Pre V2 Rx BPF Filters

[ZL2APV]

In late June a small group of us started work on assembling and testing an Rx filter board. Initially it was just chit chat about ordering parts and waiting for them to arrive but more recently they have arrived and assembly has proceeded. Now that we have some of it built up and testing has begun the interchange has gone from friendly chatter to a technical discussion more appropriate to the list so I am scanning the old emails for the technical bits and will place them as a series of posts under this thread.

I would like to acknowledge John W9JSW who designed the board layout, did a BOM and procured a set of components for us, boxed them up and shipped them. This was a huge commitment in time and a big savings in postal costs to us in the antipodes.

Glenn VK3PE who is leading the construction and testing and has a lovely set of instruments for the job. Glenn has been preparing a set of the filter plots and will be adding them to this thread soon

Andrew G4XZL is overseeing the filter designs.

73, Graeme ZL2APV

[ZL2APV]

@Graeme: Have you tried the Rx filters on a receiver yet? I am very interested to know if it creates any noise and thought a good test would be to connect any receiver (your PICaSTAR?) to a dummy load and note the noise level then connect it via the Rx filter to the load and see if the noise floor rose at all. Naturally the Rx filter board should be powered and it probably would be a good idea to run through all the bands plus the through position to see if their is any noise floor change with any particular position.

@Glenn: Loss seems to be  <1dB with the switches in circuit compared to a direct sweep of the filters only. Isolation around 80dB. These numbers are about right according to the Data sheet.

I see nothing of the 25KHz internal to the switch devices.

[ZL2APV]

I meant to mention that the "Special" library in Kicad does not have anything in it and is quite safe to remove from the project to avoid the "The following libraries were not found: special" message each time you open the schematic. Just the usual Preferences/component libraries and scroll down the component libraries to special, select and choose "remove"

73, Graeme

On Thursday, July 14, 2016 at 4:10:59 PM UTC+12, ZL2APV wrote:

[ZL2APV]

We had a discussion along the lines of a Butterworth response for the mesh bandpass filters may be easier to reproduce for those without a big set of instruments. The Chebyshev response although falling off more sharply seemed to be much more critical of component tolerance...

Here is another Butterworth design which again would be suitable for Tx or Rx although not very good attenuation for 10.1MHz second harmonic. This filter was plotted with a Q of 40 to match the Bournes inductors but it improves with T37-6 toroids and calculates to 19 turns for the 1.1 uH. Easy to get capacitor values are a good advantage for this filter also.

It would be good in the future to put this filter onto an operational amplifier and measure the harmonic output of the complete unit as the 2nd harmonic will be somewhat down on the fundamental and the total of this and the filter attenuation may give a perfectly acceptable output as it stands. The advantage of using the one filter for both Rx and Tx is high and I recall from using mesh filters previously they were not as critical in terms of getting the inductance dead right and it was more a case of making the 3 inductors physically identical. I will wind up a toroid and measure its inductance at 12 MHz (midband) and make sure that 19 turns as calculated from formula is in fact correct.

Graeme

[ZL2APV]

Graeme,

I bread boarded your 30-20M filter this morning using 19turns at about 80-90% coverage on T37-6 cores.

I also built the same filter using AADE measured inductors which had about 2-3 turns less. As you pointed out, the AADE measures at a very low frequency which is not condusive to accuracy when using other than air cored inductors. This was proven out in the AADE filter, as it was centred rather high.

The 19turns resulted in  very close BW and attenuation to your Elsie graph.

Glenn

vk3pe

[ZL2APV]

OK Glenn, I have just finished winding up the toroids and building the filter. I have attached the plots for your info and will do something about writing this up for the list later. I would prefer that you present your own results as you did all the work and deserve the credit but I will pave the way for your info.

The plots are interesting in that I did not get the same difference in the passband as you did although it is still better at the 10 MHz end than the 14 MHz end. The return loss came out good with worst case VSWR being around than 1.1:1 so it would be great in Tx service although the 20.2 MHz 2nd harmonic of 10.1 MHz is only down 25dB but the 28 MHz point is nearly -50dB

[ZL2APV]

We had a discussion on the effects of capacitor Q on a filter and techniques for measuring inductor values. Glenn and I both have an AADE style LC meter and we were aware of the test frequency used and how that may affect the inductance seen at the operating frequency of the filter. We looked at alternative methods of measuring L and in particular concentrated on the method described by Jacques Audet VE2AZX here and the spreadsheet here which give very good results. The issue of capacitor and inductor Q is well addressed by Wes Hayward here where he gives some eye opening information on chip capacitors.

73, Graeme ZL2APV

[Glenn P]

Here is a summary of the BPF's I tested, per Johns PCB design. Thanks John, for the PCB's and supply of parts to keep postal costs down.

The conclusions drawn are my own and open to discussion of course.

 See attached .DOC file which presents sweeps of all the filters and some variations which occurred, due to measurement methods of the hand wound toroids used in a couple of the filters.

glenn
vk3pe

[ZL2APV]

To examine the need for highly accurate toroids in Butterworth mesh filters I altered the toroids from 19 turns as used in the previous test to 18 turns each. They were firstly bunched up to cover 60% of the core and then spread to cover 80% of the core. The attached screenshots are of the 18 turn 80% core coverage and as can be seen there is very little difference in the performance of the filters suggesting that hand wound toroids in this kind of filter are practical. The biggest deviation I got was when I mixed one 18 turn toroid with the other two 19 turn ones giving rise to it being important to keep the toroids the same value i.e. wind all three exactly the same.

Files showing the filter response and return loss plus the physical layout are attached. Looking at the 60 MHz end of the filter response it can be seen that there are peaks in the response although still around the -50 dB mark. Pondering on this I have concluded that it might be series inductance in the shunt capacitors which in this case are leaded polystyrene. I think that the series capacitors will have their inductance largely swallowed up by the series inductors. I expect that if the shunt capacitors were chip caps with minimal lead length the rises in response would shift up to VHF regions.

This table presents through loss readings of 19, bunched 18 and spread 18 turn toroids

MHz      10.109 14.029 14.370 20.210 28.020
19 turns -0.29  -0.57  -0.80  -25.70 -47.54
18t 60%  -0.30  -0.52  -0.67  -23.97 -46.07
18t 80%  -0.35  -0.51  -0.60  -22.88 -44.50  Lowest insertion loss was -0.32 dB at 11 MHz

Conclusions: The Butterworth mesh filter is much less critical of component variations than Chebyshev or optimized mesh filters but at the expense of out of band performance. Readings obtained very closely match Elsie predictions and filters can be designed using these parameters for other bands with confidence. By being prepared to alter the passband width it is likely that a filter can be designed which fits standard 5% components. The SWR response of the filter is very good and suitable for both transmit and receive and could be used for this dual function but for the 20.2 MHz response which is only -23 dBc and assuming a second harmonic of -15 to -20 dBc from the final there would be an unwanted emission of -40 to -45 dBc which does not meet some countries regulations. The situation could change with pre-distortion or maybe a push pull amplifier but for our current system it looks like separate Tx and Rx filters will be necessary but I would like to be wrong here.

[Glenn P]

Perhaps, adding one more turn would help with the 2 x 30M rejection and not affect 20M too much, Graeme? If LPF ability was needed for PA.

glenn

[Graeme Jury]

Yes you are right about the rejection Glenn but I think the passband loss would go over a dB at 20 metres and remember these are toroids with a measured Q of 158. When we go to the smd inductors we will be looking at a Q of 40 or so with attendant pass band losses. The real issue is that the passband is around 33% of the centre frequency and it is a real big ask for a filter to have good skirt attenuation with that percentage. Taking the 17/15 band with a passband percentage of around 10% it is quite easy to achieve a filter skirt that keeps the 17 metre 2nd harmonics at legal levels. Its gone midnight here so I will post the design of the 17/15 here tomorrow so you can see what I mean. If we were seriously building filters for common Rx/Tx use we would need to make a separate 30 metre and 20 metre filter. Another factor with a common bandpass filter is that we can't do out of band transmissions due to the filters - but then why would we want to do that? We still have the instrument output for signal generator etc. use.

73, Graeme

--
You received this message because you are subscribed to a topic in the Google Groups "Hermes-Lite" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/hermes-lite/NULHi5nD9vk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[ZL2APV]

Here is the design for a 17/15 mesh filter based on a slightly modified Butterworth design to arrive at standard 5% components. As can be seen, in contrast to the 30/20M filter the second harmonics are well down on the skirts and assuming -15 dBc for the linear amp the 2nd harmonic will be down at least to -55 dBc. Again the Butterworth allows for stable non critical components.

Oh yes my remark on why would we want to transmit 5 watts out of band. How about driving a varactor multiplier?

73, Graeme

To unsubscribe from this group and all its topics, send an email to hermes-lite+unsubscribe@googlegroups.com.

[ZL2APV]

Just looking over the interconnections between the pre-V2 5 watt PA v1.1 to pre-V2 Rx filter board and v1.4 frontend board. I see the PA board places 12 volts on the common (pin 10) of the header connectors.
This suggests that the band pins (1, 3, 5, 7, 9, 11) should have pullups to 12 volts.
The CD4504 has pin 13 connected to Vcc which puts it into TTL mode but with 12 volt switching signals it should really be in CMOS mode to level translate 12 volts to 3 volts.
Alternatively 5 volts could be picked up from somewhere and applied to pin 10 of the header on the PA board although 5V is not readily available there.
Another possibility is to pick up 5V from the edge connector pins 13 & 14 on the v1.42 frontend board and strap this to pin 10 on the RELAYCNT 14 pin header. The connection to 12 volts would need to be broken on the PA board as well.

Ideally the pull-ups should mount on the V1.42 frontend board so that anything connected sees the logic high if the line is not activated. They could go on the PA and Rx BPF boards and would need to be on both in the event of one or the other boards being unplugged. Another possibility is to replace the ULN2803 with a line driver chip with totem output so there is a logic high or logic low depending on the switching state.

Glenn, what software did you use to draw up the really neat pre V2 interwiring diagram? Was it with Kicad?

73, Graeme zl2apv

[ZL2APV]

I hate having to reply to my own posts particularly when I am correcting something I have carelessly done wrong. If I had thought it through I would have realised that the pullups are provided via the relay coils on the PA board so we are set with the 12 volt option and all 5 volt references are irrelevant as is using a line driver.

What is still relevant is that the CD4504 should not be operating in TTL mode with 12 volts being applied to its TTL input and instead should be operating in CMOS mode where it can tolerate up to 15 volt CMOS levels.

Sorry about the noise, Graeme

[Glenn P]

Hi Graeme, I used Protel for the drawings but imagine Kicad could do the same thing. I may have used PCB to do it though, can't recall now. (tracks and overlay text) I have Kicad also but never really used it in anger. ( or does it make you angry?) One thing i dislike about Kicad here at least, is the very fine lines and colour that it uses in Schematic mode. These don't print very well with my Laser printer.

Would you like me to build your new filter also to confirm it?  I already have the 1.1uH pre-wound  toroids as you know.  I don't have 200p or 75pF caps here in my range but have to parallel some up, luckily easy values. Will be Sunday probably, as going to a small hamfest tomorrow.

glenn
vk3pe

[ZL2APV]

Thanks Glenn, and yes as I am familiar with Kicad I will use it for that sort of thing although it does have some aspects that are frustrating but after I figured them out they became strengths. The drawing package in Libre Office might be good too as would be Libre Cad.

I will take you up on your kind offer to build up the filter as I am going to be tied up for the next couple of days and would like to see a real life plot. It is pretty convenient that I managed to get 1.1uH inductors for both the 30/20 and the 17/15 filters and makes parts easier. 2x100 and 47//27 should be OK.

Enjoy your hamfest and I hope that you manage to score some goodies.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme and Glenn,

Thanks for sharing all the details of your investigations. I've read all the posts and am digesting them.

73,

Steve

KF7O

[Steve Haynal]

Hi Glenn and Graeme,

Thanks for the report. It is nice to know that the Murata and Bourns inductors can at least do a passible job. You both are deep into the details, but I have a few higher level questions:

• For the RX BPFs, what are the pass bands you are designing for? If these filters are only used for RX (I know you are considering TX too), what sort of pass band is necessary to improve the DR2/IP2 numbers in Adam's test report? My thinking is that you want to suppress any combination of two strong signals which might cause interference in the band of interest. This works at to a little less than an octave. For example, a pass band of 5 to 9 MHz would suppress and keep out the distortion at 5+5 MHz or lower components as well as 9-5 and higher components. If this is true, I'd imagine you'd want all filters to be almost an octave wide to support SWL, etc., and you could get by with one filter for 17/15/2/10M. 5 filters should do it. 
• In general, I don't understand why there are two filters for 17M and 15M in the current TX filter and RX BPF designs. I've been looking at the RS-HFIQ recently, and this design gets by with simpler TX filters. Note that the signal path after the PA only goes through the TX filters (the BPF filters are before the PA) and the PA is similar to John's. Are we suffering from stray capacitance in the relays? Are the RS-HFIQ stop band results not that good? What am I missing?
• What source are you using for your toroids? Are you seeing large variations due to different manufacturers/sources?

73,

Steve

KF7O

[Glenn P]

Graeme, this will stagger you..... I plotted the Mesh Filter as per your Values, measured Caps with AADE meter to better than 2%, inductors measured at about 25MHz using series resonant method and SA-TG.
Compare with your numbers/plot.  I didn't measure return loss but imagine it will accord with your plot also.

Glenn
vk3pe

On Friday, July 15, 2016 at 10:55:29 AM UTC+10, Graeme Jury wrote:

Here is the design for a 17/15 mesh filter based on a slightly modified Butterworth design to arrive at standard 5% components. As can be seen, in contrast to the 30/20M filter the second harmonics are well down on the skirts and assuming -15 dBc for the linear amp the 2nd harmonic will be down at least to -55 dBc. Again the Butterworth allows for stable non critical components.

73, Graeme

[Graeme Jury]

Crikey Glenn, that was a worthwhile exercise! Looks like there is a lot to be gained from the less critical Butterworth designs and this would be perfectly suitable for both Tx and Rx use. I am going to comment on Steve's post as soon as I finish here and will give my thoughts in the use of this filter for Tx purposes. I have done a 60/40 design which I will test also. We might be in for a change of values.

I operate a HiQSDR as my main rig with a 60 MHz Lo pass roofing filter and a 1.7 MHz Hi pass broadcast band rejection filter because I get 2 volts of RF from the broadcast stations on 80 to 108 MHz FM and the .5 to 1.6 MHz bands. I still get harmonics and some intermod plus a small rise in the noise floor of the radio even with the filters and it is terrible without them. This station description is to give a background to a subjective listening test I made with the  30/20 band filter. I removed the roofing and floor filters and replaced them with the 30/20 filter and operated the radio on 20 Metres. The noise floor was a tiny bit better and just about all of the little spurs had gone. I worked a few W stations including a W1 which is hard to get from ZL and a TA station and it just felt more alive (not very scientific). I then had a listen on 15M which was nearly dead without the filter and absolutely nothing with the filter. I then moved to 30M band and worked a W1, W5, and W7 station with the spectrum looking flat except where genuine signals existed. Moving down to 40M where the VK,s should have been S9+ nothing was better than S4 and 80M was dead. Going down to the broadcast frequency (1.359MHz) where I get my 2 volt signal it was just rising out of the noise and listening to it I could hear the receiver noise in the signal so that was a huge attenuation. This is a subjective report but run under actual operating conditions and convinced me that it is worthwhile running Rx filters. I previously didn't want them as I felt it restricted wide band operation of the SDR.

73, Graeme zl2apv

[Graeme Jury]

Hi Steve,

Glenn and I have had a discussion on how there is a lack of specification and documentation on some of the project. We are mindful of the huge amount of work that John has put into the RF section of the project and the burden of getting something together that can be tested and tweaked and are happy to pick up on that phase for him.  Extending out to your point that ~.9 octave filters should do the job, I agree that if the filter design difficulties can be overcome this should do the job. One of the problems is that there was never a specification saying that this should be pursued and John has taken the approach of a Hamband only filters radio with a through link for all out of band operations so the radio will be optimised for Ham service but still general coverage, contrasted against a quasi octave continuous coverage rig so we need to get that sorted before continuing.

As part of getting a specification together we needed to know what worked and what did not hence the further experiments on filters that were already thought to be settled. You will see form Glen's plot of the filter he constructed that 17/15 should be a single filter and John's board can already accommodate that so yes your suggestion that 17/15 be a single filter is already in hand.

My feeling is that the Rx filters should be designed for good SWR i.e. <1.3 and placed in the input to the frontend board with a Peregrine switch changing the Tx and Rx paths as is already there and the other end of the filter has another Peregrine switch to go to either the Tx input or the antenna changeover relay for a loss penalty of 0.6 dB. This may not be an option if the experiments by Jim  where we stay with a differential signal all the way to a push pull final proves to be the best way to go (and I feel it will) Due to the higher skirt requirements of the Tx with a single ended final it will need to have its own filter set. I do not know yet if a push pull amplifier will remove that need.

I got my toroids from Toroid King http://www.kitsandparts.com/ and their Al seems to be a little better than quoted but close enough to use the on site calculator to get the turns. We are designing all filters for standard inductors and are getting good results form the Bourns and Murata ones. The toroids are better but not as much as I expected.

At present we are designing for the system as it is now i.e. an SDK or equivalent with a frontend board and an Rx filter board and a 5 watt single ended Tx board. I am constantly reminding myself that my enthusiasm is taking me out of scope and everything learned here should be applied to the MKxx version, not changing this one too far.

73, Graeme zl2apv

[Steve Haynal]

Hi Graeme,

I apologize for not always setting well defined goals for the HL2. This is still very much a project I do for fun and see myself as just a guy in his garage building an SDR. Because of this mindset, I often wonder about various sometimes exotic and tangential options, and do not hold myself to tight engineering schedules and practices, etc., that I would at my day job. This may frustrate others in this group. I do value all the interest and help from the community, especially in the analog/RF/filter domains where I am no expert, and plan to learn, borrow and adapt what you do. So please keep it up!

Regarding the BPFs, I personally would like to see them as wide as possible as long as the IMD2 numbers are improved. I would like to support SWL and multiple band WSPR/JT skimming, at least on the higher frequency bands. I also want to keep the cost down. I am open and interested to various BPF plans, so please let me know what you think is best.

73,

Steve

KF7O

[Steve Haynal]

Hi Glenn and Graeme,

I am happy but surprised that you saw no internal noise from the PE4259s. I would expect you would see noise around 900 kHz and odd harmonics similar to what I saw in this post by me on May 1 to the thread "PE425x switches noise at 900 kHz and harmonics (from the EMRFD group)."  The current RX BPF board uses 14 PE4259s, and I am curious and a bit worried as to how these 14 noise sources will add up. What noise do you see if you run an experiment similar to the one I did? Also, the PE4259 datasheet says performance degrades below 10MHz. Do you notice any differences or measure any distortion on 160M?

73,

Steve

KF7O

[Graeme Jury]

Hi Steve,

Don't apologise, we are all in this together and any one of us could see that there was no spec, loose or otherwise and done something about it. The plain fact in my case is that I am still uncertain about the best direction and am spending a fair bit of time trying things before I make firm suggestions to the group. What I am sure about is that John Williams has made some very nice PCB's which need building up and testing with possibly some tuning and tweaking changes. The end result will be a 5 watt Tx and a good Rx which will be certain to give a good performance with a BeMicro CV or BeMicro SDK. I promised John that I would be one of his testers and will see this job through.

Of course you are doing the V2 board and I am hoping that the experience that I gain from John's board will help me become useful in the design of whatever is finally attached to it while at the same time trying to ensure that John's development is still going to be one of the viable options.

In a skimming or any multiple band situation where part of the signal is in one filter and another in a different filter passband there will be issues. Even a 3 or 400 KHz span going across abutting filters will be a problem. The best way to reduce IMD2 is a super linear front end or if the signals are strong enough put attenuation in the front end as dropping the intermodulation signals drops the IMD products even more depending on their order. Choice 1 is to have an uncrunchable front end with a roofing and footer LP and HP filter but that is a big ask. Next best is as sharp a filter as possible but that introduces tracking difficulties and limits displayable spectrum. Then you have the flavours of band filters and the decisions on just how much bandwidth to utilise.

73, Graeme zl2apv

[Alan Hopper]

Graeme, Steve, list

I'm intrigued with the IMD2 issue, do you have an idea as to the major cause?  With the better filtering on the 1.4 frontend is there any value in turning off the internal rx filter?  Could dithering help?  I've just ordered a quality splitter as I want to try and optimise a soft version of the rf agc, my plan is to run two hermes lites from the same antenna and tweak one for maximum jt/wspr spots vs the reference. It  might be interesting to try the same thing with different frontend setups.

I've just made a 'coupled resonator' vertical dipole that is resonant on 20,17,15,12 & 10m for skimming and general use, I'm not sure how I'd be able to use bandpass filters with it, for this use separate high and lowpass filters are more attractive.  I guess the same board  built with just highpass filters would fit my use.

73 Alan 2E0NNB

On Sunday, July 17, 2016 at 10:31:07 AM UTC+1, Graeme Jury wrote:

Hi Steve,

Don't apologise, we are all in this together and any one of us could see that there was no spec, loose or otherwise and done something about it. The plain fact in my case is that I am still uncertain about the best direction and am spending a fair bit of time trying things beforeI I make firm suggestions to the group. What I am sure about is that John Williams has made some very nice PCB's which need building up and testing with possibly some tuning and tweaking changes. The end result will be a 5 watt Tx and a good Rx which will be certain to give a good performance with a BeMicro CV or BeMicro SDK. I promised John that I would be one of his testers and will see this job through.

[John Williams]

The Pre-V2 RX board has an extra filter bank that is spare. It could be populated with a 30MHz LPF for skimming use.  That would allow one to either select a band pass filter for normal band operation or the skimmer filter for wide spectrum use.

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.

[in3otd]

Hello Alan,
I'll try an explanation for the IMD origin, may not be entirely correct, so any comments/questions/corrections will be welcome.
I think that in a direct-sampling RX there are two main sources of non-linearity. One is the quantization done by the ADC; due to its transfer function, a series of steps around the decision thresholds, its "transfer function" contains many (in theory infinite) non-negligible power terms, which contribute to the appearance of IMD of all orders also at relatively low input levels. The related IMD2 (or 3) products level do not follow the usual power law, i.e. they do not increase as the 2nd (or 3rd) power of the input level, being the contribution of a large number of power terms, but remain relatively constant or appear to increase somewhat linearly with the input power. The exact behavior may also depend on the ADC INL or DNL or other details.
As the input power is increased, the classic "analog" non linearities are seen, where just one power term is largely responsible for the IMD2 (or 3), until the input level becomes so high that the other terms start to contribute (but this usually well above the normal operating region).

Ten year ago, SM5BSZ published an interesting article about the IMD in digital receivers and had shown that some smart dithering at the input of the ADC could be used to "randomize" the ADC IMD and effectively reduce the ADC generated products below the noise. This could be a great application for the full-duplex mode in the H-L, where during RX the TX path is used to inject some known (or out-of-band) dithering on the RX side.
(BTW, in the article SM5BSZ attributes most of the IMD to crosstalk with the digital output lines, I'm not sure it this always actually the case).

It may well be that in the actual usage, with lot of uncorrelated signals at the input, the IMD generated by the ADC are automatically "randomized" and spread out over a large bandwidth, making the receiver appear better than when testing with only two main signals at the input.

In any case the IMD can be of course be reduced by filtering out the unwanted input signals. Not easy to do for IMD3, since the unwanted signal is always close to the wanted one, but for the IMD2 the two signals are at least an octave apart so filtering is easier.
How much filtering is needed for the IMD2? As usually "it depends". If I'm not mistaken, For a classic analog amplifier, the second-order sum or difference product are proportional to the product of the input tones levels, so if the input filter attenuates by N dB tones which are an octave apart the corresponding product will be reduced by N dB and so the input IP2 will appear to improve by N dB. As said, for the ADC intermodulation the behavior is different and the intermodulation products may not appear to change much at low input levels.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

John,

I thought the new front end designs already effectively had a 30MHz lpf for rx(but I'm have trouble keeping track of all the options)? For my 20-10m skimming example I would like to aim for the highest possible performance, if that means a highpass filter to cut below 20m then I'd like that option.  From an rx imd perspective are the frequencies above the frequencies of interest as troubling as those below?

73 Alan 2E0NNB

[Alan Hopper]

Claudio,

thanks very much for this, you do wonders for my education. I have a reasonable understanding of the adc effects, I wondered if any other part of the input chain had any significant effect.  The topic of dithering came up often at Friedrichshafen and its value still seems open to debate. I wonder if our lower bit depth and sampling rate might affect the balance of this argument for hermes lite.  Out of band dither in duplex mode should be fairly simple to experiment with, subtractive dither where you try to remove the known dither signal is a bit more of a challenge as the 'known signal' will be mangled by the round trip.

Whilst on the subject I also wonder about dither on tx, I think Steve mentioned it at some point.  Could the the tx path be calculated at higher bit depth and then dithered. Could noise shaping dither be applied to move the noise beyond the lpf or to a frequency the antenna will poorly radiate.

73 Alan 2E0NNB

[John Williams]

The spare can be used for any purpose. I can add extra part locations to make it more versitile. One would need to propose a high-pass for me to use as an example. I do not know the answer to your other question.

[Graeme Jury]

Alan,

I am glad that I did not try to answer your IMD question as you would have got a classic analog misinformation. Thanks Claudio for the nutshell description and the valuable link. Back to the area where I am more comfortable, I would strongly recommend building a 30MHz LPF with enough power capability to handle your transmitter and placing it permanently in the aerial line external to the radio. Buy good quality mica capacitors and use air wound coils. A design straight from Elsie works well and I use a Cauer LP 7 stage filter which gives me a through loss of around -0.2 dB and an SWR of around 1.1:1. At 100 watts out from my Tx there is no noticeable change in power on my power meter on either side of the filter. This improves all preceding filters in the radio both on Rx and Tx and is one of the best things I have added to my station.

I think that when we have finished with the filters there will be one additional filter space due to 17/15 being combined and I would be more than happy to design and test a 20 metre HP filter for you. If you are using Quisk you will be able to switch on more than one filter at once so it would be possible to switch an installed series 30 MHz LP or a 2 MHz HP in or out of circuit while allowing the normal bandpass filters to be selected. I guess you will be using your own radio so I suggest that you allow binary selection of the user outputs for similar filter switching.

Before giving an unqualified yes to using a succession of High pass/Low pass filters I would like to try seeing if there is any unwanted interaction when they series up. In theory it would be possible to build a filter band with a group of high pass filters for the start of each ham band and a group of low pass filters for the end of each ham band and you simply select the start and finish filters to give the bandwidth you want to skim. There would be no need for a filter for every band as a HP for 17M could mate with a LP for 15M and there would be no need for a filter in between. Might not be the way everyone would want to go but would be the most versatile and certainly cater for skimming and other wideband requirements.

73, Graeme zl2apv

[Steve Haynal]

Hi Claudio,

Thanks for the link. It is a very worthwhile paper to read.

73,

Steve

KF7O

[Steve Haynal]

Hi Alan,

The dither I tried was added in the DSP processing before the DAC to try an clean up spurious TX emissions. It would be interesting to try adding dither to the input of the ADC, even with the TX as Claudio suggested.  

73,

Steve KF7O

[Alan Hopper]

Graeme, John,

thanks for the suggestions and consideration of skimming options. I really didn't mean to disrupt all the hard work being done on this but these radios do lend themselves to skimming.  Building some filters is certainly on my ever growing project list!

73 Alan 2E0NNB

[Graeme Jury]

Hi List,

It has been a wet and windy winter's day here so I thought I would take a look at the most problematic filter which is the 30/20M band filter. Staying with the 1.5uH inductors I spent several hours running a number of optimisations and finally came up with ...
1.5uH - 160pF - 300pF - 1.5uH -330pF - 300pF - 1.5uH - 160pF
These values give a definite improvement and takes out the bandpass ripple but unfortunately only one of the original caps can be used (330pF) and it is necessary to find 2 x 160pF and 2 x 300pF. I prefer where possible to stay with the original inductor values and up to this point they seem to be the best choice anyway.

73, Graeme zl2apv

[in3otd]

Hello,
out of curiosity I have measured the IMD2 of my Hermes-Lite when, in addition to the two test tones, also additional noise was presented to the ADC. The test signals were at 6.1 MHz and 8.1 MHz and the IM2 product measured was the sum frequency at 14.2 MHz; the noise added came from my vertical antenna, where the signals in the 20 meter band were removed using a notch filter to avoid masking the intermodulation product.

                   _________
generator_1 o-----|         |
                  |         |
                  |combiner |----.
                  |         |    |     _________
generator_2 o-----|_________|    '----|         |
                                      |         |
                                      |combiner |----o to Hermes-Lite input
                   _________          |         |
                  | 20 m    |    .----|_________|
                  | band    |    |
    antenna o-----| notch   |----'
                  | filter  |
                  |_________|


The intermodulation product was measured also without any external noise added, connecting a dummy load instead of the antenna.

The results were quite interesting:

with the external noise, the IM2 product is greatly reduced in amplitude, likely spread out over the whole ADC bandwidth by the "dithering" introduced by all the signals present.
So maybe a direct-sampling RX works better (or appears to) when no input filters are used, hi.

73 de Claudio, IN3OTD / DK1CG

[Graeme Jury]

A very interesting post Claudio. I am currently playing around with receive filters and maybe unwittingly not doing the best thing. For some years I have run my HiQSDR with a 60 MHz Lo pass and a 1.7 MHz Hi pass filter arranged with the 60 MHz filter in the antenna line and the 1.7 MHz filter in cascade via the aerial changeover relay to the receiver. This totally stopped the broadcast breakthrough that I experienced with the nearby broadcast stations producing in excess of 2 volts into my antenna and I must say that I never really noticed any significant IMD from the normal shortwave traffic although I don't have any nearby hams. Maybe band stop filters at the second harmonic would give a better service.

73, Graeme zl2apv

[John Williams]

Do you need a RX filter board to test with? I do not recall if I sent you one. I am a fan of no RX filters on input. I see little need for it at my QTH.

John -

--

[Steve Haynal]

Hi Claudio,

Thanks as always for your terrific and illustrative measurements. They help clarify what are often just fuzzy intuitions in my mind.

It seems that there are two concerns regarding whether BPFs can help improve HL RX. First, traditional IM2 products and measurements are not that meaningful and can be reduced with dither as you just showed and as described in the paper by SM5BSZ that you linked to earlier. This assumes (correct me if I am wrong) that there is no clipping of the ADC. The second concern is that AM broadcast or other signals are so strong that there is clipping of the 12-bit ADC and the HL LNA gain must be reduced to avoid clipping. Reducing the LNA gain avoids the clipping but reduces sensitivity and so some weak signals can no longer be detected. For this second case, BPFs can still help. Do you agree with this reasoning?

If we imagine the ideal dither, it seems it should be a frequency modulated sine wave with *predictive* amplitude and frequencies to null strong signals not of interest that would cause ADC clipping and reduce MDS yet keep the input to the ADC always nearly at maximum swings. Maybe we can approximate something like this by feeding attenuated TX to the RX ADC input. What are you using for the combiner in your tests?

The AD9866's LNA is always in front of the ADC and will add its own distortion as it is an analog circuit. Although purists do no like anything in front of the ADC, I don't see how one can get good results without such an LNA with an inexpenesive 12-bit ADC.

I will forward you result to AB4OJ. He is planning to redo measurements on the HL2 and your results have bearing.

73,

Steve

KF7O

[Graeme Jury]

Hi John,

No thanks, I got one of your original ones and have done all the mods to the filter switching and Peregrine switch inputs as per Glenn.

Thank you for thinking of me.

Cheers, Graeme

[John Williams]

Sorry Graeme, I should have addressed this to Claudio. So sorry...

John

[Graeme Jury]

Hello Steve,

I think I need to make some diagrams etc. to fully answer your questions so this is a brief response to be followed up later.

Your first point. Basically the Rx filters are Ham Band only and the through filter is to be switched in for all other frequencies. i.e. if a ham band filter is selected the through filter is deselected and vice versa. In cases where the second harmonic of the lowest band of a pair can be kept 25 dB or so down the filter is made to cover 2 bands e.g. 60/40, 30/20, 17/15 and 12/10. I am also looking at a roofing filter at 32 MHz amd a flooring filter at 1.7 MHz

Your second point. Yes 17/15 is a pair and we have already got this filter built. The separate filters are old information which has remained but obsolete. I need to make a diagram of what I use for Rx filtering to get things really clear. For Tx filters John W9JSW has a perfectly good set as are the filters on the RS-HFIQ Tx. I have not done any testing to see the effect of the PIN diode switching on harmonic output but of course the relay switching by John is tried and true. The relays have around 1 pF from armature to coil which is the worst case capacitance and reinforces the need to RF ground both sides of the coil with capacitors.

Third point. I got all my toroids from "Toroid King" except for some that Glenn sent me but I have not used them yet.

73, Graeme

On Saturday, July 16, 2016 at 5:39:36 PM UTC+12, Steve Haynal wrote:

[Steve Haynal]

Hi List,

I've been thinking a bit more about this result from Claudio. If we accept that IMD2 RX products are not a problem for the HL RX and that we only care about blocking a few occasionally strong signals, then perhaps we can get by with only 3 high pass filters on the RX. It seems that most HF "noise", especially at active times, is it on lower frequencies. If we have RX high pass filters at 1.8, 5 and 10 MHz in concert with the LPFs for TX, this may allow us to cut out most of the noise that causes ADC clipping but still pass enough noise through for dithering. :) Also, these filters do not need to be very sharp as we only need some attenuation of the strong signals to avoid ADC clipping. Maybe we should even start thinking in terms of (wider) notch filters.

I've also been thinking about when ADC clipping is really a problem. I often run the LNA at levels where there is considerable clipping reported when doing WSPR skimming and never seem to notice problems. Maybe we can categorize clipping into two types:

A. Dominated by one or two very strong signals. These signals may cause the ADC to be "pegged" for relatively long periods of time at regular intervals.

B. Random clips because a multitude of signals or noise happened to add up for a clip at that moment. These may be very random in duration and occurrence, but should occur more frequently with high LNA settings and/or at active HF times.

If the clipping is of type B., and causes losses or distortion in the signal we care about only 4 to 5 percent of the time, may be we shouldn't even care for any practical uses? In other words, how much "random" clipping does it take before a signal we care about is degraded enough that it is noticeably worse, or becomes unusable? There must be academic papers on this topic.

[Alan Hopper]

Steve,

I have run a longer test with 2 hermes lites, one with band pass and one without. After about 6000 wspr and jt65/9 spots the one without the band pass was 2% ahead so at my location it would appear that I'm better off without a band pass.

I have also wondered about clipping, I ran a test with one radio with the overload indicator on most of the time and the other one not, over a few hours they were within 1% of each other.  It would be interesting to tweak the firmware to give a more accurate measure of just how much clipping is actually happening, maybe number of samples clipped per packet.

73 Alan 2E0NNB

[va7...@gmail.com]

Hi Claudio,

Steve asked me to provide some input to this discussion, as I  tested the Hermes Lite about a year ago.

http://www.ab4oj.com/sdr/hermes_lite/hl_notes.pdf

I agree with your description of the sources of IMD in an ADC. Odd- and even-order products generated by the quantisation process are very likely the main source of IMD until the input power to the ADC is sufficient to drive on-chip analogue stages in the ADC into odd-order non-linearity. This can be seen in the IFSS curve (Fig. 4 on p. 4 of my report). IMD products and reciprocal mixing noise (the latter normally very low in an ADC) are the main components of the idle-channel noise in the Noise Power Ratio (NPR) test.

Dithering and randomisation performed by an on-chip dither generator and randomiser reduce odd-order IMD by de-correlating the IMD products into noise which can degrade the ADC noise floor. The output randomiser may be more useful for offsetting the defect of "hash" from the digital side of the ADC than the dither generator. I have found in the course of testing some SDR's (e.g. the Perseus) that whilst dither reduces the impact of IMD3, it has virtually no effect on DR2 (IMD2 dynamic range). As for "incidental" or "external" dither, I feel that the jury is still out on this. I recently set up a test in which I applied a 2-tone test signal to a Perseus with dither off, measured the IMD amplitude, then injected white noise from my NPR noise generator (with all bandstop filters out) into the third input port on the combiner. As I varied the noise loading, I saw the noise floor increase and decrease with the applied noise level, but the level of the IMD products did not change. When I removed the noise loading and activated internal (on-chip) dither in the Perseus, the IMD product amplitude decreased markedly. This leads me to question the concept of incidental dither, unless my test procedure is in error. I would value your comments.

The main source of severe IMD2 in an on-air receiving environment is a mix between two strong stations outside an amateur band which throws a 2nd-order product into the amateur band. My test plan specifies test signals at 6.1 and 8.1 MHz, which will throw a product at 14.2 MHz. I have found that an RF preselector (either an amateur-band BPF or a suitable 1/2-octave BPF encompassing the amateur band) is required to reduce IMD2 to an acceptable level in these scenarios. Hence the comment in Section C.1 of my report.

I conduct IMD3 and IMD2 with the same test fixture: two signal generators, each driving a 1W MCL amplifier with 30 dB gain. Each amplifier in turn feeds a combiner input via a 15 MHz LPF (for IMD3 tests at 14 MHz) and a 20 dB pad. This configuration assures good isolation between the generators.For IMD2 testing, I use a pair of test frequencies < 15 MHz. If required, LPF's cutting off at lower frequencies can be substituted.

Although I always use the IFSS method for testing IMD3 in SDR's, I still take a single-point measurement for IMD2 testing. This is discussed in a paper i presented at the SDR Academy at Friedrichshafen this year.

http://www.ab4oj.com/sdr/sdrtest2.pdf

If it is desired to use the same receiver for wideband signal analysis, skimming etc. and for narrow-band single-signal reception, a means of bypassing the preselector will be needed. In an RF environment where even-order IMD interference is likely, the preselector will need to be switched in..

73, Adam VA7OJ/AB4OJ

On Sunday, July 17, 2016 at 6:57:07 AM UTC-7, in3otd wrote:

[in3otd]

Thanks John,
I do not have a RX filter board, but I think I won't have much time to experiment with that also, as I'll soon have the PA boards from Steve to test...

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Alan,
interesting that the one without the band pass was ahead I would have thought that nevertheless the one with the filter should see a (slightly) lower noise floor caused by any IMD and be slightly better then.
The two radio have the same sensitivity, i.e. the filter losses are negligible (or an equivalent loss is inserted for the radio without filters)?
Do you have the wspr logs to compare the SNRs for the spots received on the two radios?

73 de Claudio, IN3OTD / DK1CG

[in3otd]

Hello Adam,
thanks for your detailed comments; I agree with your analysis below about the origin of the intermodulation products.

There are not many details on how the "internal dither" is implemented in the ADC used by Perseus, it may be that it uses a small dither amplitude that is only partially effective, but this is just a wild guess.

Regarding the result of the "incidental dithering" test you did with Perseus, I think that if the origin of the IM2 was "analog", so due to the preamplifier stage where the lower order distortion terms are dominant, you may not see any improvement due to the added noise. I would expect some improvement only in the portion where the IMD is due to the ADC (imperfect) quantization, so at low input levels.

73 de Claudio, IN3OTD / DK1CG

[Alan Hopper]

Claudio,

I was also surprised. For these tests I used hardware rf agc capped at +30db. The results could have as much to do with the gain differences.  I swapped the front ends half way through to eliminate other differences.  I want to do some more tests to optimise gain setting/agc, I shall then rerun the bandpass test.  I'm doing these tests to work out what works best here so I am not adding any extra loss to compensate for differences i.e. if the cause is filter loss and that results in fewer spots it is still the less good option.  I realise this is less useful to really understand what is going on.  Once I have a fair agc I shall log all the spots for comparison.  

73 Alan 2E0NNB

[in3otd]

Hello Steve,
I would say that IMD2 RX products at low interfering levels are often not a problem in practice, since these are spread out by the "incidental dithering" caused by the other signals entering the ADC. This would also explains why users did not notice any particular issues with RX IMD even without any input filtering, except when the signals were so strong to drive the ADC into clipping. It is likely that the RX IMD in practice manifests itself just as an increase in the noise floor, which can just be mistaken for the noise floor picked up by the antenna.

In any case, sub-octave filters will improve the IMD2 performances, also at stronger input levels, where the RxPGA distortion becomes relevant. The best case would probably be with sub-octave filters and explicit out-of-band dithering applied, since in this case we will spread out the ADC IMD and protect the RxPGA from strong signals. This will also guarantee a predictable dithering: I measured again the IMD2 with the "incidental dithering" this morning and the results were of course a little different that those of yesterday evening, since the signals in the lower bands were not so strong this time:

Note that the IMD2 at higher input tones levels was not reduced with the lower signals levels present this morning.

Similarly for clipping, as you saw, some limited clipping can probably be tolerated, it will likely just increase a bit the noise floor, which may already be high due to the external noise.

Regarding the combiners, I have used a couple of Mini Circuits ZFSC-2-4 since I had them around but homemade ones will work just as well in the HF bands; some 6 dB hybrids that can be used as splitters/combiners are described on my website (http://www.qsl.net/in3otd/ham_radio/6dB_hybrid_couplers.html) but 3 dB couplers (like the Mini Circuits I used) are nicer because of the lower losses.

73 de Claudio, IN3OTD / DK1CG

[James Ahlstrom]

Hello Group,

This is such a fun thread I couldn't resist adding my $2E-2 worth.

First on Steve's point about using a LNA before the ADC.  An ADC has a noise figure that can be calculated and measured.  For the calculation see this reference:  http://www.analog.com/media/en/training-seminars/tutorials/MT-006.pdf.  Even on an ideal basis (no ADC faults) and a 14 bit ADC, the noise figure is greater than 20 dB. This is inadequate for the higher HF bands, so there is no option but to use a LNA, whether purists like it or not.  I think the built-in variable gain LNA is a great strength of the Hermes Lite project.  Combined with RF AGC I think we can outperform many of the SDRs out there in a practical application.

Rightly or wrongly, I have never believed in internal dithering, pretty graphs in application notes notwithstanding.  As hams, we often encounter signals near the noise level.  The time-honored test is to connect your antenna to your radio and make sure the noise level rises.  So the band noise must always be visible and will always provide dither.  Out-of-view signals will add additional dither.

The type of receive filters required mostly depends on the QTH.  In Europe you may need a notch filter for short wave broadcasters running 100 kw nearby.  In the USA, you may need a 1.7 MHz high pass filter for AM broadcasters.  For field day events, don't ask.  I can run my HL without any filters at all, even though I am only 40 km from NYC, but I can not advance my RF gain too high.  This provides adequate performance for me at this QTH.  We should add Rx filters but not over-think it.

I have also noticed the phenomenon that increasing the RF gain well into the clipping region degrades the audio surprisingly little.  I am not sure why that is, but I like it.

Jim

N2ADR

[Graeme Jury]

Hello Group,

Have been doing some more filter testing on the receive filters and have altered the proposed 30/20 receive filter as it had a funny dip in the passband (see plot from Glenn here). I tried to use as many of the original components as possible so it is not the best that can be done but is not too far. I have attached a plot of the response which includes the loss in the PE4259 switches and a plot with the switches turned off to show the isolation. As can be see there is some spill over in the board and the filter plot takes the general form of the isolated plot in the upper frequency regions. In my past experience it is very difficult to get more than 40 dB isolation across any device without resorting to shielded compartments. When building attenuators I never try for more than 20 dB for any one section.

I replicated this filter with T37-6 toroids and simply used the calculator on the Toroid King site for the turns without bothering to measure them. The filter had an identical response but the passband loss was much better, more than a dB improvement. I am becoming happier with the Peregrine PE4259 switches but am less happy with low Q inductors and want to build a board up with toroids, probably in the T30 size where I do not measure anything but simply construct from calculated turns. It might be a failure but at least I will know if it is possible and if it is worth it. With all the chit chat going on over filters I am starting to wonder if they are worth it at all. I certainly have had outstanding results with my 1.7MHz hi pass and 60 MHz lo pass filters with nothing else except perhaps the lo Q filtering from my antenna tuner. I decided after 50 years of being a tinkerer I would learn CW and go for a DXCC this last sunspot cycle using my SDR. With that filtering it was easily done and so convenient to be able to look anywhere on the bands. This is taking a long time and getting convoluted I am taking on board Jim's N2ADR comment "We should add Rx filters but not over-think it".

73, Graeme ZL2APV

[Alan Hopper]

Claudio,all,

I've been thinking about generating dither in duplex mode. There is a small snag in that you cannot currently use RxPGA gains over +19db whilst transmitting, this appears to be an undocumented feature of the AD9866.  I'm interested to see if my band pass result can be improved with dither.  Your morning and evening results are very interesting and might well explain why opinion on "incidental dithering" is split.

73 Alan 2E0NNB

[in3otd]

Hello,
ah, didn't know of that limitation, maybe it was done to limit the possibility of crosstalk??

Regarding the "incidental dithering", it makes sense that the dithering signal has to be at least comparable to the interfering signals, so all the ADC codes are "swept" by the dithering signal.
This may also be the reason why VA7OJ did not see any effect when using wideband noise: seeing the noise floor increase from the added noise is likely not enough, the added noise amplitude (not the power!) should be at least comparable to the interfering tones amplitude, so much higher than the noise floor. So if noise is used for dithering I think it should not be (too) broadband and most important not in the band of interest.

73 de Claudio, IN3OTD / DK1CG

[Steve Haynal]

Hi Adam,

Thanks for the feedback. I think your numbers and Claudio's agree when there is no external dither as you can see the IM2 levels rise at around -75 dBm power-in for both sets of data. It is interesting to think of why Cluadio saw better results with incidental dither than you did with random noise. I think Claudio's incidental dither contained more energy, at least for specific frequencies outside of the band of interest, than your random noise generator. On an active evening in Europe, I suspect there were tens of signals with power levels of more than -75 dBm, or greater than S8 to S9. As these "external dither signals" add and subtract in various ways, it would have caused the input voltage to swing across most of the ADC codes and thus better average out level quantization errors. In his paper, SM5BSZ recommends relatively high amplitude frequency modulated dither(s) out of the band of interest similar to what Claudio experienced. It would be interesting to repeat your experiment with a band stop filter on the noise for the band of interest, and increase the random noise generator output considerably for larger out of band dither. The same effect may not be observed with a Perseus given the different ADC. Perhaps this experiment can be tried the next time you have access to a HL.

Another simple experiment for users in Europe to try is to just pick two strong signals in the 6 to 8 MHz range some evening, and see if you can pick out any distortion "bump" at these two frequencies added.

73,

Steve

KF7O

On Sunday, August 7, 2016 at 1:02:23 AM UTC-7, wrote:

Hi Claudio,

Steve asked me to provide some input to this discussion, as I  tested the I think there are (at least) 3 differences:

 Hermes Lite about a year ago.

http://www.ab4oj.com/sdr/hermes_lite/hl_notes.pdf

I agree with your description of the sources of IMD in an ADC. Odd- and even-order products generated by the quantisation process are very likely the main source of IMD until the input power to the ADC is sufficient to drive on-chip analogue stages in the ADC into odd-order non-linearity. This can be seen in the IFSS curve (Fig. 4 on p. 4 of my report). IMD products and reciprocal mixing noise (the latter normally very low in an ADC) are the main components of the idle-channel noise in the Noise Power Ratio (NPR) test.

Dithering and randomisation performed by an on-chip dither generator and randomiser reduce odd-order IMD by de-correlating the IMD products into noise which can degrade the ADC noise floor. The output randomiser may be more useful for offsetting the defect of "hash" from the digital side of the ADC than the dither generator. I have found in the course of testing some SDR's (e.g. the Perseus) that whilst dither reduces the impact of IMD3, it has virtually no effect on DR2 (IMD2 dynamic range). As for "incidental" or "external" dither, I feel that the jury is still out on this. I recently set up a test in which I applied a 2-tone test signal to a Perseus with dither off, measured the IMD amplitude, then injected white noise from my NPR noise generator (with all bandstop filters out) into the third input port on the combiner. As I varied the noise loading, I saw the noise floor increase and decrease with the applied noise level, but the level of the IMD products did not change. When I removed the noise loading and activated internal (on-chip) dither in the Perseus, the IMD product amplitude decreased markedly. This leads me to question the concept of incidental dither, unless my test procedure is in error. I would value your comments.

The main source of severe IMD2 in an on-air receiving environment is a mix between two strong stations outside an amateur band which throws a 2nd-order product into the amateur band. My test plan specifies test signals at 6.1 and 8.1 MHz, which will throw a product at 14.2 MHz. I have found that an RF preselector (either an amateur-band BPF or a suitable 1/2-octave BPF encompassing the amateur band) is required to reduce IMD2 to an acceptable level in these scenarios. Hence the comment in Section C.1 of my report.

I conduct IMD3 and IMD2 with the same test fixture: two signal generators, each driving a 1W MCL amplifier with 30 dB gain. Each amplifier in turn feeds a combiner input via a 15 MHz LPF (for IMD3 tests at 14 MHz) and a 20 dB pad. This configuration assures good isolation between the generators.For IMD2 testing, I use a pair of test frequencies < 15 MHz. If required, LPF's cutting off at lower frequencies can be substituted.

 I think there are (at least) 3 differences:

[Steve Haynal]

Hi Jim and Alan,

Currently, when the ADC clips it starts a counter so that the clip signal is stays high for enough time to actually see a LED go on. It may be that the clips which are occurring are very short but fairly frequent which makes it looks worse with the LED or software indicator. We should enhance the firmware to provide better statistics and software feedback on clipping.

73,

Steve

KF7O

[Steve Haynal]

Hi Alan and Claudio,

I really hope this limitation is a misunderstanding on my part. I could find no documentation of it in the datasheet. But currently the firmware has to lower the LNA to no more than +19 dBa for full TX power to occur during duplex.

73,

Steve

KF7O