2012/8/8 Jean-Samuel Najnudel - BJT PARTNERS SARL jsn@bjtpartners.com
Hi Andrey,
Thank you very much for your reply.
I understand your point of view regarding the ADF4350. I agree with you, the ADF4350 still have quite good performances. I also agree that easier software support is also important. It would just be a great news if we could pass the GSM spec, at least as micro BTS.
Regarding the GSM 1800 version of the selectivity board, it will not be a problem to change the mixer and the filters. However, are you sure we will be able to get the correct matching if we only change the passives ? Are you sure we would not need to modify the PCB traces to get a good matching for a 1800 MHz version of the board ? What do you think ?
Most of cooper would be 50 Ohm, so only FR-4 losses a bit higher. Also RF cables losses and so on, therefore 1800 parameters will be a bit worse, but it typical. About components, I've checked again and found that RF SAW filter for 1800 UL require coil at input (added).
Best regards, Andrey Sviyazov.
Best regards.
Jean-Samuel. :-)
On Tue, Aug 7, 2012 at 7:27 PM, Andrey Sviyazov andreysviyaz@gmail.comwrote:
Hi Jean-Samuel.
1/ A few weeks ago, we discussed together about phase noise issues to be
able to pass the GSM spec. You explained HMC830LP6GE has much better performances than the ADF4350. I know the HMC830LP6GE is a bit more expensive but, as discussed together before, I really believe this worth the extra cost. This would allow us to pass the macro or at least micro BTS specs without the need for designing another slectivity board. This would save a lot of work. Moreover, we espcially need this preselectivity board for a micro or even macro BTS. In this use case, I really believe price difference (like 40 or 50 USD) will be very low, especially compared to all other costs of a mid-long range micro/macro BTS installation. Unless you really disagree, I would really suggest to use the best performance components as we reasonably can.
First of all, Robin said that ADF4350 already supported by UHD therefore it could save time to implement it for UmTRX. 2nd, I think that parameters differences isn't really as significant (few dB) as price and soft changing time cost. 3rd, it isn't last design and I am sure that we will change it for more functions, performances and for housing adaptations. Also I'm planing to buy HP4352 to design and measure PLL's, so of course I'll test HMC830 too, but now we just have to be sure that it will work properly and without risks.
2/ In my deployment in Mayotte, I will also deploy GSM 1800 band base stations because most of the spectrum I have is in this frequency band. To avoid making another PCB for GSM 1800, could you design the board to allow to populate either GSM 900 or GSM 1800 filters and matching passives with the same PCB copper ? A single PCB would be easier to test and to maintain. Moreover, this would save some costs as we will be able to order lager PCB volumes. This could make a quite significant saving. Please let me know what you think about this.
As I mentioned before, there are band specific components with same footprint. So, there are no changes required in PCB design for 1800 band, only BOM. For example, HMC485 instead of HMC483 and FAR-F6KA-1G7475 instead of FAR-F5KA-897.
Best regards, Andrey Sviyazov.
Anyway, thank you very much for this design. With UmTRX, this would really make a cost effective and great performances base station. Thanks a lot.
By the way, as I am abroad, I may take a bit more time than usual to reply to your e-mails. I really appologize for this and I will do my best to reply as soon as I can.
Best regards.
Jean-Samuel. :-)
On Mon, Aug 6, 2012 at 5:07 PM, Andrey Sviyazov andreysviyaz@gmail.comwrote:
I've found first error: Missed AC block required for Vref pin of ADF4350. Also translated some components descriptions to English to avoid misunderstanding of Russian. All newest files here in zip.
Best regards, Andrey Sviyazov.
2012/8/6 Andrey Sviyazov andreysviyaz@gmail.com
Sorry. Zip file here :)
Best regards, Andrey Sviyazov.
2012/8/6 Andrey Sviyazov andreysviyaz@gmail.com
Hi all.
Here first release of preselector project, schema and BOM. Any questions and suggestions welcomed.
Best regards, Andrey Sviyazov.
2012/8/8 Sylvain Munaut 246tnt@gmail.com
Which spec is that exactly ?
Its about far-near problem (05.05 p.5.1), i.e. blocking by near mobile with high signal at only 600kHz offset. Therefore, calculated that LO noise at 600kHz offset should be around -140dBc/Hz to keep nominal Rx sencitivity for wanted far mobile.
Regarding the GSM 1800 version of the selectivity board, it will not be a problem to change the mixer and the filters. However, are you sure we
will
be able to get the correct matching if we only change the passives ? Are
you
sure we would not need to modify the PCB traces to get a good matching
for a
1800 MHz version of the board ? What do you think ?
Unless I missed something fundamental, this board is only RX side right ?
Yes, it contain LNA, DivSwitch, down converter and IF filter for Rx path only. Late I'll make band dependant boards with 1-2W TxPA and ceramic duplexer. But, possible no reasons to do it, I don't sure yet.
To get any significant loss in RX due to mismatch, you need a very
large mismatch.
Yes.
Best regards, Andrey Sviyazov.
Hi,
Its about far-near problem (05.05 p.5.1), i.e. blocking by near mobile with high signal at only 600kHz offset. Therefore, calculated that LO noise at 600kHz offset should be around -140dBc/Hz to keep nominal Rx sencitivity for wanted far mobile.
Mmm, interesting, I had never thought about this.
Can you confirm my computation is the right one :
Minimum Ec/No of 8 dB, thernal noise of GSM ~ -120 dBm, so the weakest mobile would be a -112 dBm.
The interference I see specified in 05.05 is a -26 dBm sine, so with -140 dBc/Hz you'd get a -166 dBm 'leak' of the interference in the signal of interest at any given position, which you need to integrate over the entire useful bandwidth of the GSM signal ( 271 kHz -> 10 * log10(271e3) = 54 dB ), which would give a -112 dBm noise.
Cheers,
Sylvain
08.08.2012 22:46 пользователь "Sylvain Munaut" 246tnt@gmail.com написал:
Hi,
Its about far-near problem (05.05 p.5.1), i.e. blocking by near mobile
with
high signal at only 600kHz offset. Therefore, calculated that LO noise at 600kHz offset should be around -140dBc/Hz to keep nominal Rx sencitivity for wanted far mobile.
Mmm, interesting, I had never thought about this.
Can you confirm my computation is the right one :
Minimum Ec/No of 8 dB, thernal noise of GSM ~ -120 dBm, so the weakest mobile would be a -112 dBm.
The interference I see specified in 05.05 is a -26 dBm sine, so with -140 dBc/Hz you'd get a -166 dBm 'leak' of the interference in the signal of interest at any given position, which you need to integrate over the entire useful bandwidth of the GSM signal ( 271 kHz -> 10 * log10(271e3) = 54 dB ), which would give a -112 dBm noise.
Cheers,
Sylvain
Yes, your computation confirm that we can get noise flour -109 dBm in summary. But I can't understand this stupid requirement. Really blocker it is mobile signal, therefore it will make direct jumming due to GMSK spectrum mask. What do you think about it?
Best regards, Andrey Sviyazov. (Sent from my mobile client)
Hi all.
I have question about preselector improving: Do we need on-board eeprom for identification? It also can contain band and IF freq values, T'C look-up table for SAW, etc.
Best regards, Andrey Sviyazov.
2012/8/9 Andrey Sviyazov andreysviyaz@gmail.com
08.08.2012 22:46 пользователь "Sylvain Munaut" 246tnt@gmail.com написал:
Hi,
Its about far-near problem (05.05 p.5.1), i.e. blocking by near mobile
with
high signal at only 600kHz offset. Therefore, calculated that LO noise at 600kHz offset should be around -140dBc/Hz to keep nominal Rx sencitivity for wanted far mobile.
Mmm, interesting, I had never thought about this.
Can you confirm my computation is the right one :
Minimum Ec/No of 8 dB, thernal noise of GSM ~ -120 dBm, so the weakest mobile would be a -112 dBm.
The interference I see specified in 05.05 is a -26 dBm sine, so with -140 dBc/Hz you'd get a -166 dBm 'leak' of the interference in the signal of interest at any given position, which you need to integrate over the entire useful bandwidth of the GSM signal ( 271 kHz -> 10 * log10(271e3) = 54 dB ), which would give a -112 dBm noise.
Cheers,
SylvainYes, your computation confirm that we can get noise flour -109 dBm in summary. But I can't understand this stupid requirement. Really blocker it is mobile signal, therefore it will make direct jumming due to GMSK spectrum mask. What do you think about it?
Best regards, Andrey Sviyazov. (Sent from my mobile client)
On Mon, Aug 13, 2012 at 1:08 PM, Andrey Sviyazov andreysviyaz@gmail.com wrote:
I have question about preselector improving: Do we need on-board eeprom for identification? It also can contain band and IF freq values, T'C look-up table for SAW, etc.
Good idea.
-- Regards, Alexander Chemeris. CEO, Fairwaves LLC / ООО УмРадио http://fairwaves.ru
-
Alexander Chemeris -
Andrey Sviyazov -
Sylvain Munaut