Title: UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 26. May 2018, 08:25:44
The UHSDR firmware now has a good TX image rejection setting, but it can not improve carrier suppression that depends on the hardware and is of about -30/-35 db.
Of course we could add a hardware balance as it is in the Eagle TRX, but its calibration is valid for one frequency only and does not solve the problem
To avoid transmitting unwanted signals in the band away from the carrier frequency, it is possible to adapt the UHSDR firmware to switch from a TX audio shift approach to a carrier shift approach as better described in the HB9DRI article (http://www.linkrf.ch/IQ+XT.html (http://www.linkrf.ch/IQ+XT.html))?
’73 Paolo IZ6MAF
EDIT:
corrected link
|
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 26. May 2018, 09:09:27
I do not see there the way carrier (LO) suppression is made. It is impossible that it is only an oscillator with less phase noise - phase noise is not the reason. It is the isolation of TX mixers.
Additional, we already have "carrier shift". When we change frequency not the audio signal is tuned to new frequency, but the LO as it is described in the pereliminary paper.
What I am thinking about is to switch from "Xlate TXing" to "non-translated TXing" in all modes. The result would be that the unwanted LO signal is very near to your TX signal (in fact it is same as in conventional analog TRX). Possible impact: suppression of unwanted sideband is decreasing.
At OVI40 RF I go a complete other way for CW: CW is not generated by mixing but by "switching LO signal directly". I have developed a schematic that suppresses output in not-keyed state > 80dB and does not have any spurs (except hamonics of 3rd, 5th, 7th and so on order).
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 13. June 2018, 20:28:25
Hello Andreas, I am attaching a brief summary of the tests I have done to improve the TX Carrier suppression on my mcHF/Eagle and what I hope you can do even better by optimizing the firmware
vy 73 Paolo IZ6MAF |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 06:59:28
Hi Paolo,
many thanks for your detailed description. Yes, all you wrote is right and much of this is already implemented in OVI40 TX mixer. This is not very astonishing because all this "are not secrets" and you can find hints scattered (!!) all over the web. My prototype is running well (and that is the only difference to your proposal it does not use high precise opamp). Swapping the opamp does not result in measureable improvements. But the rest does (and concept gets simpler and better reproduceable by leaving away components)...
I am already thinking about txing without any shift but because that is not hardware and can be done at any time I have not completed my rhoughts (I am very hard working at the OVI40 RF hardware).
So let us discuss about changing to non-xlate txing!
In theory this results in a complete loss of spurs in 2 * xlate difference to your TX signal but in increasing txing "the wrong sideband". I have not done any investigations but may this be problematic?
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 14. June 2018, 08:28:51
Hi,
first of all, thank you Paolo for your work! All the mods you've done are pushing old mcHF project much forward.
We can do option in menu for test (Debug menu) and have some tests against problems. Will look into the code. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 09:02:36
Hi Slawek,
it is easy I think. You must look for code which does decision in TX to send xlated or not and if you set option "TX not xlated" it must always result in "no translation". We already Tx in CW non-translated always so this code parts can be the anchors for looking.
I have tested it by manually disabled xlate in RX and TX and found no significant difference in spurs at TX. But the hardware improvements are mandatory (some of them are already in the papers from Joe Tayloe which are much older than mcHF...)
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 14. June 2018, 09:10:56
Yes, ofcourse Paolo can test this idea with Xlate off without firmware change, and then we decide to add it or not to menu, ok?
|
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 09:50:56
Yes - 100% my idea. Or - extended to: remove all xlate stuff for TX.
I only have done empiric tests (receiving my own signal which is produce by a mp3 sound source). I recognized complete removing of all spurs outside +/- 3KHz from TX frequency (except intermodulation spurs but they are very, very weak). But now "mirror suppression" is working as a suppression of "unwanted sideband" (of course). It is neccessary in non-translated mode, too but if we decide to roll back to non-translated mode we would only need ONE TX IQ setting and not two which we have actually (SSB and CW).
I am tending to always transmit in non-translated mode and remove all stuff which is used for xlating in TX. But my tests are only empiric (no measurements done). Paolo: can you please measure TX behaviour if you switch to non translated and let us know the result? I think this would be a big step for mcHF. Not sure if Chris would accept this for his new release but it could be added to our modification page of mcHF immediately. For OVI40 it is not neccessary because of hardware concept includes your ideas already.
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 14. June 2018, 09:59:23
| I will also test this with spectrum analyzer, and publish some pictures here at evening. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 10:22:37
Very good. We need reliable results which are manifested by measurements.
You can test it without the hardware modification of Paolo. The only difference will be a lower suppression of carrier. All other spurs observations are portable.
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 14. June 2018, 10:32:37
| I already have almost all modifications described by Paolo, but without voltage bias decrease for TLV2464 and thus FST3252. I played some with potentiometers but as Paolo described - it works but only for one band. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 11:04:43
The reason is that the effect is produced by unsymmetrical capacitive elements on both switching channels. They are frequency dependent. One solution is to use switches with a much lower capacitive influence.
For OVI40 RF I am using GHz multiplexers from Potato. They are capable of multiplexing much higher frequencies and do have of course significant lower capacitive influence.
Disadvantage for mcHF:
They are not available in same package so no drop-in replacement is possible. If Chris would change the layout mcHF can use old switches and Potato similar - package of Potato is availble for "standard multiplexers", too...
But Chris has ignored existence of Potato divider since years so I do not think this will happen.
EDIT:
We are discussing in the UHSDR board. So main interest is the firmware part not the hardware part... Ideas of Paolo and everyone are of course highly welcome and may result in an integration to OVI40 - now or later. OVI40 will live because of such contributions ::)
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 14. June 2018, 11:19:15
I quickly repeated the measurements at 3500 and 28500 khz with and without Xlate and as you can see from the images without Xlate there is a considerable production of spurious. What will they be due to?
'73 Paolo |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 14. June 2018, 11:27:28
That must be discussed.
Are you using TUNE function with TUNE frequency 750Hz? We must calculate what can be the reason for the spurs...
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 14. June 2018, 17:05:53
I am using tune, single tone, 750 hz.
It would be very useful if others could repeat the tests I performed to make a comparison.
vy 73
Paolo |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 14. June 2018, 22:35:06
Hi,
some measurements of xlate on/off. I think can confirm your observations with spurs, ony last spur is visible because of my spectrum analyzer RBW limit. After weekend will test on 10Hz R&S one. Anyway, some other interesting wideband observation inside the pdf. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 15. June 2018, 05:53:47
First conclusion:
It looks that signal with xlate=off has less spurs than signal using xlate. But there are some spurs near by the signal in xlate=off where I do not dee the maths where they are coming from.
Normally I would suggest:
f(LO) + F(mod) = our "wanted" signal
f(LO) = poorly suppressed local oscillator
f(LO) - f(mod) = unwanted sideband
where (mod) = 750Hz
But I see something ~ 3KHz. I don't understand where it is produced. Is it some "filter ringing"? Does it change if TX filter settings are touched?
If we can investigate where it is coming from we could switch to non-xlate TXing.
But I am not sure at the moment if that will work fpr AM or FM. I do not completely understand the maths which are standing behind them.
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 15. June 2018, 06:32:14
Hi Andreas
i suppose that ~3kHz is one of the spurs (leftmost showed by Paolo) that is visible out of the resolution filter limit in sprectrum analyzer. I will do furher investigation with N2PK like 150MHz VNA today. In some circumstances it can work as a spectrum analyzer with very narrow resolution filter. Unfortunatelly i do not have Zeus like SDR used by Paolo. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 15. June 2018, 16:56:38
Hi,
Some more measurements with N2PK VNA used as spectrum analyzer. This exactly confirms Paolo measurements. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 15. June 2018, 17:42:32
If I see the measurements I am asking myself if the spur at the other side of LO (in translated mode) can be better suppressed by individual IQ correction??!! Can you pse test that?
And, again, I cannot understand why non-translated mode has much more spurs than translated...
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 15. June 2018, 18:34:44
Yes, it can by software setting (see pdf). I choose the 10m band to better show the problem because spurs are bigest. On 80m it looks better. I didn't touch the opamp bias voltage (still on 2.5V). We have spurs also in translated mode but are wider spreaded (look at my yesterday's measurements with 1MHz span).
I have to look at the dac output with soundcard and fft. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 15. June 2018, 21:35:24
| Hmmm, changed the components to proposed by Paolo (except the opamp) and have not noticeable profit of it, carrier stays at exactly the same level. Seems that probably precision of opamp does the job here. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: DF8OE on 16. June 2018, 06:36:42
No. Carrier would stay without any signal, too.
We must split into TWO parts:
1) suppression of LO ("carrier")
2) suppression of spurs
1) is only a result of mixer. There are three components:
i) capacitive coupling from LO inputs to any multiplexer output (the reason is buried in design of mixer
die)
ii) output which results in unsymmetrical behaviour on the mixer inputs (DC offsets and capacitive
components
iii) R(off) of the multiplexers is not high enough
2) is more complex. It can be a result of phase and amplitude differences on the four inputs. Additional there will be spurs when signal is distorted (on one ore more inputs). Some parts can be suppressed by firmware (phase / amplitude errors). But if signal is distorted there is no chance to suppress anything by firmware. So I agree opamp should not produce any distortions or unlinearities. But this is provided by a "normal" opamp - you do not need precise ones.
So you see there must be different approaches to reduce each of this effects. Very complex at all ::)
vy 73
Andreas |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 16. June 2018, 07:11:26
Hi Andreas,
as the mod is not working i went back with the rest of the mod, soldered the coupling capacitors and have the mixer connected like in first schematic of 0.4 from Chris. I did it because i suspected input offset amplification on output of the opamp (my words about precision of opamp). The result: nothing changed. So the problem lies around the FST3253. The carrier suppression changes wit frequency, with increasing it, the carrier attenuation is worse. It points to conclusion that there is capacitive couplig across the mixer.
I think all the rest of spurs is due to AM modulation by presence of carrier. In translated mode the modulation is wider and in non translated mode it is narrow. |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 16. June 2018, 08:41:13
As Andreas said, the problem is very complex because so many variables are involved.
To better analyze the QSE mixer I started a new series of tests: I disconnected the R84 R85 resistors from the IN I and IN Q signals, and connected them to the bias in order to make measurements / modifications on the mixer that are not influenced by the UI board.
Too bad that LTC 5598/99 of Linear Technology can only be used from 5 mhz up, if we could have used it would have solved all the problems.
look at that very interesting tube https://www.youtube.com/watch?v=SUyNyIYcl6U
vy '73 Paolo |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 16. June 2018, 09:14:20
Hi Paolo,
using of integrated specialized IQ modulator involves new pcb design. Let's wait for OVI40 RF board, you will be suprised... |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 18. June 2018, 18:50:43
Hi Slawek
With the R84 and R85 resistors disconnected and connected to the bias voltage of the OPA 4192 (1,65V), only the carrier signal appears, which after changing the sn74AC74 with a PO74G74 powered by an LP2985 / 330 has an increase of 32 db when i move from 80 m to the 10 m (before was 40 db) and of course it is independent of all settings.
If I change the voltage across the resistors the signal changes in frequency and amplitude, but remains clean.
I begin to think that the spurs can come from channels I and Q coming out of WM 8731 that are not properly filtered as specified on page 13 point 3 of the data sheet.
I am very satisfied as the L.O. leakage now in 10 m is lower than 50 db of Tune single tone 750 hz transmission signal at a power of 5 w.
I look forward to the RF OVI40 card that will surely be even better.
Please could you compare my results with your tests?
vy '73 Paolo |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: IZ6MAF on 18. June 2018, 18:57:05
Hi Andreas
Why the text that I write in English (see attached) when I send the post is changed?
vy '73 Paolo |
Title: Re:UHSDR TX Carrier shift instead Audio Shift
Post by: SP9BSL on 18. June 2018, 20:40:32
Hi Paolo,
changing the 74AC74 to Potato shows that there is also a kind of phase issue. Unfortunatelly I do not have both the Potato chip and the OPA4192 here. Yes for this moment 50dB looks good, i'm afraid that this is all what can be done with mcHF RF. The problem lies in TX mixer chip which have high leakage caused by internal capacitance. I planned to check output of the dac (i mentioned that already speaking about checking dac output by fft) will try to connect an analyzer to this week.
And speaking about English - it is Google translation. Sometimes when i do not understand, simply move cursor on text, and the original is showed. I think it is due to English->German and then back to English. |
Diskussions- und Newsboard des DARC-Ortsverbandes I40 | Powered by YaBB SE
© 2001-2003, YaBB SE Dev Team. All Rights Reserved.
|