mm)  .   I doubt that at HF, though there is much to gain by changing it from whatever it is now.

 

I agree, since even at 30 MHz, where the wavelength is 10 m in air and maybe 5 m for traces in the PCB everything appears lumped, so we will need to take into account the contributions of everything connected to the traces. I mean, I expect the relays to be fairly inductive, so a bit of excess capacitance may actually help in compensating that. I opened up one of the relays, will post pictures later, of course they are not built thinking at HF performances, hi.
The filters ultimate rejection depends on a lot of details, I'll expect that the coupling between the biggest components (toroids, relays) is one of the main contributors to that.
 

In  a past life at higher freqs we used to clear the layer down to reduce capacitance effects but I think the relatively small 'C' induced will be swamped by the filter values anyway in most cases.

yes, I've done this a lot, but always in the GHz+ range. If someone has a TDR it will be interesting to see where the discontinuities are, just out of curiosity. But if you use the appropriate risetime (about 10 ns for a 30 MHz BW) I guess you won't see much.

73 de Claudio, IN3OTD / DK1CG

[Stew KF5KOG]

Hi Steve,

I was able to run some tests today.

I added the 100 ohm resistor to the input of T1.  My output at max drive is around +10 dbm.

The attached scans are at max drive using only the new frontend V1.2.  No PA attached.

I do not have enough markers so I had to do a screen capture twice for all your frequencies of interest.

I did not see the 44.728 spur, and I also have a spur at 7.509 mhz that you didn't mention.  It appears to be real because it is affected by adding attenuation to the output signal.

The spurs at, 2.456, 58, 147.456 mhz are significantly less than yours.  I hope someone else can confirm this.

I'll try the copper braid trick and see if there are any improvements.

73, Stew

KF5KOG

[Stew KF5KOG]

With copper braid(RG178) from ground of T1 to output filter capacitor grounds.

Closer in spurs(<=58mhz) were not as good as before by a few db, but I suppose those are within the margin of error of my equipment.  the others seemed to improve by a few db.

Stew

On Monday, December 7, 2015 at 1:40:33 AM UTC-5, Steve Haynal wrote:

[in3otd]

Hello Stew,
why the TX out at 29 MHz reads out at about -7 dBm? Is there an attenuation/offset in the readings that needs to be taken into account?

For comparison I'm enclosing a couple of pictures taken with my setup, which is slightly different than yours (for T1), hope this does not add confusion:

H-L with v1.2 Basic Frontend,  TX transformer T1 there with a 4 turn winding and a 6 turns with center tap on a BN-43-2402 core, 100 ohm termination on the AD9866 IOUT output (R2) . TX output filter on the frontend board 100p, 330n, 200p, 390n, 200p, 330n, 100p, no ground braid added.
TX at 29.0 MHz, power output +10 dBm, which is slightly below the maximum I can get with this configuration.

Spectrum without the TX filter (directly at the T1 output):

Spectrum with the TX filter (at the P1 output on the frontend, T1 output jumpered to the filter input):

As you can see, I have no 7.509 MHz spur, but I can see the 44.728 MHz one (and the 2.456 MHz is maybe a couple of db higher with the filter)...

73 de Claudio, IN3OTD / DK1CG

[Glenn P]

I hope to do same test within a day or so also.

I have the 100R fitted to mine also. Same values for re construction filter so it will be interesting.

glenn

glenn

[Stew KF5KOG]

Hi Claudio,

I am using a step attenuator.  During those tests it was set to -16db.  I meant to only have 6db but my aging eyes missed that the 10db dip switch was also on.

I have since tried to replicate your results without attenuation and my results are not very good.  Matter of a fact Steve's results are better than mine.  I decided to gradually add attenuation, 3,6,10 db.  What I have found is the carrier decreases by the amount of attenuation selected, but the spurs decrease by as much as twice the attenuation!  even when I zoom in on any particular spur, the readings are the same.

Stew

[Glenn P]

sounds like you are overloading the SA ?

glenn

[in3otd]

Hello Stew,
as Glenn said, to input to the SA is likely too high. How much attenuation is needed depends on the particular SA type, of course, but FYI I have an external 20 dB attenuation (this is the reason of the "20 dB offset" message on the screen) and the SA is adding another 20 dB internally.

73 de Claudio, IN3OTD / DK1CG

[Stew KF5KOG]

Thanks Claudio and Glenn,

That makes sense.  I did not notice that ATT reference, my software doesn't have that option.

Now I can find the sweet spot, and that may cure some of the erroneous spurs I see.  I'll rerun the scans in a few days.  12hr shifts at work the next few days.

Stew

kf5kog

[Steve Haynal]

Hi Stew and Claudio,

Thanks for the measurements and discussions. At 10 dBm output things look very clean. With some experiments several months ago, I saw as expected the natural harmonics decrease much more rapidly than the fundamental as the fundamental was decreased. What I didn't see was spurs from the DAC (5th harmonic aliased back) or those suspected from the AD9866's interpolation filter (73.728-Ftx) decrease more rapidly than the fundamental. For example, the 2.456 MHz spur when TX is 29 MHz varies in step with the fundamental amplitude. Only when reducing the range of values sent to the DAC does it diminish more rapidly than the fundamental.

I hope to test a few more configurations this weekend too.

73,

Steve

KF7O

[Glenn P]

I did a few tests today also....

Basically compared the basic RF V1.0 to the V1.2 board.

The V2 board was tested with and without the 100R across T1.

All tests done at 29MHz as Stew has done.

I didn't interpret the result too much, others may draw some conclusions from them. But it would seem the output is very clean apart from the consistent output at approz 150MHz (which may actually be 2 x the 72.xMHz Ref xtal?  SA's are not renowned for absolute freq. accuracy.

With the 100R fitted, output of around +10dBm is achieved with drive setting of 100.

Sorry, no time to measure other bands.

see attached pdf:

glenn
vk3pe

[Stew KF5KOG]

Hi Glenn,

What I noticed from your measurements, the spur that appears to be at 2.4mhz drops ~7db with 100ohm resistor across T1 input.  It also appears to be drive related, at 7.4dbm and less the spur is in the noise floor.

Stew

[Glenn P]

Yes, I overlooked that end of the spectrum !  I will do some more tests later in a narrow bandwidth

glenn

[Glenn P]

OK a test using a narrower BW of SA (50MHz showing 2.4MHz levels at drive of 100 and 32.

The drive of 32 was done to reduce the wanted 29MHz by 10dB to see the effect on the 2.4MHz and other 'spurs'.

Definately affected by 'something' in the AD9866

see attached:

glenn
]vk3pe

[Steve Haynal]

Hi Glen,

Thanks for the ;measurements and reports. They are very informative. 

Do you think your ~150 MHz spur could be 73.728*2=147.456 MHz? In the full-duplex version, this clock is generated on the interface of the FPGA and may be radiating some. You could try shielding the Hermes-Lite. You could also try the half-duplex firmware as that runs on v 1.22 too.

It is interesting that you see a reduction in the 2.4 MHz spur for power reduction. My experiment where I did not see reduction in this aliased back DAC spurs was for the 22.981 MHz spur when TX at 24.895. What behavior do you see at that frequency? Other differences are the 100 Ohm resistor, and the reconstruction LPF in place.

With my build of the LPF (100p, 330nh,200p,390nH,200p,330nH,100p) using AIML inductors, I see losses in the filter from about 1 dB on 40 M to 4 dB on 10 M. It doesn't appear that others see this. Is that true? There may be a problem with my build.

73,

Steve

KF7O

[Glenn P]

Not seeing losses like that Steve.

I will have a look closer in at the '150MHz' spur to see what it is asap.

Off the subject, but:
MY H-L is in pieces right now. I decided to tidy it up a little by mounting all the boards on top of a blank piece of PCB material. Mounted on metal spacers.  Found it no longer works !   The problem or cause is that if the  LAN adapter board is grounded through metal spacer also, then nothing works. Power drain is only 50mA...... pulled out the spacers on LAN and its working again.   Have to find some plastic ones.  (the return path is from RF board mount screws, ground track is very close to mounting screw heads and shorts it seems.

glenn
vk3pe

[Glenn P]

and 24.895 also.........

[Glenn P]

OK Steve, the spur is not 2x 73.728. It appears to be the 3rd harmonic of the LAN interface xtal (50MHz)   Touching around the xtal brings the level up and down.

I did a test at 24.895MHz and can see what appears to be your alias of 22.981MHz although the marker is a little out in reading freq. (but centre marker is ok so not sure) Dropping drive to reduce the wanted signal by 10dB also drops the '22.891' a like amount, or very close, 11dB

The closer in spurs appeared also at the 29MHz test and are approx +- 1MHz from wanted signal.

RE the reconstruction filter, i recall that doing an analysis of it using RFSim99 shows that this filter does have  a fair degree of loss approaching 30MHz.  Tolerances may affect this even more. My build uses a 390nH supplied by John and 330nH WW from my collection.

You may note that I actually get more output (+11dBm at 24.895MHz) than what I was getting in other tests at 29MHz which would confirm a roll-off higher up..

On Monday, December 14, 2015 at 12:57:06 PM UTC+11, Steve Haynal wrote:

Hi Glen,

Thanks for the ;measurements and reports. They are very informative. 

Do you think your ~150 MHz spur could be 73.728*2=147.456 MHz? In the full-duplex version, this clock is generated on the interface of the FPGA and may be radiating some. You could try shielding the Hermes-Lite. You could also try the half-duplex firmware as that runs on v 1.22 too.

It is interesting that you see a reduction in the 2.4 MHz spur for power reduction. My experiment where I did not see reduction in this aliased back DAC spurs was for the 22.981 MHz spur when TX at 24.895. What behavior do you see at that frequency? Other differences are the 100 Ohm resistor, and the reconstruction LPF in place.

With my build of the LPF (100p, 330nh,200p,390nH,200p,330nH,100p) using AIML inductors, I see losses in the filter from about 1 dB on 40 M to 4 dB on 10 M. It doesn't appear that others see this. Is that true? There may be a problem with my build.

73,

Steve

KF7O

snipped:

[Glenn P]

forgot the picture........

[John Williams]

That is a AIML-1206 part...

[Steve Haynal]

Hi Glen,

Thanks for the measurements. Interesting that the 150 MHz is from the LAN interface. Are you using the waveshare?

I also found the 22.981 MHz to be consistently lower than the fundamental by 50 some dB. The spurious free dynamic range (SFDR) did not improve for this one when lowering the fundamental. Reducing the range of codes (amplitude of the input audio) does improve the SFDR. I thought the ~2 MHz spur when TX at 29MHz would behave in similar fashion as it is a 5th harmonic aliased back, but apparently it does not.

On Sunday, December 13, 2015 at 10:53:09 PM UTC-8, Glenn P wrote:

OK Steve, the spur is not 2x 73.728. It appears to be the 3rd harmonic of the LAN interface xtal (50MHz)   Touching around the xtal brings the level up and down.

I did a test at 24.895MHz and can see what appears to be your alias of 22.981MHz although the marker is a little out in reading freq. (but centre marker is ok so not sure) Dropping drive to reduce the wanted signal by 10dB also drops the '22.891' a like amount, or very close, 11dB

The closer in spurs appeared also at the 29MHz test and are approx +- 1MHz from wanted signal.

RE the reconstruction filter, i recall that doing an analysis of it using RFSim99 shows that this filter does have  a fair degree of loss approaching 30MHz.  Tolerances may affect this even more. My build uses a 390nH supplied by John and 330nH WW from my collection.

You may note that I actually get more output (+11dBm at 24.895MHz) than what I was getting in other tests at 29MHz which would confirm a roll-off higher up..

https://mail.google.com/mail/u/0/#inbox

[Glenn P]

Hi Steve, excuse my ignorance but what is waveshare ? Its definately coming from the LAN interface though.  Possibly when the unit is better shielded things will improve but the LPF's should help to knock it down anyway.

I have no explanation for the ~2MHz alias not following the wanted signal.  

glenn

[in3otd]

Hello,
I have measured the output level of some harmonics and spurs/images when varying the fundamental output level. I have used the Quisk routines for the "spot level" control to change the TX output from 0 to 1000, with the TX frequency fixed at 29 MHz. Level were recorded without the TX output filter, i.e. just after the T1 transformer on the V1.2 frontend.
With the spot level at 1000 I get around 13 dBm, while for having 10 dBm the level to use is around 665 here. As known, the output spectrum degrades rapidly when going towards the maximum output power.

Graphs and the full data file are also on my website.

I'll try to do the same graphs for the TxDAC output.

73 de Claudio, IN3OTD / DK1CG

On Tuesday, December 15, 2015 at 7:51:45 AM UTC+1, Steve Haynal wrote:

Hi Glen,

[James Ahlstrom]

Hello Claudio,

As you know, the Quisk "Spot" control changes the level of samples sent to the DAC.  So if you reduce the level by 6 dB, you have reduced the sample level by a factor of two, and so you have changed from 12-bit samples to 11-bit samples.  The loss of sample depth is undesirable.  It is better to adjust power by using the analog Tx level controls when actually transmitting.  It is OK to change the digital level for test purposes.

I look forward to your tests of the TxDac output.  My tests showed a much cleaner spectrum.  Remember that to test the TxDac at full output, you need to change the firmware.  You need to turn off the IAMP, and set the TxDac to full output.  Change register 0x0A to the default, and 0x0E to 0x01.  I got 5.5 dBm output from the TxDac with no back termination.

Jim

N2ADR

[in3otd]

Thanks James,
yes, if I understood the code correctly, the spot level changes the DC level at baseband, right?
The TX power level was fixed at 255 but I can try to run a test varying it; I guess this control has a more limited range looking at the AD9866 datasheet (did not check what the verilog code does)

For the TxDAC testing, if I understood correctly, I need to change the AD9866 init code (the part done in MyHDL); last time I looked at that it was not so clear to me how it worked...

73 de Claudio, IN3OTD / DK1CG

[Sid Boyce]

Any change in the following recommendations before I proceed further
with the build.

From 2 emails I have printed off.
1. 12/8/15 -- Steve says T2 is 8T (no centre tap) with 100 ohm across
the winding --> CN1 and 14T --> filter.

2. The frontend Filter can be bypassed with T1 secondary wired directly
to P1 as the PA provides the filter - John 11/05/15.

Already did the Regulator Bias voltage mod in Pre-V2-Amplifier-Errata.md.
73 ... Sid.

--
Sid Boyce ... Hamradio License G3VBV, Licensed Private Pilot
Emeritus IBM/Amdahl Mainframes and Sun/Fujitsu Servers Tech Support
Senior Staff Specialist, Cricket Coach
Microsoft Windows Free Zone - Linux used for all Computing Tasks

[John Williams]

Can't answer now, away from computer, but errata article is correct.

--
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.

[James Ahlstrom]

Hello Claudio,

yes, if I understood the code correctly, the spot level changes the DC level at baseband, right?

Yes, the Spot feature sends a 16-bit DC level 0 to max.  So that should result in a 0 to 12-bit max DC level for transmit.

The TX power level was fixed at 255 but I can try to run a test varying it; I guess this control has a more limited range looking at the AD9866 datasheet

The variation is limited.  I think it is 7.5 dB.

For the TxDAC testing, if I understood correctly, I need to change the AD9866 init code (the part done in MyHDL); last time I looked at that it was not so clear to me how it worked...

You need to change myhdl/ad9866.py, and then use MyHdl to convert that to Verilog. 

Jim

N2ADR

[Steve Haynal]

Hi Glenn,

The waveshare is the ethernet adapter required for the BeMicro CV. Are you using a BeMicro CV, SDK or CVA9?

73,

Steve

KF7O

[Glenn P]

Thanks John, I realised that later on.....

[Steve Haynal]

Hi Claudio,

Thanks for the measurements. I always appreciate your clear graphs. I too would be interested in the differences when adjusting TX power versus DAC amplitude. The TX power range should be ~19 dB. If you alter ad9866.py, you have to "compile" and generate Verilog that Quartus consumes. You can also hack the verilog file ad9866.v.

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

I meant to include that I am interested in the direct TX DAC tests too. The ~19 dB range is with the IAMP, ~7.5 dB range with direct TX DAC output. 

73,

Steve

KF7O

[Glenn P]

Using a CV8 plus waveshare LAN

glenn
vk3pe

[in3otd]

Hello Steve,
I have briefly played with the TX power control and the range is about 19 dB as expected, but the last codes give all the same (max) ouptut power and not the expected 0.5 dB steps, so there is a bigger step toward the end. I tried to understand how the power is controlled, but it's not completely clear to me. AFAIU you are changing the TxPGA code and the IAMP current ("primary path gain"). Shouldn't be enough to change the TxPGA code? I guess the steps which are actually constant are these. I tried changing them but I got 10 dB less power on all steps, which does not make sense to me.

Besides, I saw that the AD9866 datasheet briefly mentions the IOFF internal current sources that can be used for a "marginal improvement in distortion performance under large signal conditions", without further details and I wanted to check whether this can help for the spurs we see. I hacked ad9866.v to include the previously unused step #17 and change the IOFF1 value using the register at 0x12, tried to set it to 2 and 4 but in all cases I get a corrupted, noise-like, output. Did you already try to change the IOFF1 value? Do you have further details on how that should be used?

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello,
by popular demand (hi), here are the fundamental, harmonics spurs, etc, vs the TX drive level (C1 byte for C0=0b00010011).

TX frequency was 29 MHz, spot level at max (1000). As before, the measurements were done without the TX output filter, i.e. just after the T1 transformer on the V1.2 frontend.

First graph is the fundamental alone, just to show the bigger step at the end (note the small downward slope there, initially I thought it was a thermal drift, but apparently it's not)

and here are the graphs for the other relevant signals

It seems that the bigger increments in the distortion products occur when changing the IAMP gain (but the 2nd harmonic is somewhat an exception). The true images follow mostly the fundamental output, as they should. Note that the fixed spur at 2*fs behaves like a distortion product (?).

73 de Claudio, IN3OTD / DK1CG

On Friday, December 18, 2015 at 7:55:43 AM UTC+1, Steve Haynal wrote:

Hi Claudio,

Thanks for the measurements. I always appreciate your clear graphs. I too would be interested in the differences when adjusting TX power versus DAC amplitude. The TX power range should be ~19 dB. If you alter ad9866.py, you have to "compile" and generate Verilog that Quartus consumes. You can also hack the verilog file ad9866.v.

73,

Steve

KF7O

[in3otd]

...sorry for the repeated postings,
I thought it was interesting to show how the graphs look like when plotted in dBc, i.e. level of the harmonics/products/spurs w.r.t the fundamental level.
It turns out that the TxDAC is actually cleaner when at maximum gain (in general) (while increasing the IAMP gain causes a increase in the distortion, as expected):

...a somewhat strange behavior...

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Claudio,

Thank you for these detailed measurements.  I would like to compare them with my single measurement of the 2Fs - 5Ftx spur for Fs = 73.728 MHz, Ftx = 24.895 MHz, spur at 22.981 MHz.  This is a very troublesome spur.

For Spot = 1.000, no back termination and a 1:1 transformer, I measured 16.45 dBm for the fundamental and -38.8 dBm for the spur.  That is -55.25 dBc.  For 29 MHz, your spur is at 2.456 MHz.  Your numbers at 29 MHz and using the blue graph for 2Fs - 5Ftx are 13.3 dBm for the fundamental and -49 dBm for the spur, or 62.3 dBc.

I measured the output of the TxDAC alone at maximum output power and got 5.6 dBm for the fundamental and -71 dBm for the spur, or -76.6 dBc.  From your graphs at a Tx level of 110, the fundamental is 0 dBm and the spur -70 dBm, or - 70 dBc.

My measurements are not directly comparable to yours (not to mention much less extensive) but they point to the same result.  We can drastically reduce spurs by using the output of the TxDAC without the IAMP, and including an amplifier on the front end board to get back to 12 dBm output.

Jim

N2ADR

[in3otd]

Hello Jim,
I have now done some measurements on the TxDAC output, with the IAMP disabled, using a center-tapped 1:1 transformer, as described in your post and in the AD9866 datasheet. The TX frequency was 28.895 MHz this time. Indeed, the output of the TxDAC alone is much cleaner:

Strange that the IAMP has a strong effect on the 2fs-5ftx type of spurs, as it doesn't have a very strong component at fs at its input; I thought it might come from the TxDAC when loaded by the IAMP, so I tried to do some experiments loading the IOUTP output in various ways when using the IAMP, but did not manage to improve the distortion.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Claudio,

You are getting into some of the sordid history of the Hermes-Lite firmware! The original firmware had a constant 20 dBm output level. When I added support for adjusting the power level, I did so without the remapping of gain codes that you refer to in the RTL. But people complained that we no longer saw the near 20 dBm output. There is a comment in the AD9866 datasheet saying that under certain settings the gain LUTS are corrupted. (Search for LUT.) I concluded there was a bug in the AD9866 firmware as whenever we attempted to adjust the TX gain, we could never go back to the full 20 dBm output. That is when I wrote the remapping you are referring to. The only way I could get near 20 dBm output was to use the nonintuitive codes you see. This really is a hack and maybe there is a better way, or just to live without being able to achieve full output once that gain settings are changed. 

I did try the recommendation in the datasheet for marginal improvement in spurs. I saw no significant improvement. There is a post somewhere in this list with more details.

73,

Steve

KF7O

[Steve Haynal]

Hi Claudio,

Thanks for all the measurements! I'm away now, but hope to take a close look at them soon.

73,

Steve

KF7O

[James Ahlstrom]

Hello Claudio,

Great work on measuring the TxDAC distortion!

By Fs I think you mean 73.728 MHz.  But, as you know, the AD9866 has a 2X  iterpolator, so the actual Fs for the TxDAC is 147.456 MHz.  The 2Fs - 5 Ftx spurs are exactly in the form of an IMD product, and I always guessed they resulted from the IAMP due to the lack of a reconstruction filter between the TxDAC and the IAMP.

Jim

N2ADR

[in3otd]

Hello Jim,
yes, I meant fs=73.728MHz and as you said the TxDAC is a sampled system which works at 2*fs so I can understand that its circuitry can generate harmonics of the transmit signal ftx which are then aliased around 2*fs.
But looking at the datasheet the IAMP seems a "normal" amplifier (continuous time, not sampled) so I will expect it to amplify (and generate intermodulation products) of its input signals. Looking at the measurements the carrier at 2*fs at the output of the TxDAC is relatively low (-50 to -60 dBm) and the tx fifth harmonic is between -70 and -80 dBm. If the 2*fs-5*ftx product at the output of the IAMP was due to intermodulation in the IAMP I would have expected it to be much smaller then the input components. To have aliases of the harmonics the IAMP generates it should also be a sampled system, which seems not be the case, at least looking at the datasheet.

To expand a bit more on the experiments I mentioned in the previous post, I noticed that when the TxDAC+IAMP are used the common mode of the TxDAC output (on the IOUTP pins) is around 1.3 V. But when using the TxDAC alone, the datasheet recommends to connect the transformer center tap to ground, so the common mode is now 0V. The measurements above are done it this condition.
I have then lifted the transformer center tap from ground, bypassed with a good 1 uF capacitor and changed the voltage on the tap, effectively changing the common mode of the TXDAC output. I saw that until 1.2/1.3 V the bias current does not change, so that seems to be the allowed common-mode range. But I saw that already when using a 1.0 V bias on the center tap the products at the TxDAC output degrade quite a lot, some even by 20 dB. I have the measurements data somewhere.
So it seems that also the bias on the TxDAC output can have a significant impact. But, as known, unfortunately the TxDAC and IAMP are DC coupled so we cannot do anything on this.
Strange is that the datasheet says "the TxDAC voltage compliance is around 1.2 V" but then it is biased to 1.3V when using the IAMP. Maybe the circuit is much more complex than shown in the datasheet, so the allowed common mode changes when the IAMP is used.

Another experiment that I wanted to try is to put a series trap at 2*fs between the IOUTP_+/- pins; honestly  do not expect it to do much. If this is confirmed the products are likely already at the TxDAC output and due to the different biasing conditions when using the whole chain.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Claudio,

Yes, it seems you are right on this; the distortion is in the TxDAC.  Looking at the data sheet on page 30 we see the recommendation to "Limiting the peak positive V IOUTP+ and V IOUTP − to 0.8 V" and the peak-to-peak to 1.2 volts.  If not too much trouble, I would be very interested to know if the TxDAC output was clean with 0.4 volts common mode, and a signal of 0.4 volts peak.  That is, the voltage on each TxDAC pin of 0.4 volts +/- 0.4 volts.  This data is needed if we use the TxDAC output directly without the IAMP.

Jim

N2ADR

[in3otd]

Hello James,
will try to do this measurement in the next days, currently I had to put aside my H-L to make room on the bench for servicing a much bigger non solid-state amp, hi.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Ahhh.  Hollow state technology.  I remember it well.

Jim

[Joe]

Jim,

I've been running some test based on previous suggestions I'm using PowerSDR

as the radio program with a BeMicro SDK and a V1.21 main board. 

I added a 1:1 transformer connected to CN3 I didn't remove the original ouput

transformer connected to the IAMP.

First the good if the SDR drive is kept below 75 the DAC output is good on 12meters 

to -70db if the drive is increased above this level then the spur appears. BTW the 10meter 

spur is also gone.

Now the problem, it would appear that both the gain of the DAC and IAMP are used

to control the main output.  The output of the DAC doesn't follow what I see on the main output which

goes up as the TX drive is increased but the DAC output rises and falls in several steps as the drive slider

is increased.

Joe    wa9cgz

[James Ahlstrom]

Joe,

Yes, the firmware must be modified to use the TxDAC.  The regular firmware will not work.

First you must turn off the IAMP.  Then you must modify the Tx power logic so it controls the TxDAC.  The adjustment range is zero to -7.5 dB.  My notes say to set register 0x0E to 0x01; set register 0x0A to the default (no write to this register).

Jim

N2ADR

[Steve Haynal]

I will release firmware supporting both TxDAC or IAMP as well as gain mapping adjustments for both via recompilation hopefully by Jan 4.

73,

Steve

KF7O

[Joe]

 Since I'm software challenged while waiting for Steve's new code I tried a interesting experiment. I had added some pin jacks to CN3 to patch in a balun for testing of direct output of the ADC and since it seems that the IAMP is causing the spurs on 12 and 10meters  I wondered if adding a series resonate thank circuit across pins 1&2 which is between the ADC and IAMP would have any effect on spur levels. The interesting thin is

that the spurs can be reduced fror -54db to almost 65db with careful tuning of the series tank circuit. I tried a number of different frequency

frequency series tanks but the one that seemed  to be the most effective ended being 234MHZ as measured on my VNA . Now keep in mind that CN3 has rather long leads back to the AD9866 and the only way to see what you are doing is with a spectrum analyzer.

While the above does reduce the spurs below -60 the better approach maybe eliminate the IAMP altogether.

Joe  wa9cgz     

[in3otd]

Thanks Joe,
interesting result with the series tank, maybe it's actually resonating at 2*fs, about 147 MHz, with the long PCB traces?

If you wish, I can recompile the current code for you, disconnecting the IAMP and recomputing the gain steps, let me know which BeMicro board you have there (SDK/CV/CV A9 ?).

73 de Claudio, IN3OTD / DK1CG

[Joe]

Thanks Claudio,

My test unit is a SDK non FD 73MHZ.

Joe

[in3otd]

Hello Jim,

here is the TxDAC output vs. common-mode bias (on the transformer center tap).

The TX frequency was 24.895 MHz, output power at max (around 5 dBm), which is more than the 0.4 V peak you mentioned, but should give some margin, hi.

As you can see, the 2*fs-5*ftx spur is fine until about 0.6/0.7 V bias.

a

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Joe,
enclosed is the current code compiled for the SDK (non-FD, 73.728 MHz clock), modified to disconnect the IAMP (and setting its gain to zero) and remapping the 0 to 255 TX level code to the sixteen 0.5 dB TxDAC levels (so the output actually changes every 16 TX level steps). I have compiled and tested the code for my CV A9 also and it works fine there, so there are good chances it works on your SDK as well, hi.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Thanks Claudio.  I plan to use 0.4 volts common mode, and +/- 0.4 volts max amplitude as advised by the data sheet.

Jim

N2ADR

[Joe]

Claudio,

 I installed the new code but it doesn't run correctly

the audio on RX has a very short interruption  about every 2 seconds and when in TX

output frequency is measured it's incorrect look at examples below.

TX          Measure   in MHZ

1850      2.22

7295      8.754

21000    25.2

24935    29.922

nothing on 10 meters but just to make sure I'm

using a SDK non FD at 73.xxx MHZ clock.

Joe 

[in3otd]

...oops, didn't notice that the frequency in the SDK code was set to 61.x MHz by default, while for the CV A9 is 73.728 MHz; now recompiled with the correct frequency...

73 de Claudio, IN3OTD / DK1CG

[Joe]

Claudio,

Firmware runs with no issues and 12Meters now shows the spur at

about -60db and 10 meters about -55db see iIN3OTD 12&10 of analyzer below.

also included of the same bands on old firmware with series resonate trap not a lot of difference

both 12 and 10 meters continue to be a issue however a TX filter would help on 10.

Anyway if both spurs can be reduced to -60dbm  it's probability good enough.  

[in3otd]

Thanks Joe,
not sure I'm reading the graphs correctly, I will expect even lower spurs from the TxDAC output. My measurements (and Jim's) show that when transmitting at 24.895 MHz the spur at 22.981 MHz is at around -72 dBm; the carrier is at around 5 dBm so this will be -77 dBc, your numbers seems to be much higher...

73 de Claudio, IN3OTD / DK1CG

[Joe]

 I agree it seems that I got better results before when the IAMP was still connected.

Just to make sure where on the same page I have a 1:1 balun with the center tap grounded.

I'm going to run the test over a will keep you informed.

Joe   wa9cgz

[in3otd]

yes, the TxDAC is connected to a 1:1 balun with center tap, as in Figure 63 of the AD9866 datasheet (Rs is not used). I have used a BN43-2402 core with 4 turns and 2+2 turns of 0.2 mm wire.

73 de Claudio, IN3OTD / DK1CG

[Joe]

Claudio,

The only difference is my balun which is 6 turns, the table below are my results I retested with both firmware

at my QTH it would appear that your firmware has a slight advantage only if the output is reduced to +5dbm at +10

I don't see much advantage over the original firmware with a series trap except your firmware version doesn't

require any test equipment  for adjustment.

RESULTS WITH NEW FIRMWARE

 

BAND         PWR OUT     SPUR       SPUR FREQUENCY

12 24.950  +4.8DBM       -70            22.6

                    10.9             -58

10 28.000  +5.0               -65           7.46

ORIGINAL FIRMWARE WITH TRAP

12             +19DBM        -68 worst case a second spur occurs at -70 

                 +10                -65 and -69 (going from +10 to +19 actually causes a reduction)

                 + 5                 -65

10            +19                 -68 worst case a second spur at -70           

                +10                 -66

                +5                   -66 

Both 12 and 10 meters show these issues. 

Everything below 12meters is -75 which is the limits of my spectrum analyzer

I didn't see much change from my previous readings what we need is someone else to check. 

Joe  wa9cgz

[in3otd]

Happy New Year to everyone!

Joe,
in the "new firmware" table below, there is a 10.9 (dBm) in the "PWR out" column; can you obtain that power from the TxDAC alone? In theory you can get only slightly more than 5 dBm when using the 1:1 center-tapped transformer and this is the power I and Jim measured.

In the "original firmware with trap" you have a +19 dBm output, but from the IAMP with the 2:1 transformer and 100 ohm of backtermination (R2 on the main board) you should get +14 dBm at maximum.

I think there are some differences in the actual circuits we are testing...

73 de Claudio, IN3OTD / DK1CG

[Joe]

Claudio,

You where correct on the new firmware the transformer wasn't 1:1 and now the results look excellent with max

power out of +5.2DBM as far as the +19 output with the old firmware it's because transformer was wound

according to the original instructions with no 100 ohm terminating resistor. But just to be on the  same page

I'm using a 43-2402 core with 4 turns center tap grounded to 4 turns toward the 50 ohm side.

I'm going out today but will post some analyzer pictures later.

Thanks Again  Joe